Contents


Glenn Gombert ggombert@infinet.com
Mar 1, 1997

I am getting some differences in the image-scales that I have been calculating for some of Tom's Images that are somewhat differnet than those that Michael got when he put togther his TASS Technical Note. I was wondering if anyone had tried to measure these? I am especially interested in this since it directly affects the accuracy of my Astrometic calculation.

The centroiding algorithm that I am using gives better than 1 arcsecond resolution when used on digitized POSS data so I am sure that it is fairly accurate.


Arne Henden aah@nofs.navy.mil
Mar 1, 1997

Glenn G. wrote about image scale calculations. I did an independent measure in my astrometric note, and got 13.73 arcsec/mm on average.

Which POSS digitized data are you referring to when you quote 1arcsec resolution? And what do you mean by 'resolution'? How are you measuring the accuracy of the centroid? Actually, 1arcsec accuracy for centroiding the POSS plates, if you are working from the original plate material, is not as good as the professional groups are getting. For instance, the PMM is quoting 200mas astrometric accuracy over the entire sky, including all of the bending and optical distortion corrections, and local accuracies higher than this. However, we are probably comparing apples and oranges here! Note that photographic plates DO NOT have gaussian-shaped images as they are inherently non-linear. The PMM uses a Fermi-Dirac function for its fitting.

If you want to test your centroiding, get a series of TASS images of the same region of the sky on different days and test the repeatability of a frame of isolated stars within a section about 3 degrees square. I'm sure someone has already done this, and so please accept apologies if I haven't read the proper emails!


Arne Henden aah@nofs.navy.mil
Mar 1, 1997

Chris replied to Tom's request regarding file naming conventions. I'm glad someone had the courage to attack the problem and open themselves up to flak!

My strong preference is to continue with the DOS/ISO9660 8.3 filenames. You have to match the lowest common denominator. Chris has done so.

I would prefer that, in any naming convention, the date is contained in the leading characters. This makes sorting and directory listings easier.

TM3GET has a unique idea in using a modified Julian Date in the file name. This saves space, as a 4-digit day number is valid for nearly 30 years! It is a little harder for a human to identify, so my preference is, if enough characters are available, to use the yymmdd format.

Chris suggested using common 3-character extensions like TXT and FTS for the files. If you are compressing a lot of information into the file name, it doesn't seem sensible to be extravagent in the extension and waste characters.

Therefore, here is another suggested naming convention.

yymmddxy.mnn

yy   year - 1900
mm   month
dd   day
x    site  a-z
y    filetype a-z
m    camera 0,1,2
nn   sequential file number for that camera
Note that yymmdd could be replaced with jjjj (modified Julian Date ala tm3get), leaving an extra 2 characters that could be reserved at this time.

After flatfielding, I do not see further processing of the FITS frames. All steps following extraction will have astrometry tied to them, so I see no need to keep individual extractions separate; rather, a single star list for a given camera for a given night should be made.

So maybe another file format would be:

jjjjxmnn.ccc

jjjj  modified julian date
x     site
m     camera
nn    sequential file number (set to 00 for starlists)
ccc   file type.

Now the file type can be a squandering of 3 characters as suggested by Chris, with perhaps RAW for the raw fits, and RED for the flatfielded fits, and some other extensions for other filetypes.

Chris suggested a 3-digit julian date, which won't work as that only gives you 1000 days before rollover (and it depends on what day you start as to when the rollover occurs).

Just some mutterings. I've been shovelling too much snow lately and I'm afraid my brain has shut down...


Nick Beser beser@aplcomm.jhuapl.edu
Mar 1, 1997

Chris Albertson wrote:

> One thing to watch out for is that you will be running two programs,
> your rsh'ed temperature monitor, and the tm3get11 program.  both will
> be accessing the same hardware.  This is a clasic problem in
> computer science.  You need some scheeme to insure that this does
> not cause problems.  Believe me with no such scheeme you will have
> problems. (one program sets a register a millisecond before
> the next one re-sets it)  One easy way is for the highest priority
> process (tm3get11) to toggel a flag that controls access by the second
> process.
It gets simpler than that. tm3get11 is not running when I do a rsh to monitor tass. After enough time has elapsed, I stop issuing rsh to monitor, and issue a three rsh:
  1. Copy a new tass.rc file with the correct number of lines to collect,
  2. replace the autoexec.bat with one that will launch tm3get11 prior to coming back on line.
  3. Issue a reboot.

Now as far as I understand the configuration (and we have not tested this), TM3GET11 will be running to completion before windows 95 comes back up. There will not be a rpc capability in the system until windows 95 comes back up and starts the rpc task. So there will be no conflicting task asking for status. Since the homepage script is the program that is issuing the rsh for monitoring, we can cut it off until a message is sent from the windows 95 system saying it is up. If the program terminates early, when windows 95 comes up, it will alert the unix system, who will know what time it was expected to come up. The unix system will close the remote doors, and then shut off the power to the TEC.

> I assume the data is showing up on one of the UNIX machine now via
> NFS mount of whatever.  If it is, why not just read the temperature
> out of the raw FITS data file.  It is (should be) recorded in one of
> high numbered columns of peudo-pixels.
At this time, there is no nfs mounted file system or smb file system. The problem is that tm3get11 runs best in DOS mode, and does not run very well under windows 95. I plan to run the program during the boot cycle for windows when it runs stand alone dos utilities.
> > Until we have a Linux version of tm3get11, I plan to copy a autoexec.bat
> 
> The _best_ thing would be to have the driver broadcast the enginerring
> data to any client who requests it.  I have just about finished this
> part of the Linux driver code.  I have not heard from Norman in
> ages.  If anyone does, would they ask him to e-mail me at the address
> below on my sig line.
> 
> P.S. What you have is not (if you use common technical terms) a
> "remote procedure call (rpc)".  It is best called a telnet or rsh
> server as that is the protocol used.  An RPC is used by a
> programmer to call a subroutine or function tha physically resides
> on another machine.
Chris I agree with your definition of the RPC. I am using the Winsock RSHD/95 utility so I should call it a rsh server. I would much rather be using a linux system, but since I am stuck at home recoverying from surgery, I can't get into the lab to write or debug it.


Arne Henden aah@nofs.navy.mil
Mar 1, 1997

First, I want to make a point. Most of my messages seem to be negative, and could be interpreted as downplaying the quality of TASS data. This is in no way true! I wouldn't be commenting at all if I didn't feel that TASS had a place in the scientific community and that the final data quality would not be comparable to many professional sites. I'm sure Richmond and Paczynski feel the same way! There are some really bright people associated with this project, am I'm learning new things every day.

Chris asked about the astrometric accuracy to perform a couple of projects.

  1. To uniquely identify a star from the TASS scans, an accuracy of +- 1/2 pixel is probably good enough (at least to cross-identify scans from different sites and cameras). To cross-identify with the GSC or other catalogs, I would prefer arcsecond-class coordinates.
  2. To track a discovered solar system object. For near-earth asteroids, the current 3arcsec accuracy will often be good enough. Comets, because of their diffuse nature, are always difficult objects to get highly accuracte coordinates. Distant asteroids require higher precision; usually, reported coordinates with greater than arcsec errors are thrown out of the solution.

But TASS is almost there. Factors of 2-4 can usually be made up at this preliminary stage. For example, Tom's camera1 (V) images are far better than the I-band images, and the reported fwhm is poor only because there is trailing on the frames. You may have to stop down the I-band cameras to improve image quality; switch to R-band filters; or use two V-band cameras and one I-band camera. Likewise, once the hardware is adjusted, improved centering techniques using knowledge of the typical TASS camera image profile may increase the astrometric accuracy.

As a professional, I would much rather reported data that had: (1) 0.5arcsec coordinates instead of 3arcsec; (2) 0.01mag photometry instead of 0.10mag photometry. I think both are doable with TASS without major effort.


Tom Droege droege@wwa.com
Mar 1, 1997

It seems to me that a few arc seconds is quite good enough to do the first (easy) part of this experiment. We only have to do 1) below, that is sort the star measurements. Then we can monitor our strip of sky for a year and find lots of good stuff. Then we can write a paper on what we have found, and wait for a shower of money (or a lot of copy cats) to do it better.


Tom Droege droege@wwa.com
Mar 1, 1997

I have been running Mike Gutzwiller's star program on two days of "B" fields. I am very pleased with the result.

Running at 5 sigma, I get pretty much the same stars in all six fields. This is very encouraging to me. The astrometry would appear to be good enough to sort the stars, which is all that is needed to make the first pass at analysis for variable stars.

I like to look at data by hand with a pencil and paper. It allows one to see obvious junk in the data that can be missed until the program gets very sophisticated. So I printed out the first 100 stars in the B field from each of the six measurements and looked at them.

  1. The camera pointing is not ideal. But I don't plan to change it unless you all really beat me up. One can keep fussing and never take any real data. So I am after getting some measurements to meet the June deadline. I hope that after June, we can use the preliminary data to get all lined up and focused properly and then run for a rear for a quality experiment. Then we will throw the Mark IIIs away, and run with Mark IVs if anyone can still stand this project.
    Camera 0       + 0.60   to    - 2.28    degrees declination 
    Camera 1       + 1.41   to    - 1.38
    Camera 2       + 1.23   to    - 1.46
    

    These fields were determined by looking at the extremes of 100 stars so the edges may not be quite right.

  2. The data I was using was marginal at best. The moon was near field B and there was also some thin clouds. I would usually throw such data away, but it is all I have since I tried to point the camera better. Just picking out stars at random and comparing the six measurements I was very pleased.

    The sigma of the four I magnitudes measurements in general was less than the measurement sigma. With my non statistical background this indicates to me that the systematic errors are less than the measurement error, which is where we want to be. This was more true for the faint stars than for the bright ones, which probably suffer from saturation effects.

  3. Data from the two cameras on a particular day matched about as well as data from two different day's measurements from the same camera. This again is encouraging.

If the data gets no better than this, and the analysis gets no better, then we can still do a nice search and catalog experiment. We should find a thousand or more interesting objects. This is just great!


Nick Beser beser@aplcomm.jhuapl.edu
Mar 1, 1997

Bernie Kluga did a little redesign on the sensor alignment hardware for TASS#5. Marty and I are going to try to get him to document what he did and report it to the group. He replaced some of the mechanical linkage with teflon and other hardware so that it would not be temperature sensitive, and would produce a more accurate rotation of the sensor. The knob now does not have a large dead zone, and is easier to align. Now if the weather would clear I can give you examples of the data. The first run we did February 27 showed two of the three sensors aligned pretty well. The problem was that it was cloudy, and the third sensor could not see anything. We really only got about two files of party cloudy data before conditions closed in.

We are hoping that this will take care of the alignment problems we have been having, and will allow us to collect data for the paper.


Tom Droege droege@wwa.com
Mar 1, 1997

> First, I want to make a point.  Most of my messages seem to
> be negative, and could be interpreted as downplaying the quality
> of TASS data.  This is in no way true!  I wouldn't be commenting
> at all if I didn't feel that TASS had a place in the scientific
> community and that the final data quality would not be comparable
> to many professional sites.  I'm sure Richmond and Paczynski feel
> the same way!  There are some really bright people associated with
> this project, am I'm learning new things every day.
Views of a professional are much appreciated. I see you as someone who pushes us to achieve the best possible limit. This is just great. It keeps us on our toes. But I want to point out (again) the real goal of this project. It is to get people interested in all sky surveys, and go develop a pool of software tailored to the job.

The long term goal is to solve the earh colliding astroid problem with a group of amateurs before the government gets around to making it a big project out of it. There are a lot of other possible goals in between.

So I tend to think a little differently about presenting data from the project. What will make people more excited and get more people working on the project? Fussing too much about the last decimal point, though good science, will turn off some of the adventurous types that might want to join a treasure hunt. Thus I think it will be good to early on find a few variable stars and possibly an asteroid or two and then announce that we have done so.

> 2) To track a discovered solar system object.  For near-earth
> asteroids, the current 3arcsec accuracy will often be good enough.
> Comets, because of their diffuse nature, are always difficult objects
> to get highly accuracte coordinates.  Distant asteroids require
> higher precision; usually, reported coordinates with greater than
> arcsec errors are thrown out of the solution.
As a survey project, we just have to detect that we have a moving object. Then others (who are interested and have better instruments for the purpose) can follow up.
> But TASS is almost there.  Factors of 2-4 can usually be made up
> at this preliminary stage.  For example, Tom's camera1 (V) images
> are far better than the I-band images, and the reported fwhm is
> poor only because there is trailing on the frames.  You may have
> to stop down the I-band cameras to improve image quality; switch
> to R-band filters; or use two V-band cameras and one I-band camera.
> Likewise, once the hardware is adjusted, improved centering techniques
> using knowledge of the typical TASS camera image profile may increase
> the astrometric accuracy.
Don't get me wrong. I will certainly encourage us to to the best that we can possibly do. But I would like to go in stages. First discover a few things, then learn to perfect their measurement.
> As a professional, I would much rather reported data that had:
> (1) 0.5arcsec coordinates instead of 3arcsec; (2) 0.01mag photometry
> instead of 0.10mag photometry.  I think both are doable with TASS
> without major effort.
As an amateur, I want to have fun. My temprement is more that of an explorer than that of a precise cataloger. I think science needs both types. The structure of tass (anarchy) is such that any tass member can choose either approach.


Herb Johnson hjohnson@pluto.njcc.com
Mar 1, 1997

A few thoughts on file names for raw TASS files. My first thought is that Tom's current scheme or something very close to it is reasonable enough. We have "bigger fish to fry", and the FITS header can provide a LOT of information: the filename need not do it all. The names need to be reasonably unique and informative.

*> I assume many people are still limited by DOS's 8.3 filenaming
*> restriction. 
Yes, notibly the manufacturer of the cameras (!). Me too.
*> I will propose a convention very close to what Tom used when
*> he posted his data to Storm's incoming directory. 
*>
*> It is still possible to mix up files written by multiple sites
*> but isn't that what directories where made for.
Yes, until you create other directories and use other people's files and so on. I would encourage a filename format that would encourge each filename to be *unique*. Tom's scheme for filename (ignoring extension) would allow this if you replace the "T" with a site or camera designation, a preassigned letter for each camera (no site has two Mark III cameras). That easily supports 26 cameras, adequeate for the next year or two, long enough.

Tom uses the file extension for fractional date: I'm not so comfortable with that. Chris's alternative has some merits in that way, extensions are needed to identify file TYPES. But Tom's use of it for raw images may be an acceptable exception: raw data will have a numeric extension. Even so, programs that DEMAND a .FTS extension (or .fits) may stumble over raw data with the same filename. This can probably be lived with.

*> One last idea:  The file F0789654.TXT would comtain an
*> explaination of the contents of F0789654.FTS but one could
*> supply a file called FX789XXX.TXT to explain all flat files
*> made on day 789.
If such info can be put into the FITS header, it should. A simple program to extract such comments and create this .txt file is easily constructed. Fewer files is better.


Glenn Gombert ggombert@infinet.com
Mar 1, 1997

I have just uploaded to Storm /incoming a file called STARPOS.ZIP which contains the astromertic reduction of the following files:

        File              Star-List
        G0483967.FTS      A0483967.CAT
        G0483977.FTS      A0483977.CAT
        G0493669.FTS      A0493669.CAT
        G0493678.FTS      A0493678.CAT
        G1493669.FTS      A1493669.CAT
        G1493678.FTS      A1493678.CAT
        G1493946.FTS      A1493946.CAT
        G1493955.FTS      A1493995.CAT
        G2483921.FTS      A2483921.CAT
        G2483931.FTS      A2483931.CAT
        G2493623.FTS      A2493623.CAT
        G2493900.FTS      A2493900.CAT
        G2493909.FTS      A2493909.CAT


Michael Richmond richmond@astro.princeton.edu
Mar 1, 1997

There are now two items described near the top of the TASS home page, each chock full o' data.

One is the page containing data from Tom's Feb 1997 collection. You can find the raw images, plus star lists generated by Gutzwiller, Richmond, and Gombert, plus astrometric analysis by Henden, all under

http://www.astro.princeton.edu/~richmond/tass/tom_feb97/tom_feb97.html

The other is a (small but growing) collection of star lists for the special fields "A" and "B" which make up the "Stupendous Man Scan Plan", or SMSP for short. Mike Gutzwiller has provided star lists from 4 different nights, and I'm hoping that Tom Droege will send some data he has mentioned. You can find this data under

http://www.astro.princeton.edu/~richmond/tass/smsp/smsp.html

Again, anyone who has data or analysis on either dataset, please contact me directly so that I can continue to add more material to these pages.


Tom Droege droege@wwa.com
Mar 1, 1997

I have put a dozen files up on storm under incoming. They are

S30t0504.714
        .723
     505.714
        .723
s31t0504.677
        .686
     505.668
        .677
s32t0504.677
        .686
     505.677
        .686
Neither day was particularly good. The moon was right there and so were some thin clouds. These were all recuced with Mike G.s sigma setting at 3. This gets down to mag 13 + which I want to see for political purposes. I think a cut on position will sort things out nicely from the noise. But what do I know about statistics, so I will recompute these with the setting at 5 if you all wish.

I compared one of Mike's files to one of mine by printing out a 100 stars. The two lists look much alike. The stars found on both lists all compared within the stated error (well within 3 sigma).

I continue to be encouraged.


George Turner turner@bigbang.astro.indiana.edu
Mar 2, 1997

> So maybe another file format would be:
> 
> jjjjxmnn.ccc
> 
> jjjj  modified julian date
> x     site
> m     camera
> nn    sequential file number (set to 00 for starlists)
> ccc   file type.
Arne's suggestion is a good one; I have thanked the stars so many times that we decided to use JDs with our RoboScope data. It is a very powerful tool to be able to manipulate large amounts of data in a numerically sequentially way. You've got to keep in mind that computers will be dealing with the data much more than humans in a one on one way Often times a simple sort of filenames gives me the sequence for input data further down in the data reduction pipeline. I would however move the sequence number to after the MDJD to be able to take advantage of this sorting ability; How's about :
 jjjjnnmx.ccc

 jjjj  modified julian date
 nn    sequential file number (set to 00 for starlists)
 m     camera
 x     site
 ccc   file type.
I also have swapped the camera & site because all filenames will have a certain predefined value that has the same meaning at all sites; the site variable will be different to everybody by definition. Simple MJD to local time conversion programs can easily makeup for any handicap resulting from humans not being able to decipher the filename.

Usually when a human wants to look at the data, it'll be because you want to go to that frame to find out what went wrong in the data reduction pipeline. Your data repository will be very much like a library; think about going to the library card catalog, getting the address of the data you want then going to that location and looking at the data. You'll often want to compare data from that point both forward and backwards in time. The simplest way would be to take the MJD of the time the file is opened . Since it's a drift scan, the length of the file will give you some sense of time with each row representing some kind of cosmic tick of the TASS clock.

(With RoboScope running on VMS & Un*x systems we use the full JD out to 7 decimal places with the decimal used as a period in the filename; this gives you 1/100 of a second which is the limit of most computer systems clocks. Setting your clock to that accuracy will be the topic for another discussion.)

Another idea is to use hex numbers for the sequence number; that'll give you 255 numbers instead of 99 and computers can easliy deal with them. Heaven forbid someone has enough trouble in one night with the thing starting & stopping that much; but hey, RoboScope can very quickly generate tremendious amounts data, mail messages, filled disks, overflowed buffers and other computer disconcerns when things don't go right.


Arne Henden aah@nofs.navy.mil
Mar 2, 1997

Looking back over my emails, I think that I am replying too quickly and often without thinking. I am going to make it a policy to hold off 24hrs before answering a mail-list comment/question to give others a chance to voice their equally valid opinions, and to give me time to better compose messages that might be taken incorrectly!

Glenn G. wrote:

>        The centroiding algorithm that I am using gives better than 1
> arcsecond resolution when used on digitized POSS data so I am sure that it
> is fairly accurate.
I replied:
> Which POSS digitized data are you referring to when you quote 1arcsec
> resolution?  And what do you mean by 'resolution'?  How are you
> measuring the accuracy of the centroid?...
This dialog might be construed to mean that I think Glenn's results are poor. He has been getting results as good as anyone else! The POSS comparison is a good reality check, and shows that his centroiding is in the right ballpark. The point I was making is that the POSS material is considerably different than TASS images, and that to refine your centroiding, you need to compare TASS images with other TASS images. However, I am genuinely interested in the questions presented above...they were not meant to badger!

Once the quality of the images has improved (or at least everyone agrees to leave them in one state or another), then I think an investigation into different centroiding schemes could be valuable. The best article I've seen on centroiding techniques is Ron Stone's (AJ,97,1227,1989). We've found that most techniques work well on mid-brightness stars; they differ for saturated objects, extended sources, and stars near the sky level.


Arne Henden aah@nofs.navy.mil
Mar 2, 1997

Tom wrote:

> ...But I want to point out (again) the real 
> goal of this project.  It is to get people interested in all sky 
> surveys, and go develop a pool of software tailored to the job.   
Thanks for properly reminding me of the second letter in TASS! With the results from Tom's posted images, it is clear that you can cross-match star lists, get 'ok' astrometry and photometry, adequate for many projects and personal interests. I forgot to mention the obvious in replying to Chris' comment about what quality of data was necessary -- you don't need high quality astrometry to discover moving objects.

At the same time, if a couple of nights of effort can improve the quality of the images, you win big-time, even for the asteroid problem. For example, changing the 3.4 pixel fwhm to say 1.2 pixel fwhm:

  1. a sharper peak makes star detection easier.
  2. compressing the signal into a smaller image makes psf-fitting and centroiding easier. You have more signal on the image slopes where most of the fitting occurs; the image is more Gaussian in shape.
  3. a smaller image gives better photometry (you include less sky and nearby stars in the aperture) and a fainter limiting magnitude. I would guess you would go a magnitude fainter.
  4. a smaller image means you can work closer to the galactic plane before crowding becomes important.
  5. since a smaller image reduces the fitting error, the other astrometric errors (such as optical distortion and non-great-circle clocking rate) will become more apparent and easier to correct.
I'm not saying spend the next year tweaking parameters, but I am saying you should work for the best images out of the hardware that you can.


Chris Albertson chrisa@wavenet.com
Mar 2, 1997

>   At the same time, if a couple of nights of effort can improve the
> quality of the images, you win big-time, even for the asteroid problem.
> For example, changing the 3.4 pixel fwhm to say 1.2 pixel fwhm:
> (1) a sharper peak makes star detection easier.
One of my questions was "Can you do this?" Assuming the lens resolves 80 cycles/mm at the center and we use 9um pixels.

What _should_ this lens be producing? Remember it is far from defraction limited. The 80 figure is just a reasonable guess but the figure could be between 60 and 100.


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 3, 1997

>*> FWHM in x and y for each.  The software then finds the median FWHM in x and 
>*> y for all the candidates and assumes this is the FWHMs of a "typical" star. 
>*> The software then throws out any of the candidates from the first step 
>*> whose FWHM in either x or y varies by more than 25% from the median.  This 
>*> throws out spurious peaks, **saturated stars**, etc
>
> Does this account for the lack of bright stars (> mag 8 or so) in your list: 
> they've saturated some pixels and are rejected?
Most of the bright stars are rejected because I mark as invalid any pixel at 32767. No peak is allowed to have an invalid value so almost all bright stars are thrown out apriori. The brightest stars detected in an image are in the mag 6.5 to 7.0 range.
>*>> Along these lines, some figure of merit for "best astrometic fit" to all 
>*>> the appropriate catalog *positions* would be useful.
>*>> .... I can probably get you a good
>*>> PROCEDURAL example from the (radio) Ohio (State University) Survey 
>*>> publications in the Astrophysics Journal: let me know.
>*>
>*> A good idea.  There are lots of other figure of merit as well, such as 
>*> constantness of the sky background, etc.  In general in any halfway decent 
>*> image all SAO stars are seen unless they are too 
>*> bright and saturate the CCD or A/D converter.
>
> Hmmm. The implied question by my comment is: how good is your determination 
> of RA and dec for this image? You've used the SAO catalog, and "fit" some 
> of the stars in it to corresponding stars in the image. How good is the 
> fit? (I'm not presuming in this you are using the MAGNITUDES from the SAO, 
> incidently: for example, Michael Richmond does not use this information
> in his location/astrometric algorithms. That would be another kind of 
> "fit", I think.)
Unfortunately the fit to the catalog varies from image to image depending on the vagaries of the star matching routines and the distribution of the brighter stars. If the image and the test section of the catalog overlap only in a corner but enough stars overlap for a good match to occur then the stars in the opposite corner have poorer RA and Dec determinations than usual. Even in these cases though the RA and Dec should be good to about 1/2 pixel.


Arne Henden aah@nofs.navy.mil
Mar 3, 1997

George Turner suggested some good mods to my second cut at filenaming.

I originally suggested using a sequential file number (00-99) in the filename rather than julian day fraction because: (1) I didn't see any need to know the starting day fraction of an exposure since each row in the frame has a different time tag; and (2) it was easier for me to identify the stored file at a later date. If it had a file number of 06, then I knew that there were 5 earlier files. 99 files seemed to be a reasonable number based on the 15-minute intervals in each FITS file.

After thinking about it, I'm not sure that approach is best. The starting julian day number will often change during a night (for Arizona, it changes just before dawn; for Europe, it changes after dusk), and then how do you handle a sequential file number? Perhaps it would be best to use the day number and day fraction like tm3get and Roboscope. Or, alternatively, take the approach that many observatories use: keep the julian date as the starting julian date for the entire scan, even if it rolls over during the scan.

If you use the day number and fraction, then you run out of characters without modifying the extension. For example (Henden v3),

    jjjjfffm.xcc
where
  jjjj  Julian Date -2450000
  fff   JD fraction
  m     camera
  x     site
  cc    file type
or, if you are happy with 14.4 minute granularity for the date (Henden v4),
    jjjjffmx.ccc
or, like George mentioned, you could make jjjj, ff or both into hex or base-26 (a-z) to save a digit or two at the expense of human readability.

I currently like version 4, as it works with the current TASS scanning strategy. It has the limitations mentioned by George (such as when you are starting/stopping with greater frequency than the current 15-minute file intervals), but meets the 8.3 spec. If you find it constricting, then use base-16 or base-26 to encode the date/time.

Our filenaming with FASTT is quite different. We use the year, daynumber, and sequential file number like

     y97d053.105
primarily due to historical reasons and to keep commonality between all of the CCD systems (stare as well as drift scan). Also, FASTT does more than TASS-like drift scans, as it has a shutter and can move in declination. Therefore, a typical night's observing includes a variety of projects, from scanning long strips to looking at single objects as they transit. This means that coming up with a good file-naming convention for TASS is just as new to me as it is to you!


Michael Richmond richmond@astro.princeton.edu
Mar 3, 1997

Following Arne Henden's suggestion, I have created a new subset of the HST Guide Star Catalog for TASS data-munchers to use. It contains one star per 20x20 arcmin grid, all the way around the celestial equator, covering Declinations -4 to +4 degrees. The stars range from 8 < V < 13.

The file has a simple ASCII format, to allow easy access. It contains 26,653 stars, and occupies 1.14 Megabytes. You can download the ASCII version, or zip'd version which is about 336K.

You can find a more complete description on the "TASS Catalogs" page,

and also download the catalog from that page.


Herb Johnson hjohnson@pluto.njcc.com
Mar 3, 1997

I'm at work making comparisons between the various posted lists of identified stars from one of Tom's test images g0483.967 (an I band image at RA 200 to 206, decl -3.0 to 0.0). Richmond will put the results on the Web page as I complete them. They are not fancy, simply "do your stars look the same and are in the same place as my stars?" for one of Tom's many files. I'm a little frazzled as this involves printing sorted lists and matching stars by trial and error; and hand calculations of many little numbers rather than writing programs and using spreadsheets. My sample is small, too: the first 50 or so stars out of hundreds, for one file out of several. So far, I'm looking at position and magnitude for comparison. I hope to uncover any catastrophic problems between these list: a "reality check", not much more.

One question arises in this process: To M. Richmond, M. Gutzwiller, G. Gombert (two Michaels's here force this name convention):

What are you guys using for an absolute magnitude reference in your posted star lists? Richmond I believe is using the Landolt stars, where such appear in the image field. If they don't, what do you do? (My image file of choice apparently has no Landolt stars.). Gutzwiller, are you using the SAO magnitudes - my file of choice has about 40 of them in it, but it was not obvious to me they "appear" in your output lists as matching identified stars. Seems all these are close enought that there is *some* effort to use absolute candles!?!

Simple results:

MAGNITUDE: Richmond offers both "instrumental" and "abs. magnitude" values: I ignore the former for comparisons, he says they differ by a constant from the latter. Gutzwiller and Gombert provide magnitudes. Richmond's and Gutzwiller's seem more consistent (over my sample, within +/- .2 mag) than Gombert's (over my sample, Gombert's varies from about half a magnitude higher, with occasional highs to one magnitude).

POSITION: using Richmond's findings as a reference, Gutzwiller RA varies from Richmonds by +/- .0015 or so; Gutzwiller's dec varies by a similar value roughly. Gombert uses pixel positions only: his X value (column) is 15 pixels higher than Richmond's, his Y value (row) one pixel higher. Additional variance is about .1 pixel in column and a little more in row.

REFERENCE STAR LISTS:

As noted above, for one file the SAO offers 40 stars, the Landolt zero.

More clever calculations of distributions and noting exceptions can be done by the principles, or by me at a later time. (Whew!)


Tom Droege droege@wwa.com
Mar 3, 1997

Here is a man after my own heart. One who actually prints out data and then looks at the numbers and adds things up with a pencil and paper. Well, Herb, there are not many of us left.

I have been doing the same thing. First I printed out a few star lists, then I wrote a QBASIC program which helped with the process so now I can have it find all the data for a star and put it up on the screen. It is easy to see on the screen when you look at the data that the high sigma is due to one funny point from anothe star in the data.

The two of us with our hand methods will keep the big computer program users honest. I have seen such efforts in physics turn out complete nonsense until some of us actually looked at the real data. (Reactions in the target shielding were producing off track events. The big programs were confused, but as soon as one hand plotted a few raw events you could see what was happening.)


Chris Albertson chrisa@wavenet.com
Mar 3, 1997

Does anyone have a good clean and deep image that overlaps one of more of the TASS fields that were posted. The Palomar survey is now published on a set of eight CD-ROMS. Would this work or has the compression algorithm killed the data?

I would like to be able to answer two questions about a list of objects reduced from a TASS frame:

  1. what fraction of the real stars where detected as a function of magnitude or some other parameter?
  2. what fraction of the list "objects" are bogus and do not exist in the real sky.

I know there are other ways to do 1 and 2 (especialy 2) but I'd still want a clean image that goes a mag or two deeper then TASS.


Chris Albertson chrisa@wavenet.com
Mar 3, 1997

aah@nofs.navy.mil wrote (with edits):

>   If you use the day number and fraction, then you run out of characters
> without modifying the extension.  For example (Henden v3),
>     jjjjfffm.xcc
> where
>   jjjj  Julian Date -2450000
>   fff   JD fraction
>   m     camera
>   x     site
>   cc    file type
> 
> or, if you are happy with 14.4 minute granularity for the date (Henden v4),
>     jjjjffmx.ccc
Just a few comments.
  1. The convention of using the last three characters is so old and entrenched that it is not likely to go away. The dot-three part of 8.3 goes way back and pre-dates DOS or even CP/M. I'll bet 11 characters fits real nicely into the word size of some very old computer. I think we are stuck with "ccc".
  2. file names only have to be unique within a directory. What will be mixed in one directory?
  3. I picked jjjfff as the unique id because it is easy to derive from the filename written by tm3getxx. You just drop one "j". I can write a simple script to do this but more complex mapping require a program to do the filemane changes from tm3get's convention. Also the transformation should be computable in one's head both ways. I can handle dropping a J put base 16 I can't.
  4. the file namming convention need only apply to data posted in a public place. Once it is on my machine I can rename it using longer file names.
  5. I still think you need a tag to specify content. D for Dark frame, X for eXperimental, P for Program of whatever.
  6. no one will follow a standard if they do not like it. So all we can do is make suggestions and if they are good ones people will accept them.


Glenn Gombert ggombert@infinet.com
Mar 4, 1997

I can probably come up with an ST-6 image taken with a 16 inch F/4.5 Newtonian that might be useful for comparison purposes. I will check my "image archive" and see.

The "Real_Sky" data takes up a LOT of space ~15 megabytes for a 1 degree square field (thats the largest area the software will let you select).

You have made an excellent suggestion, one that we should try to make some measurements/comparisons on...


Nick Beser beser@aplcomm.jhuapl.edu
Mar 4, 1997

> Does anyone have a good clean and deep image that overlaps
> one of more of the TASS fields that were posted.  The Palomar
> survey is now published on a set of eight CD-ROMS.  Would
> this work or has the compression algorithm killed the data?
I think I can answer your question about compression on the Digital Sky Survey. The algorithm was designed by Rick White at Space Telescope Science Institute. It is the hcompress algorithm using the haar wavelet. Rick did a study that evaluated different compression ratio's. There is loss, but in the region that we are working, I don't think the loss will be noticable. The DSS plates are digitized to 25 microns at a precision of 15 bits. (I am reading from the Guide Star Sampler), The Digital Sky Survey Home Page says that the first generation was 1.7 arc seconds, and the second generation plates were 1 arc seconds. The compression algorithm distorts this somewhat, but if the data were reduced to the accuracy that we see with TASS, the effect of compression should drop out. The compression algorithm gives higher accuracy to low pass data (reduced resolution of the image) (that assumes that hcompress works like a conventional wavelet.)

Rick White did a paper on hcompress (available on the net) that references some studies on image quality for Astronomical data. His conclusion was that Astrometry is hardly affected by compression of modest ratio's (20:1), and degrades for higher ratios.

The DSS is available on the net, however, it is in very small field of view. Do you have access to the actual CD-ROM's?

in a later message

The URL to the Digital Sky Survey is:

http://stdatu.stsci.edu/dss/dss_form.html

There are other methods for retreiving the data. Take a look at xephem 3.0. I have not been able to get it working yet, but it is supposed to read fits files, and retreive the DSS of the matching area.


Arne Henden aah@nofs.navy.mil
Mar 4, 1997

The first half of my March observing run occurs this weekend, and so far long-term forecasts look good. My intent (along with other projects) are to provide 3 photometric sequences in each of the SMSP-A and SMSP-B fields. The ones in SMSP-B are at:

       8:49:00  +01:15
       8:52:50  -00:10
       8:50:00  -01:15
And each is 11 arcmin square. These sequences will have +-0.01mag errors at V=17, and so should be good for both additional calibration stars over and above the available Landolt standards, and for determining limiting magnitude and precision at faint magnitudes. I will provide the data at both the 12arcsec aperture size that Landolt used, as well as a huge 50arcsec aperture size that the current TASS images seem to require.

The SMSP-A sequence centers have not been determined, but they will be at roughly the same relative location (but centered on known variable stars). Note that the CCD images for these sequence fields can be obtained from me after Monday if you want high-quality images in the SMSP fields, and I can probably provide CCD images from FASTT for the SMSP-A field at any time (I just have to restore them from a tape archive).


Arne Henden aah@nofs.navy.mil
Mar 4, 1997

> Does anyone have a good clean and deep image that overlaps
> one of more of the TASS fields that were posted.
Chris, do you mean one of Tom's fields? I have FASTT scans covering most of these fields (except g1493946 and g1493955) that go to R=18 or so. They are stored as 2048x2048x16bit FITS images (8MB each, 50x50arcmin). The resolution (3 arcsec artificially blurred seeing) may be poorer than the Digitized Sky Survey, but the relative photometric accuracy will be better. What do you want to do with the 'clean and deep image'? If you just want something pretty, then the Digitized Sky Survey will be better.


Herb Johnson hjohnson@pluto.njcc.com
Mar 4, 1997

Key to authors:

*>> Herb
*>  Mike Guztwiller
*>> What are you guys using for an absolute magnitude reference in your 
*>> posted star lists?
*>
*> The 8.5 and 100000.0 were determined to give approximate matching of the 
*> calculated magnitudes and the SAO catalog.
OK, thanks for the details.
*> At this point a more interesting question might be how the 
*> relative magnitude 
*> varies between the lists.  If all the programs were consistent then the
*> magnitudes calculated between two lists should differ by a constant for all 
*> stars.
Lacking at the moment a reliable and extensive absolute reference list, and seeing as we need to present findings of position and magnitude for this AAS paper(s), I first chose to look at self-consistency among the lists. The SAO list gives only 40 candidates, but I'll try them next as a reference: at least they'll force me to look over different parts of the image.
*> Since Michael R. is using the GSC and I'm currently using the SAO the 
*> varience [in position] between our lists is expected.  I plan to use the
*> GSC or possibly the USNO-SA1.0 instead of the SAO soon.
...such radical changes being the reason I don't spend more days on a more exhaustive review. If I can automate it I might. Meanwhile, is the varience *small enough* that it can all be attributed to differences in lists (and different PSF, and dark and flat correction, etc. etc...)?

All in all, I'm still a little *amazed* that there is not a better established list of reference stars. Having been around astronomy for awhile, I'm not *surprised*, however.


Arne Henden aah@nofs.navy.mil
Mar 4, 1997

According to my notes, the pixel size is given by

       scale (arcsec/pix) = [206.265*p]/f
where p=pixel size in microns, f=focal length in mm for the 135mm lens and the 9micron pixels, this gives a scale of 13.75 arcsec/mm assuming 135.0mm focal length and 9.000micron pixels (neither of which is probably true, and probably not constant between the lenses and/or V-I focal points).

Using this, the drift scan rate on the equator is given by

time (sec/pix) = 86164.09 (sec/sid day)
                 -----------
                 [[360*3600 (arcsec)]/13.75(arcsec/pix)]
or 0.9142 seconds/pixel. This changes to 0.9143 at 1 degree, and 0.9147 at 2 degrees off the equator.

Tom was getting 0.9178 for his Feb12 scans, considerably different than the numbers above. I trust that the pixel size is probably very close to 9 microns, so my guess is (and assuming Tom has done his arithmetic right) that the true focal length of the lens is more like 134.5 mm.

What are the scan rates that other sites are coming up with?

Also, a related site-specific question: for those sites actively scanning, what fwhm are you finding for the 3 cameras? I'm curious as to whether the 3.4 pixel fwhm from Tom's February data is comparable to other sites.


John Gwinner gwinner@northnet.org
Mar 4, 1997

> I'm still learning this terminal emulator at home, so sorry
> about the errors in the last mailing!
>   Continuing, with corrections:
Pardon me for jumping in; my first post so there's a brief intro at the end.

Anyway, my basic question with regard to the software your running is why stop with eight characters?

If you're using Win95, DOS itself (exit to DOS) will handle long file names without any problem. I don't run Linux (we use NT in the shop for a variety of reasons) but I would think it should be able to handle long file names.

As an introduction: I'm a software developer with an interest in Astronomy. My major strength is 3D real time graphics on PC's. I have to admit I haven't been keeping up with the list, more or less just download a few notes now and then.

I have thought about taking a look at some of the software mentioned on the list .. I use a very fast DBMS system (CTree) that could help with some of the routines I think, but I'm pretty busy at this point getting my next software contract.


Michael Richmond richmond@astro.princeton.edu
Mar 4, 1997

Since several people have asked for a "deeper" image of one of Tom's fields, I've created some GIF pictures from the Digitized Sky Survey CD-Rom. You can find them on the page dealing with Tom's dataset,

http://www.astro.princeton.edu/~richmond/tass/tom_feb97/tom_feb97.html

The GIF images are a portion of the field visible in images g0493946 and g2493909, two I-band images at high galactic latitude.

If people find this useful, I can make available the FITS data for this area, and/or create a GIF image for an area covered by some V-band scan in Tom's set.


Michael Richmond richmond@astro.princeton.edu
Mar 4, 1997

Bohdan Paczynski kindly provided a subset of the General Catalog of Variable Stars, selecting stars with declinations (1950) between -3 and +3 degrees. I have slightly re-formatted the information, and precessed the coordinates to equinox 2000. You can download this ASCII text file containing 386 stars from the "Catalogs" sub-page under the TASS home page:

http://www.astro.princeton.edu/~richmond/tass/catalogs/catalogs.html

I have heard that the positions in the GCVS aren't very precise, so don't be surprised if they aren't where you expect...


Nick Beser beser@aplcomm.jhuapl.edu
Mar 4, 1997

> There is an ADC on the Data Board and a DAC on the Control Board.
> It is not necessary to load the DAC unless you are trying to change the
> set point of the temperature controller, or the frequency of the
> drift scan.
> 
> It sounds like Norman's routine will do just what you want.  However,
> I notice that Norman's program does not set the temperature DAC until
> it completes the clear process and starts ramping up.  What I do at
> home is to start Norman's program, and wait until it completes the
> clear.  Then I abort the run and delet nnnnnn.tm3.  Then after things
> have stabalized, I start the run again.
> 
> It looks like you will have to find the place in Norman's code where
> he loads the temperature command value into the DAC.  Then you can load
> it up.  Or you can just do what I do and after a power on start and restart
> the program.
> 
> Note the DAC will hold it's set point until the power is turned off.  With
> the VCO not very good, it may pay to keep the computer on all the time.
You raised some interesting points. Let me go over the way we will operate the system:
  1. The windows 95 system is up (all the time), and it runs a remote shell routine that will let me start a DOS task from the unix computer.
  2. The unix computer will start a remote task that will set the temperature DAC, and sample the temperature and voltage.
  3. At this point I really want to monitor temperature to see if we are in trouble. Is there any harm constantly resetting the temperature DAC, or should I leave that alone and every minute, run a task that just reads the temperature and voltage?
  4. At 1 hour or so later, (or when it is dark), I will command the unix system to open the doors to the telescope.
  5. At the same time, I will down load a new autoexec.bat and tass.rc file that after a reboot will cause the tm3get11 program to start. The tass.rc is loaded so that it ends at a expected time.
  6. After the doors are open, I will remote shell a command to the windows 95 system causing a reboot (I have a dos executable that reboots). The reboot will launch tm3get11, data collection will continue until the number of scans programmed in tass.rc has been completed.
  7. I will either periodically ping the system (it should not respond until windows 95 comes back on line). As an alternative, I will try to execute a remote shell, that should not work until windows 95 comes up. That will be my signal that the program has finished. The program has failsafe measures that shut it off if the temperature goes bad. I may be able to tie a serial I/O line into the system to send critical data back to the home page system. Has anyone tried to run tm3get with I/O stored on a NFS or SMB (samba) system?
  8. After the system comes back up, I will monitor the temperature and time. If the time limit has expired, I will close the doors, and then shut off the power. I could do the same if it returned early (monitor temperature to see if there is a problem, and shut down).
Any comments?


Nick Beser beser@aplcomm.jhuapl.edu
Mar 4, 1997

> > 7. I will either periodically ping the system (it should not respond
> > until windows 95 comes back on line). As an alternative, I will try to
> > execute a remote shell, that should not work until windows 95 comes up.
> > That will be my signal that the program has finished. The program has
> > failsafe measures that shut it off if the temperature goes bad. I may be
> > able to tie a serial I/O line into the system to send critical data back
> > to the home page system. Has anyone tried to run tm3get with I/O stored
> > on a NFS or SMB (samba) system?
> 
> Could you place a program in the Win95 "start up" folder that would
> notify the UNIX system when Windows has come back up?  Maybe even
> place an rsh command in the startup folder?
> 
> Did you ever try to run tm3get under dosemu with Linux?  Just wondering
> if it would work.
I will have to look into a rsh from Windows 95 to unix, that would be easier and cleaner.

I tried running tm3get11 under a dos window in Windows 95. It couldn't keep up. I don't have the tass control card installed in a linux system yet (I am waiting on the linux version of tm3get11 before I install linux on the TASS computer.)


John Gwinner gwinner@northnet.org
Mar 5, 1997

> You are right.  Every major operating system being sold today can handle
> long filename.  Problem is some people are still using older systems.
Well, we could also backdate the software for 286 computers too :-) My bias: I'm an upgrade bigot, at least you'll know that from the start!
> Also many DOS and even windows programs, cannot accept long filenames
> even though the OS can.
True, 16 bit programs will not; 32 bit programs will, unless special effort is made in the software to screw things up.
> If it where me making decisions I'd use longer more desciptive names on
> the archive and let people with 8.3 systems rename the file _after_ they
> are downloaded. 
That would work .. changing file names is required for Web work anyway.
 
> Is this database Free or covered by the GNU or GPL style lincense?
> That is the problem with many DBMSes.  If you require our users to
> buy a DBMS product you code is not so popular.  
True. I guess it depends on what we're doing; I had in mind compiling the routines in and distributing the compiled code, but if we are all compiling at home, then we'll need something free of course.

I just couldn't find anything that was free and sufficiently high performance, so that's why I went with CTree. There may be other solutions, I was just brainstorming.

Because I've got a dark sky site and, due to work, lot's of computers standing around, I've thought I could be a good site for a TASS setup after things get settled down. I'm in Upstate NY. However, we've had three days of starlight since December :-(


Arne Henden aah@nofs.navy.mil
Mar 5, 1997

>One of my questions was "Can you do this?"  Assuming the lens
>resolves 80 cycles/mm at the center and we use 9um pixels.
9um pixels is 110 pixels/mm. 80 cycles/mm means 160 resolution elements/mm, so roughly the resolution should be about the same as a pixel. Since seeing is <14arcsec, then theoretically, the TASS images should have fwhm of <1pixel. I don't know for sure what the quality of the Aubel lenses should be. One way of checking would be to take a lens of known quality (something that has been lab tested in Modern Photography or something) and put it on one of the TASS cameras.

The 135mm f/2.8 should be about 50mm in diameter, which gives a diffraction limit of

       difflim (arcsec) = 0.252 * lambda(nm)/D(mm)
or 2.5arcsec for the V bandpass. So you are right: if the resolution is 100 l/mm, then the lens is not diffraction limited. That is ok, since the pixels undersample the diffraction pattern anyway.

Bottom line: I still think you should be able to get 1.2-1.5 pixel fwhm with the TASS cameras.


Arne Henden aah@nofs.navy.mil
Mar 5, 1997

Chris wrote:

> (1) The convention of using the last three characters is so old and
> entrenched
> that it is not likely to go away.  The dot-three part of 8.3 goes way
> back and pre-dates DOS or even CP/M.  I'll bet 11 characters fits real
> nicely into the word size of some very old computer.  I think we are
> stuck with "ccc".
I think we are stuck with 8.3 because of the ISO9660 standard for CDROMs (I know, there are extensions, but lowest-common-denominator again) and DOS x.x. I am not so certain that the dot-three needs to have any special format as each program has developed their own extension. Witness WORD with .doc and .rtf; AUTOCAD with .dxf, etc. There is nothing unique about these three characters, just the 8.3 format itself.
> (2) file names only have to be unique within a directory.  What will be
> mixed in one directory?
The problem arises when datasets are combined from different sites. George touched on this. For example, it may make more sense to keep the merged (but raw) nightly star lists separate from the photometrically calibrated star lists, so the directory tree may look different at the reducer's site than at the camera site.
> (3) I picked jjjfff as the unique id because it is easy to derive from
> the filename written by tm3getxx.  You just drop one "j".  I can write a
> simple script to do this but more complex mapping require a program to
> do the filemane changes from tm3get's convention.
Note that George Turner's mods and my mods and your suggestion are all similar. George and I moved the julian date to the front of the filename for sorting ease. We use a 4-character julian day number so that you don't get rollover as quickly. Also, why use a script to do the mapping -- there is nothing preventing the modification of tm3get to output the agreed-upon file name.


Arne Henden aah@nofs.navy.mil
Mar 5, 1997

Michael noted that the coordinates for the General Catalog of Variable Star objects are rumoured to be not very precise. My findings, in looking at several hundred of these stars over my career, is that the coordinates are within 20arcsec or so of the true position about 95 percent of the time. There are some real outliers, but they are rare.

What is more common is misidentification, both in terms of whether the star is even variable and in determining the variability type. Do not believe that a star listed as an eclipsing binary is a binary...it could equally easily be an RR Lyrae or Cepheid variable. Periods are often aliases of something shorter.

Also, note that the GCVS has 386 stars in the +-3 degree zone. We found in our variable star survey a statistical 4000 variables in 1/10 of that zone, so there are about 100 times as many variables in the +-3 degree zone as are currently known.


Herb Johnson hjohnson@pluto.njcc.com
Mar 5, 1997

I was confused by Nick's recent message about Win 95, restarts, changing autoexec.bat, etc. etc. until I saw this:

*> 3. At this point I really want to monitor temperature to see if we are
*> in trouble. Is there any harm constantly resetting the temperature DAC,
*> or should I leave that alone and every minute, run a task that just
*> reads the temperature and voltage? 
Aha! Some time ago I suggested to Tom et al that an overtemperature circuit, a single chip, can detect excessive Peltier temperatures and cut them off before they cook themselves. I've sent Nick the details for further consideration. I'd do it myself, if I had a camera in hand.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 5, 1997

I want to make it clear that the reason we are playing with rsh, windows 95, and restarting is not to monitor temperatures remotely. We want to run the data collection remotely. An ideal solution is a linux version of tm3get11. What I am setting up for the APL TASS camera is a stop gap procedure that we will throw out after the linux version comes on line.

The Tass Monitor task will let one unix computer monitor a DOS machine prior to a data collection run.


Arne Henden aah@nofs.navy.mil
Mar 5, 1997

For those of you interested in detecting asteroids in the TASS frames, especially in the SMSP field A, there is an interesting paper by D. L. Rabinowitz, "Detection of Earth-Approaching Asteroids in Near Real-Time," 1991 AJ 101,1518. He describes the algorithms used by the Spacewatch telescope to detect 'streaks' (fast motion in a single frame) and 'motion' (slower movement between frames). He also gives an indication of the magnitude of motion you should expect for asteroids at different distances.


Benoit Schillings benoit@be.com
Mar 5, 1997

I played a little bit with different 135mm f/2.8 lenses on my KAF-400 based camera and found that most lenses are of very bad quality when used without narrow band filters... they generally show a large blure around a pretty sharp core... my guess is that is comes from chromatic abberations... I took some images using one of the lens that Tom uses and found that It will give some very good images when used with a narrow (10nm) band H-Alpha filter... I still think that a high quality lens could improve the detection by nearly 1 magnitude...

For reference, the fwhm with the 10 nm filter was 1.2 pixel.


Arne Henden aah@nofs.navy.mil
Mar 5, 1997

> All in all, I'm still a little *amazed* that there is not a better established
> list of reference stars. Having been around astronomy for awhile, I'm
> not *surprised*, however.
All-sky photometry at the 0.01mag level is truely difficult. That is why there are only a handful of astronomers pursuing the task of setting up standards. Space missions, where you get above the atmosphere, are the obvious solution, and the Hipparcos/Tycho results are anxiously awaited (even though they didn't choose true B & V passbands!).

The solution for TASS, as mentioned earlier, is to set up our own reference frame, tied as closely as possible to the Landolt standards. Then, hopefully, only a zero point will be necessary to compare TASS results with other observers in the future.


Glenn Gombert ggombert@infinet.com
Mar 5, 1997

While I still had the data on the hard disk I went through and calculated the "average" FWHM of the detected objects in each of Tom's images. Here are the numbers that I came up with:

        File:           SkyBackground(ADU)     RMS Sky        FWHM(Average)
        G0483967.FTS    11167.6                55.4055        3.61
        G0493669.FTS    11698.1                97.5382        3.88
        G0483977.FTS    10926.8                49.9414        4.02
        G0493678.FTS    11753.2                95.8544        3.81
        G1493995.FTS    11439.9                52.8192        3.62
        G1493669.FTS    13155.8                63.7128        2.66
        G1493678.FTS    13154.2                62.4993        2.91
        G1493995.FTS    11705.8                52.7644        3.44
        G2483931.FTS    99972.7                45.9182        4.23
        G2493623.FTS    12642.9                96.1175        5.49
        G2493900.FTS    10704.7                49.7170        4.08
        G2493909.FTS    10772.3                50.5891        4.10

I will try and repeat the same measurements from some of the data taken here in Dayton over the weekend.


Allen Gilchrist gilchris@ipxpress.aws.waii.com
Mar 6, 1997

I have seen the same thing using standard 35mm camera lenses on a Cookbook camera. I don't think the problem is limited to cheap lenses. One of my lenses is an Olympus lens that ought to be pretty good. The problem is that the designers never considered having to color correct these lenses out to 10000 Angstroms.


Arne Henden aah@nofs.navy.mil
Mar 6, 1997

Benoit wrote that he achieved 1.2pix fwhm with a 10nm filter, which sounds reasonable to me. The 1.2pix number comes from the fact that the error budget for imaging is less than one pixel, but you _never_ measure fwhm that is only 1 pixel; it always gets spread a little.

Chris wrote:

> Why is this?  My 80 l/mm estimate was based on tests done with 
> photographic film.  The film's spectral sensitivity acts like Benoit's
> narrow band filter.
I thought photographic film was polychromatic and therefore does _not_ act like a narrow band filter. Lenses are usually achromatic and, like Chris said, only perfectly corrected at two wavelengths, but these wavelengths are chosen so that the lens is almost perfectly corrected throughout the wavelength range of color film. Allen wrote:
> I have seen the same thing using standard 35mm camera lenses on a Cookbook
> camera.  I don't think the problem is limited to cheap lenses.  One of my
> lenses is an Olympus lens that ought to be pretty good.  The problem is
> that the designers never considered having to color correct these lenses 
> out to 10000 Angstroms.
Allen, were you talking about using the lens unfiltered, or at least without a red blocker? Then I can fully understand poor image quality. Allen makes the very valid point that photographic lenses are not color-corrected at 1000nm. In fact, the 'IR' marking on most of these lenses refers to 650nm or so.

My guess on this is as follows (and I mentioned it before):

     (1) the camera with best focus is #1 (V).  For Tom, it was 2.8pixels
         fwhm or so.  Michael R. told me earlier that he has seen 1.8pix
         fwhm on his camera1 without careful focussing.  No one else has
         given me any image quality data for their TASS cameras, so I
         don't know the spread in this measurement.
     (2) the cameras with worst focus are the I bandpass.  Few camera lenses
         expect the user to have film sensitive to 800nm light, and I doubt
         they correct their lenses out here.
     (3) Benoit showed that the Aubel lenses work ok at H-alpha (650nm),
         essentially at the R bandpass.
Therefore:
     (4) I think better focus can be achieved at V than you are currently
         getting.  You may want to try an experiment of using a red blocker
         with the V filter to remove any hint of long-wavelength aberrations.
     (5) I think for the 'I' bandpass you are either going to have to stop
         the lens down to remove some of the aberrations, or else switch
         to an R filter to get back into the 'corrected' range of the lens.
Stopping the lens down, using a red blocker and using an R filter should be easy tests that don't have to be done by all sites. Paying careful attention to focus should be done by all sites, preferably on cameras pointed at the equator and then only for the middle columns.


Herb Johnson hjohnson@pluto.njcc.com
Mar 6, 1997

*>  All-sky photometry at the 0.01mag level is truely difficult.  That is
*> why there are only a handful of astronomers pursuing the task of setting
*> up standards...
*>  The solution for TASS, as mentioned earlier, is to set up our own reference
*> frame, tied as closely as possible to the Landolt standards.  
A "wholesale" approach might be to use a large set of stars that, in the aggregate, have sufficient precision in magnitude. I suspect, however, that the distribution of errors is not uniform, and of course it begs the question to determine it. I've noticed that some star catalogs are many decades old: would an average (mean? mode?) of readings across the decades, hopefully at similar passbands, provide a better value for a given star; or at least remove some long-term variables?


Michael Richmond richmond@astro.princeton.edu
Mar 6, 1997

> I've noticed that some star catalogs are many decades old:
> would an average (mean? mode?) of readings across the decades, hopefully
> at similar passbands, provide a better value for a given star; or at least
> remove some long-term variables?
No, I don't think so. For the great majority of variable stars, the amplitude of variation is small enough that systematic differences between different stellar catalogs would swamp any true variability. That is, a catalog made in the 1930s, based on photographic measurements, and a catalog made in the 1960s, based on photoelectric photometry, will have effective passbands which are far enough apart that individual stars which DON'T vary, in truth, will still appear to have different magnitudes, simply due to the passband mismatch.

A few variables (like Mira) _do_ change be enough that they would still stick out -- but not many.

We've got to make our own photometric catalog, I think.


Peter McCullough pmcc@astro.uiuc.edu
Mar 6, 1997

On Feb 19, 1997, Tom Droege wrote:

> Hmmmm!  What does it mean when the last page of the last
> issue of a magazine says you are an arrogant maverick?
> I like it!  It comes very close to meeting a particular
> goal of mine. - Tom Droege
From that I was left with the impression that the article was uncomplimentary (although T.D. felt it was a compliment). In fact, the article was a compliment in the traditional sense too. Below I attach the paragraph for those without CCD Astro Magazine.

A quote from CCD Astronomy's Final Issue (last page, by Leif J. Robinson):

>  ...The Amateur Sky Survey (TASS)...aims to keep 1,000 square
> degrees of sky under continuous electronic surveillance to
> magnitude 15. (You can find out more at 
> http:/www.astro.princeton.edu/~richmond/tass/tass.html.) Comets
> and variables stars are obvious prey, but it's the promise of
> the unforeseen -- gamma-ray bursters,perhaps -- that excites
> my gray cells. I'm sure [Fritz] Zwicky would have loved the 
> arrogance of this observing project. 
So it wasn't T.D. that was said to be arrogant, but the TASS project itself, and in a very complimentary, positive way.


John Gwinner gwinner@northnet.org
Mar 6, 1997

Folks, forgive a really dumb question:

is there a reason that TASS doesn't just use some kind of RFT Newtonian? A wide angle Newtonian, with the imaging chip placed at the focal point (non-visual) would be fairly well color corrected, wouldn't it? I'm thinking that a short F ratio conventional Newtonian would have less contrast due to the size of the secondary, but you could reduce the secondary by putting the CCD in the tube, couldn't you?

If it's a dumb question, flame me privately and I'll go away .

Don't get me wrong, so far, the project sounds really important and superbly run. I hope to eventually be able to provide a site, as dark skies and computing hardware is in abundance here, but so far we've had only about four clear nights all winter :-( I'm in Upstate NY and lake effect clouds seem to be the norm in the winter (this one's been particularly bad).


Mike Schwartz pfactors@ix.netcom.com
Mar 6, 1997

>Sometimes my sense of humor gets me into trouble.  ;^)
>
>I did not really mean to mis-quote the article.  

Hi Tom -

I've been watching on the sidelines for a long time but have been completely absorbed with my SNe studies. But I thought that the CCD comment was the ULTIMATE complement. TASS is representative of true 21rst century agressive astromony done with pure guts and dedication and little money. If nothing else comes of it other than the proof that such cooperative effects can be done, then you have been an immense success. You should be very proud that you have had the capabilities to keep this moving and flowing. Not many people have these qualities. Congratulations!


Tom Droege droege@wwa.com
Mar 7, 1997

Looks like I again got a bunch of pretty good data last night.

Boy, it is a struggle to get it processed and up in a timely fashion.

I have been trying to put the star lists together and make sense of them. One thing is clear, however bad the astrometry is, it is good enough to match stars. The photometry is the big problem.

Some runs are much better than others when you compare the photometry. Out of the seven runs I was studying for the V filter, three were pretty good matching to about 0.1 mag at mag 12, three were so so, matching to about 0.3 mag at mag 12, and one had 1 mag variations. I purposefully don't say what I mean by "matching" above.

It is obvious we need lots of data. It is also obvious that we have to adjust/select each star list as is is "added" into the "sum".

This all takes a lot of discipline. More than I have, since I am doing a lot of other things. I will concentrate on taking data and hope one of you can make good sense of it.

I encourage everyone else to take all the data they can. Particularly the B field before it goes away (soon).


Tom Droege droege@wwa.com
Mar 7, 1997

All questions are acceptable. It is not easy to say what is dumb.

The problem with the Newtonian is that if you have 135 mm focal length, then you you get only a 34 mm dia lens at f/4. Putting a 3" diameter CCD camera in the tube does not leave much light gathering ability. ;^) At f/4 you have a chance to get the camera outside the tube, so it can be done. But you can't begin to match camera lenses for possible f/ ratios.

The problem all comes from the size of available CCD's and the desire to cover a large angle in the sky. To cover 3 degrees (and we want to cover more per camera) the focal length is determined by the radius that forms a 3 degree angle with the CCD width. This is close to the "standard" 135 mm focal length telephoto lens.

There are other problems with reflectors. If we had big CCDs, then it would still be hard to get big reflectors because it is hard to get a large flat field. This can possibly be done with Schmidt cameras, and we are exploring that. But at the moment, the cheap and easy solution seems to be camera lenses.


Arne Henden aah@nofs.navy.mil
Mar 7, 1997

Tom and Mike G. so far have been gathering data on SMSP-B and posting that data on Michael R.'s computer. Both of these sites have been using Mike G.'s extraction software.

I would like to encourage these sites to archive their raw scan data. There is still some discussion regarding how to properly flat-field these frames, and much discussion as to the proper way to extract the data and apply astrometric and photometric corrections. If the raw data still exists, then we can go back and try different techniques.

Also, regarding Mike G.'s package, I would request that he add a couple lines of code and output the fwhm in x and y for each detected object. I know that he applies some shape parameter cuts before accepting an object as real, but there will be many blended objects that pass such tests and adding shape information to the output file will help us in determining how to handle such objects.


Allen Gilchrist gilchris@ipxpress.aws.waii.com
Mar 7, 1997

I think the reason for using the 135mm lenses was one of cost and field-of- view. I don't think it would be possible to buy mirrors (or more correctly, mirror-lenses) of that focal length. You may get a better answer from some of the other guys on the list. Tom?


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 7, 1997

I have been enhancing the "star" program and in the process discovered a systematic error that, for once, is actually good news. I had created a catalog for the "star" program based on the mini-GSC files Michael Richmond created a while back. This was for both astrometric and photometric considerations. Hopefully the astrometric values are better in the GSC and since it went deeper in magnitude I could match more stars.

The new version of "star" will adjust the magnitude scale of the list to best match the catalog. While testing the software I found the amount of shift in magnitude increased as the catalog I was using went to fainter magnitudes in the GSC. By the time the catalog got to mag 12 the shift stopped moving though and the shift was the same for a mag 13 catalog and the mag 14 catalog. Looking at the data by hand I discovered that brighter stars were consistently measured below their GSC values. To illustrate this effect I have uploaded an Excel spreadsheet to:

storm.fnal.gov/incoming/graph669.xls

The graph is a scatter plot of GSC magnitude vs. the difference in magnitude between the star list and the GSC magnitude in Tom's g1493669.fts image taken to 3 sigma. The graph shows a flat distribution for all stars between mag 9.0 and mag 13.5. A notch appears beyond mag 13.5 when the limiting magnitude is approached. From the graph I estimate the limiting magnitude at 14.5 (50% efficiency).

For brighter stars (mag < 9.0) a systematic error begins to take effect so that the magnitude error increases as the GSC magnitude decreases. The effect is such that a matched star with a GSC magnitude of 7.0 is typically measured at 8.0 in the list.

I believe what's happening is that bright stars which have TASS V or I magnitudes lower than the GSC magnitudes (B?, P?) are preferentially selected since they are bluer and do not saturate the A/D converter as easily. ("star" ignores any star which has a saturated core.)

The end result is that the magnitudes published by the "star" program based on its (100,000 ADU = mag 8.5) baseline are about 1 magnitude too low. I now estimate the TASS 3 sigma limit to be near mag 14.5 as a result.


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 7, 1997

Arne wrote:

> Tom and Mike G. so far have been gathering data on SMSP-B and posting that 
> data on Michael R.'s computer.  Both of these sites have been using Mike 
> G.'s extraction software.
>  I would like to encourage these sites to archive their raw scan data.
> There is still some discussion regarding how to properly flat-field these 
> frames, and much discussion as to the proper way to extract the data and 
> apply astrometric and photometric corrections.  If the raw data still 
> exists, then we can go back and try different techniques.
I have archived all my runs on tape going back to October 1996 and will continue to do so. I too believe in keeping the raw data including dark images for reanalysis with better techniques. This tendency probably goes back to my high energy physics days where teasing 20 events out of 3 months of data was considered a horn of plenty for my thesis.
>  Also, regarding Mike G.'s package, I would request that he add a couple
> lines of code and output the fwhm in x and y for each detected 
> object. I know that he applies some shape parameter cuts before 
> accepting an object as real, but there will be many blended objects 
> that pass such tests and adding shape information to the output file 
> will help us in determining how to handle such objects.
I cringe whenever somebody asks for "a couple of lines of code" but in this case it actually won't be more than about 10 lines since I already calculate FWHM while creating the PSF. It will be in the next version of "star" coming out next week.


Herb Johnson hjohnson@pluto.njcc.com
Mar 7, 1997

*> I have been trying to put the star lists together and make sense
*> of them.  One thing is clear, however bad the astrometry is, it 
*> is good enough to match stars.  The photometry is the big problem.
This is my impression also, as I'm working on a second set of star list comparisons. We can find stars, across all three sets of lists. The two who are determining RA and dec (Richmond and Gutswiller) seem to correlate well: I'm still looking at SAO and GSC catalogs to see where those stars fit in. But magnitudes are another matter.

I seem to be good at this kind of dogwork, so I'll continue on it. Richmond is posting my reports on the Web page (under Tom's "original" set of images and further analysis"): I encourage comments.

Meanwhile, one comment not yet in my reports. Richmond and Gutswiller apparently use the same pixel coordinates in their starlists: they report the same stars in the same places. But Gombert's star lists (the two I've checked) are consistently reporting "X positions" or pixel column values 15 pixels greater; and "Y positions" or pixel row values 1 pixel greater. Mutual agreement would be convenient.


Glenn Gombert ggombert@infinet.com
Mar 7, 1997

After gathering ~160 megabytes of data last nigth (still unreduced) I have outfited one of the "I" band TASS cameras with an infrared blocking filter that I use with my ST-6. It should block any wavelength's longer than about 0.7 microns. I am interested in finding out how the PSF varies as the near-infrared is removed from the "W" TASS camera.

I will calculate the mean PSF that I have calculated from the data run taken last night and compare the two it should be a very interesting experiment. It may help to sort out how the wavelenght transmission through the lens varies as a functions of wavelenght for the TASS triplets.

I will post a raw image or two on Storm when I get some data taken (tonight if the weather clear's up).


Glenn Gombert ggombert@infinet.com
Mar 7, 1997

Iam *very* curious as to where the "offset" is coming from, after matching up numerous stars to postions measured on the image I get < +/- 1 pixel variation from the "raw" image to the postion shown in the star lists. It would be nice to get to the "bottom" of this. There should be little variation between the three technqiues (especially on btighter stars's X & Y positions)

I have posted several example(s) of this in the last several weeks and will show this in the TASS Technical Note that I am putting together on astrometric and photmetric data reduction.


Benoit Schillings benoit@be.com
Mar 7, 1997

An interesting thing to try :

since a good deal of the star light is spread over n pixels, reducing the aperature of the lens might actually improve star detection when the camera is used in light poluted areas... The core of the star image might stay about the same while the background level would go down.


Glenn Gombert ggombert@infinet.com
Mar 7, 1997

> but there will be many blended objects that pass such
> tests and adding shape information to the output file will help us in
> determining how to handle such objects."
Arne has made a point I think that is critical to accurate data reduction. I have seen several examples in the last couple weeks where a faint background star is in the aperature of the star that is being measured. I think that a way to detect this and make sure that it can be identified for each star that is being measured is very important for accurate results...


Arne Henden aah@nofs.navy.mil
Mar 8, 1997

Glenn G. wrote:

>        After gathering ~160 megabytes of data last nigth (still unreduced)
> I have outfited one of the "I" band TASS cameras with an infrared blocking
> filter that I use with my ST-6. It should block any wavelength's longer than
> about 0.7 microns. I am interested in finding out how the PSF varies as the
> near-infrared is removed from the "W" TASS camera.
I hope that this is a misprint, or that Glenn did not include a couple of key words! If you put an IR blocker along with an I filter, you will get zero throughput, since the I filter peaks around 8500 A. The IR blocker should accompany the V filter ONLY.

Regarding pixel addressing: The 15 pixel offset is of course because one program ignores the beginning blind pixels (i.e., only uses the true image area) and the other one includes them. Likewise, a +-1pixel error usually is due to whether the program is counting rows starting at 1 (like Fortran and IRAF) or at 0 (like C).

I have yet another request for folks writing the extraction programs. You list the RA,DEC,x,y,mag, etc., but you should also include the UT or JD for each object (since they are not identical). This is the only way we can properly handle variable stars. The UT can be calculated by using the beginning UT for the frame and adding the 0.91sec/row offset.


Herb Johnson hjohnson@pluto.njcc.com
Mar 8, 1997

*> The new version of "star" will adjust the magnitude scale of the list to 
*> best match the [GSC] catalog.  
*> While testing the software I found the amount of
*> shift in magnitude increased as the catalog I was using went to fainter 
*> magnitudes in the GSC.  By the time the catalog got to mag 12 the shift 
*> stopped moving though and the shift was the same for a mag 13 catalog and 
*> the mag 14 catalog.  Looking at the data by hand I discovered that brighter 
*> stars were consistently measured below their GSC values.  To illustrate 
*> this effect I have uploaded an Excel spreadsheet to:
I don't have Excel up at this time. I'd appreciate a simple plot of magnitude vs. this shift you mention, maybe exported as a .GIF file or whatever is reasonably universal. This shift begs a question: is it an artifact of the GSC? My discussions with Richmond on the GSC suggest it is not a good *photometric* reference: for instance, when a given area of sky has "enough" reference stars, the list ends for that area - so the GSC is not a random or "complete" sampling.

This is a very tough practical problem.

*> I believe what's happening is that bright stars which have TASS V or I 
*> magnitudes lower than the GSC magnitudes (B?, P?) are preferentially 
*> selected since they are bluer and do not saturate the A/D converter as 
*> easily.  ("star" ignores any star which has a saturated core.)
Interesting....of course this is a statistical argument and also dependent on galactic location.
*> The end result is that the magnitudes published by the "star" program based 
*> on its (100,000 ADU = mag 8.5) baseline are about 1 magnitude too low.  I 
*> now estimate the TASS 3 sigma limit to be near mag 14.5 as a result.
Odd. My reports, on the first 50 or so stars between you and Richmond, did not show a 1-magnitude difference between the two of you. My sample was small however (and with an I band image. I'm now looking at a V band image.) Are you saying "one magnitude too low *relative to the GSC*"? Richmond tells me the GSC magnitudes, as determined from the Paliomar Survey E band plates, are closer to V band results than I band results: so using the GSC magnitudes for the I band is even less appropriate.

The location of my reports on the Web page is described in another TASS mail message today. I would be encouraged by any suggestions as to how to make additional and useful stellar magnitude comparisons. The "astrometrics" of location seem to be a solved problem.


Herb Johnson hjohnson@pluto.njcc.com
Mar 8, 1997

*> since a good deal of the star light is spread over n pixels, reducing
*> the aperature of the lens might actually improve star detection when
*> the camera is used in light poluted areas...
*> The core of the star image might stay about the same while the background
*> level would go down.
There would be fewer saturated star images with a reduced aperture, and some dim stars would be lost, so the range of *detected* magnitudes would shift to the brighter stars. This might be a plus in terms of finding reference stars as there seem to be brighter stars available in some lists. As it stands now, we need a 9th magnitude or better catalog!

It's occured to me that another way to reduce saturation is to use a rotating mask, that is to block the field of view for some percentage of the time. Since we do a row shift every .9 seconds, a half-moon circular mask rotating at, say several rotations per second, would reduce starlight by 50%. In effect you are sampling the stellar field at a rate much faster than the shift rate. This scheme would preserve the focal ratio and other optical parameters in use - an easy camera modification for most of us. (Keep the motor away from the camera.)

But we need 9th magnitude or better to find asteroids and other objects, so this scheme is at best a way to calibrate to brighter, better-known stars.


Glenn Gombert ggombert@infinet.com
Mar 8, 1997

The intent is to use only and IR blocking filter and not an "I"/"V" band filter(s) to use all the visible light shorter than 0.7 microns. Then conpare this to see what kind of effect this has upon stellar FWHM in as compared with "I" band images.


Alain Maury maury@ocar01.obs-azur.fr
Mar 8, 1997

From my perspective, I read those comments as congratulations more than anything else.

I am however puzzled by the position of Leif Robinson. Indeed there is a lot of new astronomy that can be done using CCD cameras, and TASS is only a small aspect of it all. I'd then appreciate if Sky and Tel would show less amateurs pushing glass, and guiding nice astrophotos, and more amateurs doing astronomy instead. If Sky and Tel is not showing the way, then who's role is it ?

When the average John Doe starts astronomy, he looks in sky and tel to see what the others are doing, and usually he does the same, he starts looking at NGC0001 and stops at NGC9999, then photographs the same. Hasn't done much astronomy in the end, and Sky and Tel is ( to my opinion ) partly responsible of this state of affairs. CCD cameras are changing the way we do astronomy only if people are shown that this can be the case. Reducing the CCD camera to a faster way of making nice images of the sky is in the end ( taking into account all that's left for amateurs to discover ) quite pityful.

Sky and Tel ( since CCD astronomy does not exist anymore ) should open a new regular page on amateurs doing astronomy. And there are enough of them to make a full page in each issue. I am aware of those amateurs discovering variable stars, novae, supernovae, following satellites, following asteroids, doing spectrographic studies, surveying the sky with blurry 135mm cameras ;-), you name it.

So Leif, put the paper where your mouth is :-)

( I hope this remark is as mild as I hope it is, cross my finger that this translation is not agressive ! ).


Nick Beser beser@aplcomm.jhuapl.edu
Mar 8, 1997

> So Leif, put the paper where your mouth is :-)
> 
> ( I hope this remark is as mild as I hope it is, cross my finger
> that this translation is not agressive ! ).

Sky and Telescope is attempting to reach a common denominator, as an observer rather than a industry leader. This gets into the fuzzy area of the role of the journalist as a maker of the news instead of the reporter. I feel that you are correct in that by not showing amateurs crossing the line into professional research they are not reporting the whole story. You may also be correct that by limiting what the amateur sees as a scope of effort, S&T may also be implying that extending the state of the art is not the realm of the amateur.

I think it may be useful to forward this thread to Sky and Telescope. Their e-mail address is:

skytel@skypub.com

(This is listed for General editorial inquires. There is a long list of e-mail address at:

http://www.skypub.com/s_t/s_tmail.html


Tom Droege droege@wwa.com
Mar 8, 1997

Well, I finally gave in and bought a second ZIP drive to move data from the basement to the 2nd floor data analysis machine.

The problem was that the 2120 in the basement was only slightly compatable with the TR3 drive on the analysis computer. It would loose about 1 file in 10. A real pain. So I thought the Zip would be easier. ;^(

Installing the ZIP took all morning. The Zip software would just hang the computer when it got to the program that loaded the program that found a drive letter for the Zip. This lead to the standard problem. Something in autoexec.bat was hanging the computer but I coulden't edit autoexec.bat because it was hanging the computer. Now where is that DOS boot disk when I need it.

I finally got it going after extensive excercising of my four letter vocabulary. It turns out that the Zip drive did not like mouse.com in the autoexec.bat file. Now just why should that be?

In any case, I am now set up to send 100Mb data sets out to data munchers that don't have a telescope. I am preparing a disk for Arne, perhaps he will pass it on to someone else that wants it.


Glenn Gombert ggombert@infinet.com
Mar 8, 1997

Herb's point is well taken that the lack of good "standards" make it difficult to make good photometric calibrations of TASS images. I have added a "Tech Note" to the Dayton TASS Camera Home Page, as well as asking Michael to add this to the TASS Home Page on photomertic calibration using the image of M67 that I posted to storm several weeks ago. (http://www.infinet.com/~ggombert/tass.html)

It was taken with a 16 inch Newtonian telescope and SBIG ST-6 ccd camera with a "V" band photometric filter. The image has 11-12 well known reference stars in the image that have well published photometric magnitudes.

Dr. Mattei at the AAVSO was kind enough to supply a copy of the AAVSO cailbration chart for the M67 region that was very helpful in calibrating the catalog output for the image.

I carefully tried to set the "zero point" of the Sextractor program and then checked the magnitudes that were generated for the other 11 reference stars in the image. It is an excellent way to compare the photometric accuracy of any method for generating star-lists. I have shown the star-list output as well as the photometric magnitudes for 11-12 refernece stars (with comments out to the side).

It looks like a few percent photometric accuracy should be possible on a routine basis with careful calibration. The other star magnitudes in the GSC/SAO catalogs are just not accurate enought (as Arne has pointed out several times). This makes it very difficult to tell how accurate photomertic magnitudes as being calculated in TASS images.

I would be happy to supply as set of the AAVSO M67 field reference data upon request. (Just give me a mailing address). I think that this image should prove useful in evaluating various ways of calculating instrumental stellar magnitudes.


Bob Goff goffaxe@azstarnet.com
Mar 8, 1997

Ya know, the Japanese have a real astronomy magazine. It puts S&T to shame as an instructional device for observational astronomy.


Herb Johnson hjohnson@pluto.njcc.com
Mar 9, 1997

The following is a report on my cross-comparisons of three star lists of Tom's image S31T0493.669; list made of course by Richmond, Gutzwiller, and Gombert. Michael Richmond, please archive this as file "rpt1493.txt" in the area of the Web page with my other report. It's short enough that I thought I'd simply put it out on the maillist. (Note: the report exceeds 80 columns of text.) I'll make no further astrometric comparison reports, as it seems we have no problems there. I'll do no further photometric reports until I come across "better" catalogs. As Glen Gombert noted, the catalogs in hand are not sufficient. - Herb

---------------------

Comparison of star list magnitudes of Droege S31T0493.669 image
Herb Johnson March 9 1997

Image is V band. GSC catalog stars from E plates of Paliomar Sky
Survey plates. SAO is ??

SAO
---

SAO catalog for RA 117.4 to 117.999, dec 0.X to -2.4999

Richmond .AST stars for same area, by Richmond number and .AST magnitude
  (Radius R of 3.5 pixels).
Gutzwiller stars for same area, by magnitude
Gombert stars for same area, found via Richmond .AST and .PHT lists only,
   by Gombert number and magnitude

  SAO    RA      dec    mag   Richmond  Gutzwiller Gombert (via Richmond)
135145 117.4584 -2.0664 8.5  30  8.343      7.992   59  8.4326
135150 117.4934 -1.4697 9.0  46  9.608      9.232   90  9.6976
135154 117.5285 -1.6822 8.0
135157 117.5583 -1.4593 8.2  76  8.062      7.725  146  8.1444
135161 117.6426 -1.6099 8.5 107  8.075      7.698  226  8.1587
135171 117.7191 -2.4953 8.3 142  8.329             267  8.5695
116019 117.7222  0.0792 6.6
135175 117.8137 -0.4750 9.0                 8.850
135176 117.8221 -0.4620 9.1                 8.896
116031 117.8242 -0.1050 9.0 195  9.213      8.850  350  9.3497
135178 117.8348 -2.0626 8.6 199  8.256      7.906  366  8.3489
116040 117.8938  0.3830 8.7 222  8.215      7.878  400  8.3967

GSC
---

GSC stars RA 117.4 to 117.999, decl -0.0 to -2.4999

Richmond .AST stars for same area, by Richmond number and .AST magnitude
  (Radius R of 3.5 pixels).
Gutzwiller stars for same area, by magnitude
Gombert stars for same area, found via Richmond .AST and .PHT lists only,
   by Gombert number and magnitude

        GSC     RA          dec      mag    Richmond  Gutzwiller Gombert (via Richmond)
GSC0483201258  117.5306   -0.3429    10.74  62 10.304   9.947   109 10.3989
GSC0483601452  117.5929   -2.2431    10.77  90 10.571  10.171   162 10.7619
GSC0483200773  117.6202   -1.3077    11.62 100 11.681  11.316   175 11.7486
GSC0483600053  117.6446   -2.0127    11.23 108 11.245  10.930   203 11.2856
GSC0483200475  117.6488   -1.6444     9.37 110  9.426   9.066   227  9.0574
GSC0483200400  117.6892   -0.0030    11.69 125 10.982  10.624   234 11.1109
GSC0483201504  117.7124   -0.9779    10.95 140 10.638  10.302   263 10.7166
GSC0483200700  117.7204   -0.6528    10.76 145  9.995   9.645   269 10.1181
GSC0483200544  117.9566   -0.6752    11.94 256 11.519  11.213   448 11.6452
GSC0483201812  117.9848   -0.3181    11.98             11.223
GSC0483600017  117.9922   -2.0378    10.00 268 10.362  10.012   476 10.4492

Other stats:
-----------
Richmond's list: 1400 stars (to #268 identified on GSC; #222 id'ed on SAO)
Gutzwiller's list: 2182 stars (to #388 identified on GSC; #322 id'ed on SAO)
Gombert's list: 2207 stars (to #476 identified on GSC; #400 id'ed on SAO)
No overlap for the 12 SAO and 11 GSC stars in this range of .6 degree RA
  and 2.5 degrees in dec (1.5 degrees square).

Methods
-------

All files sorted by RA.
Catlog file (SAO, GSC) edited for stars of RA 117.X, stars out of dec
   range removed from list.
Richmond .AST file manually searched for matching RA and dec to about +/- .02.
Gutzwiller file manually searched for matching RA and dec to about +/- .02.
   (In most cases, the most adjacent stars are off by much more than this.)
Matching stars in Richmond .AST list found in Richmond .PHT list by
  Richmond number; pixel address converted to Gombert pixel coordinates;
  (add 1 to Richmond row address; add 15 to Richmond column address)
  corresponding Gombert star found.


Herb Johnson hjohnson@pluto.njcc.com
Mar 9, 1997

*> I'd then appreciate if Sky and Tel would show less amateurs
*> pushing glass, and guiding nice astrophotos, and more amateurs
*> doing astronomy instead. If Sky and Tel is not showing the
*> way, then who's role is it ?
*> When the average John Doe starts astronomy, he looks in sky and tel
*> to see what the others are doing, and usually he does the same,
*> he starts looking at NGC0001 and stops at NGC9999, then photographs
*> the same. Hasn't done much astronomy in the end, and Sky and Tel
*> is ( to my opinion ) partly responsible of this state of
*> affairs.
As my uncle used to say, "don't blame the cow for sour milk". My impression is that Sky and Tel (and Astronomy magazine also) is following the popular trend in (American) astronomy: a growing public interest due to major celestial events and upgrades to the Hubble Space Telescope. The events (comets and Jupiter impacts) have increased interest in observing, and general interests in astronmical stuff. The Hubble repairs, and subsequent pretty pictures, have "raised the level" of expectations for pretty astronomical pictures. Related events include lunar and martian space probes, and the growing use of CCD cameras to produce even *more* pretty pictures. Also, growing Web use by the public as supported by NASA and HST Web sites with, again, lotsa pretty pictures.

THe "pretty pictures" explanation is the only way to account for the photos of star trails in the astronomy magazines.

As a member of a number of astronomy clubs, I can personally confirm increased public interest: we have larger memberships, more people come to our events, star parties are bigger and there are more of them.

As for your comments that all this is "not real astronomy": well, again my amateur experience is that very few amateurs engage in scientific data collection. I recall, for instance, an amateur who has set a goal of being the person to have seen the most asteriods. He does this by charting where they will be, and then he looks at them on successive nights (to confirm their motion) and then he makes a simple diagram in his notebooks. *However*, to my surprise, he does not bother with efforts to note the *exact* time of his observations, nor does he bother to determine exactly *where* in the sky the object is. Thus, his observations have little scientific value. But he *is* #2 in the world now in "counting asteroids".

The point is, it is futile to argue about what "real astronomy" is or should be. The point is to do it, as best you can, and to report results and "spread the word" that even simple but CAREFUL observations can make a difference - and can be fun at the same time. Asteroid occultations are brief events, but even with a telescope and stopwatch they can be EXTREMELY important in establishing asteroid orbits and size, as well as determine very close double stars.

As we report on our observations and how we do them, someone else will be inspired to do more careful work. That's how *I* got into all this. But we must all work harder to write up and report our results. Write a paper; give a talk; post results and methods on a Web page. Most astronomy clubs and Web sites are BEGGING for original work!

As for the public interest in pretty pictures: well, the public provides donated funds for many of our astronomy clubs, and their taxes pay for many government-funded astronomy programs, so there that is.

I would not suggest complaining to Sky and Tel; I'd suggest writing articles, and *encouraging* these magazines when they PUBLISH original AMATEUR work of scientific merit. Just my opinions of course.


Herb Johnson hjohnson@pluto.njcc.com
Mar 9, 1997

*>         Herb's point is well taken that the lack of good "standards" make it
*> difficult to make good photometric calibrations of TASS images. I have added
*> a "Tech Note" to the Dayton TASS Camera Home Page, as well as asking Michael
I'll read this most carefully. Thanks!
*>        It looks like a few percent photometric accuracy should be possible
*> on a routine basis with careful calibration. The other star magnitudes in
*> the GSC/SAO catalogs are just not accurate enought (as Arne has pointed out
*> several times). This makes it very difficult to tell how accurate
*> photomertic magnitudes as being calculated in TASS images.
I presume this assumes that the seeing is consistent and clear for the length of the run. I don't think the problem will be "calibrating" the camera, although that also needs to be established over long periods. The problem we won't have obvious data on, or control of, will be the less-than-perfect skies. As amateurs, we can make a signifigant contribution in my opinion if we can work around (or through) this. The professionals of course simply throw out data from bad skies at otherwise perfect sites: we don't have that luxury.

And of course we also need "better" reference catalogs, if such can be found or made.


Herb Johnson hjohnson@pluto.njcc.com
Mar 9, 1997

*>  I have yet another request for folks writing the extraction programs.
*> You list the RA,DEC,x,y,mag, etc., but you should also include the UT or
*> JD for each object (since they are not identical).  This is the only
*> way we can properly handle variable stars.  The UT can be calculated by
*> using the beginning UT for the frame and adding the 0.91sec/row offset.
I don't want to argue, but if the UT can be easily calculated, then why add it to the lists? It would have some error anyway, unless Tom was very careful in recording the start of data collection; and unless there is no odd drift or roundoff error in the timer used to shift the pixel rows. [Incidently, how good IS that shift clock timer anyway, Tom? ;) ] Shorter list files are better.

I *would* encourage a sequence number for all lists, to make references to individual stars easier. Then of course any computed information can be easily added to a secondary list crossed by sequence number.


Chris Albertson chrisa@wavenet.com
Mar 9, 1997

> *>  I have yet another request for folks writing the extraction programs.
> *>You list the RA,DEC,x,y,mag, etc., but you should also include the UT or
> *>JD for each object (since they are not identical).  This is the only
> *>way we can properly handle variable stars.  The UT can be calculated by
> *>using the beginning UT for the frame and adding the 0.91sec/row offset.
> 
> I don't want to argue, but if the UT can be easily calculated, then
> why add it to the lists? It would have some error anyway, unless Tom
> was very careful in recording the start of data collection; and unless
> there is no odd drift or roundoff error in the timer used to shift the pixel
> rows. [Incidently, how good IS that shift clock timer anyway, Tom? ;) ]
> Shorter list files are better.
Most Database Systems (even Microsoft Access) have the concept of a "computed field" This is a kind of column that is derived as a function of other data. Is is not stored on the disk but none the less may be printed or even sorted or used as part of a selection criteria. It will not be to long before these lists will be stored in databases rather then files when this happens you will have your UT column or maybe some other column.


John Gwinner gwinner@northnet.org
Mar 10, 1997

> *>I'd then appreciate if Sky and Tel would show less amateurs
> *>pushing glass, 
>
> As my uncle used to say, "don't blame the cow for sour milk". 
Interesting, as the ATM list generally feels that less attention is given nowadays to making good optics, and instead the general lack of quality optics in commercial scopes is ruining some amateur's experiences.
> As we report on our observations
> and how we do them, someone else will be inspired to do more careful
> work. That's how *I* got into all this. But we must all work harder to
> write up and report our results. Write a paper; give a talk; post
> results and methods on a Web page. Most astronomy clubs and Web sites
> are BEGGING for original work!
I generally don't see that much on what observations are needed to be done. I've seen some web pages on lunar occultation's, and I have to admit I've been more focused on actually making a scope to be able to get out from the terminal. However, I think one good thing to do might be to start a newsgroup with 'observations needed' or something to coordinate amateur observations (web URL's cheerfully accepted). I know there's several societies out there -- heck, that's why I joined this mailing list! Making them more accessible might help. One of the things I like about TASS is that the sky is 'registered' so that if I get a camera or build one or whatever, I can feel assured that I'd get the guidance to point it in a direction that will do a lot of good. If I'm just randomly timing asteroids I'd feel that I was potentially duplicating an observation made countless times before (of course, verification is important too).


Tom Droege droege@wwa.com
Mar 10, 1997

Herb asks "how good is the shift clock timer?". I give it about 0.5% over normal room temperature. If you let it warm up, and have the computer in a room at pretty constant temperature, then it will hold to about 0.1%. The next one will be at least x10 better.


Tom Droege droege@wwa.com
Mar 10, 1997

I am preparing a couple of Zip disks with raw data. I have nearly 10 days of B data and I think 9 days of A data. I plan to ship these to Arne Hendon with instructions to ship them on to the next person on the list who asks for them. Anyone else want them? Otherwise, I will just have him send them on to Michael Richmond.


Michael Richmond richmond@astro.princeton.edu
Mar 10, 1997

I made another trip to Vermont this weekend to run the triplet. I'll write a detailed report in the future, but here's the short and sweet:

     - a wind storm blew the roof off the shed housing the camera
          several weeks ago, but fortunately, no rain/snow fell in
          while it was off.  The camera was probably shifted by
          the wind, though

     - I spent half of one night trying to re-align the camera.
          It would much better to have the triplet body mounted
          on a pier ... mine just sits, loose, on the base of the
          shed.

     - the best VCO setting for this visit was very different
          from the last one (a few months ago)

     - I couldn't use Norman's program because my triplet has some
          wiring problems (I think) which cause the CCD temperature
          and water temperature values to jump between 3 or 4 different
          values.  Norman's program halts whenever the CCD is warmer
          than the water, and the bogus values always caused this
          to happen quickly.
 
     - Tom's old BASIC program still works fine!

     - During a recent power outage, the pump stopped moving
          water in the coolant tubes, and much of it froze.
          There was just a trickle of water coming out of the return
          tube this weekend, but I judged it to be sufficient.
          I _had_ placed about 8 gallons of antifreeze into 20 gallons
          of water, but I guess it wasn't enough.

     - I got one gloriously clear night, with 8 or so hours of
          scanning. 

     - during this night, I placed an "aperture mask" over one of the
          I-band lenses for several hours, then removed it.  We can
          see if the mask (which had a diameter about half that of 
          the lens) caused the PSF to improve

     - the next morning, when I replaced the lens caps, one of the
          I-band lenses felt loose.  I jiggled it a little, and it
          CAME OFF IN MY HAND!  Yikes!  Evidently, the cold temperatures
          (about -10 or -15 Celsius) cause the metal strip Tom used
          to fasten lens to triplet body to become brittle.  I'll
          have to find some way to re-attach the lens the next
          time I observe....
Overall, a success -- I brought back about 224 Meg of data, and will try to reduce it this week.


Chris Albertson chrisa@wavenet.com
Mar 10, 1997

> I am preparing a couple of Zip disks with raw data.  I have nearly
> 10 days of B data and I think 9 days of A data.  I plan to ship these
> to Arne Hendon with instructions to ship them on to the next person
> on the list who asks for them.  Anyone else want them?  Otherwise, I
> will just have him send them on to Michael Richmond.
As it looks like Tom will be shipping data on ZIP disks I have one question for Arne Hendon first and then others who would want this data:

Do you have a way to copy this 100MB or so of data to some other media?

For example I do not have a zip drive but can translate between QIC-80, 4mmDAT, CD-ROM formats. Maybe someone else has a few other formats they can read/write.

What media does everyone have? Those of you with cameras and without cameras. If we all post our capibility it may be possible to work out an exchange order where no one has to spend money for new hardware.

I'll go first:

Chris Albertson:

   0) No camera
   1) QIC-80 tape drive, 120MB native uncompressed
   2) DAT 4mm DDS1 tape drive, 2GB native uncompressed
   3) CD-ROM Writter (at the office)
If you all would post something like this I could sort it out.

With my 4mm drive, I would not mind archiving data. The cost is low. I can store about 3GB of image data on one $9.00 tape.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 10, 1997

> As it looks like Tom will be shipping data on ZIP disks I have
> one question for Arne Hendon first and then others who would
> want this data:
> 
> Do you have a way to copy this 100MB or so of data to some
> other media?
When I return to the lab in 3-4 weeks, I will be able to read the ZIP disks, and copy them to CD-ROM (I have a CD-ROM writer). The lab pays about $7.00 for a blank CD. I have not priced them at Comp USA recently. We can make a limited number of copies of CD's (about 650Mbyte each). I'm guessing that our cost will be about $8.00 total ($7.00 for the CD, $1.00 for postage). I can get a better estimate after I return to work.


Arne Henden aah@nofs.navy.mil
Mar 10, 1997

Chris suggested we list what media we can support, as a guide to data transfer capability.

At work, we use Exabyte almost exclusively. Similar to DAT (just older technology) and about the same cost/GB. Our drives are 8500's, 5GB/tape, $2/GB. Very slow access, though.

I archive all my CCD data on CDROM as we have an old writer at the Observatory. Cost is higher: $15/GB or so. Very fast random access, but at 120MB/night, the current round of CDR only holds 4-5 nights of raw frames.

We also have the 3.5" MO 240MB drives. Commonly only used for backups for the PCs. Expensive media, but rewritable.

At home, I use a 2120 (CMS Jumbo 250) for most of my backups. Tom has convinced me that I need a Zip drive for compatability with him, so I will go out and buy one this week and probably install it on an Observatory PC so that I can get his files onto the local network. Zips are ok, kinda slow and more expensive/GB than the other alternatives. Drives are cheap, though. The Jazz might be a better choice since it holds 1GB/disk.

My personal preference is to save the raw TASS frames on tape (DAT/Exabyte) as that is the current cheapest method. Tapes do not have long-term stability, but once you get the extraction software debugged and stable, you should rarely have to go back to the raw files. Tapes are difficult to write so that they can be read at other sites, so they should probably be local archive only, and some smaller media selected if you want to ship data site-site. Internet is ok between two sites with high-speed links, but you wouldn't want to transfer 100MB/night over a 28.8 modem. Hopefully, this round of raw fits file transfers will not be the norm, and we can just transfer the extracted data in the future.


Herb Johnson hjohnson@pluto.njcc.com
Mar 10, 1997

*>> *>I'd then appreciate if Sky and Tel would show less amateurs
*>> *>pushing glass, 
*>
*>> As my uncle used to say, "don't blame the cow for sour milk". 
*>
*> Interesting, as the ATM list generally feels that less attention is given
*> nowadays to making good optics, and instead the general lack of quality
*> optics in commercial scopes is ruining some amateur's experiences.
I think Alain was a bit excited when he made that particular remark. His point was otherwise about "pretty pictures" over good science.
*>> As we report on our observations
*>> and how we do them, someone else will be inspired to do more careful
*>> work. That's how *I* got into all this. But we must all work harder to
*>> write up and report our results. Write a paper; give a talk; post
*>> results and methods on a Web page. Most astronomy clubs and Web sites
*>> are BEGGING for original work!
*>
*>I generally don't see that much on what observations are needed to be done.
It's not just observations, but the *methods* and *record keeping* that are critical. My example made that point: an amateur who observed thousands of asteroids (I recall a number like 6000) but who did follow good practices (recording time, determining RA/dec, etc.). He was a great observer, but not a good scientist.
*> However, I think one good thing to do might be to start a
*> newsgroup with 'observations needed' or something to coordinate amateur
*> observations (web URL's cheerfully accepted).
*> direction that will do a lot of good.  If I'm just randomly timing
*> asteroids I'd feel that I was potentially duplicating an observation made
*> countless times before (of course, verification is important too).
Verification is the core of the scientific method. The whole IDEA of modern science is that YOU can see the same thing I see, if you perform the same experiment in the same way. THis is why science is not mysticism: it is not *personal* experience, but collective and repeatable experience.

This is getting off the subject of the TASS project, but it's worth making the point that if the work we do is to be useful, it must be done in a repeatable, verifiable way. I'm over 40: I was taught the scientific method in JUNIOR HIGH school (8th year of public school). Today, all that kids probably see of science is a bunch of "first discoveries" rather than the patient, persistant, repetitious nature of most scientific work. In this context, sometimes I feel foolish just for asking for old data: who wants to be "second"?


Arne Henden aah@nofs.navy.mil
Mar 10, 1997

Just an observing update for those of you working on the SMSP-B field. I've had two clear nights so far, with another clear one predicted for this evening (this is why the Southwest is a desired astronomical site). I have data now on three SMSP-B fields, a total of several hundred stars in BVRI filters and from V=9-17. I should be able to get it reduced by the end of the week and will send the results to M. Richmond to post. These sequences should give workers a good start in determining the faintness limit with the cameras, and more stars with which to perform color transformations. I also have data in the SMSP-A field, but am not concentrating on it this month. Also, this setup of secondary standards is a special case for the SMSP fields. I don't intend to do it throughout the entire equatorial region as it just takes too much observing and data reduction time. I feel it is important to do it for these special cases for proper calibration and understanding of the TASS camera performance.

I can't work much brighter than V=9 here. If someone wants to get some data on stars from, say, V=6 to V=9 in the SMSP-B field, I think that would be very useful in checking saturation limits and magnitude effects on the cameras. I am willing to help such an observer in reducing the data and setting up the proper 'cookbook' for standard star observing.


Arne Henden aah@nofs.navy.mil
Mar 10, 1997

Chris A. wrote:

> Most Database Systems (even Microsoft Access) have the concept of a
>"computed field"  This is a kind of column that is derived as a function
> of other data.  Is is not stored on the disk but none the less may be
> printed or even sorted or used as part of a selection criteria.  It will
> not be to long before these lists will be stored in databases rather
> then
> files when this happens you will have your UT column or maybe some
> other column.

This is a good point, and I'm glad Chris brought it up as many folks may not have used _real_ database systems before. However, I don't feel it is all that useful for the TASS database. There will be a different set of parameters for each FITS file, and so many for a given night, and that doesn't count the multiple cameras/site and multiple sites. When you start combining these datasets, especially if you start indexing on an individual object rather than a given frame, there may be considerable added complexity. I _much_ prefer to tag each observation with its unique time, and then you can combine any way you want.

Also, I still feel that all TASS software needs to be public domain and not a purchased commercial product. We do this bookkeeping with homegrown programs at the observatory with little effort, and more customized than the typical database program. That is not to say that an individual can't munge the TASS dataset with whatever software they have available, but I would like to have some basic tools available to all.


Tom Droege droege@wwa.com
Mar 10, 1997

Hmmmm! GB Weld works great to patch a hole in your auto manifold, possibly not so good at -20 in Vermont though. I used this over standard Epoxy because I though it would be stronger. Next time you fly in a jet liner remember they are mostly glued together.

Just remove the lens ring (four screws) and scrape off the surface so you can glue it square, and use 10 Ton epoxy to glue it back on. Make sure both surfaces are clean. I have had one or two lenses pop off, mostly from a shock.

The VCO setting will depend on the tempreature of the room with the PC. My thinking was that people lived where PCs live and generally like to be at about 20 C. The next design is better. I have also set it up so the computer can read the temperature of the VCO and make a second order correction.

I don't understand the temperature jumping problem. With the roof off the shed, I would worry about condensation on the printed circuit board in the camera. Possibly just put it in a warm dry room for a couple of days. Is there such a thing in Vermont?

Even a few drops a second of water will be enough to cool the camera so no damage is done. Then the warm water will soon make it's own path.

Hearing about problems is very helpful to me. Believe me, every one affects the design of the Mark IV. I keep designing in more corrective and diagnostic features.

Congratulations on adding to the data store. We should soon have enough for some modest statistics.


Arne Henden aah@nofs.navy.mil
Mar 10, 1997

Michael Richmond noted in his technical report on Droege's February images that he saw an anticorrelation between dark current level and TEC temperature. That is, as the temperature went down, the dark current seemed to go up.

Actually, I think Michael was seeing an effect that we have on all of our CCD systems. The data Michael was using were the 'blind' columns after the image array, and those pixels include both dark current and read bias. On our CCD systems, it is the read bias that changes most with temperature after you get the CCD relatively cold, and it does change in the opposite direction. For example, on our Tek/SITe 1024x1024, the bias level is zero at -60C and 200 DN at -100C. This is one of the reasons why it is nice to have CCD overscan, as you can separate the two effects.


Arne Henden aah@nofs.navy.mil
Mar 10, 1997

Nick suggested CDR media for TASS data storage, and in general I agree that it is the most transportable and longest-lived of the mentioned media. After all, that is why I archive all of my CCD data on CDROMs!

I didn't suggest it for storage because not all sites have CDR writers. These still cost $500-1000. While Nick was willing to burn CDROMs for TASS (and I could as well here at the observatory), you still have the problem of getting the data to Nick. That is why I mentioned the Zip drives, since Tom already has one, and they are in the $150 range, install off of the printer port, and you can reuse the media (making it ideal for transport to different sites). Also, I still feel this is a short-term solution as we shouldn't need the raw fits files. Roboscope (I.U.) throws away all of its frames after processing; our 1.3m with its 6kx8k will probably do the same.


Norman Molhant molhant@ERE.UMontreal.CA
Mar 10, 1997

>      - the best VCO setting for this visit was very different
>           from the last one (a few months ago)
I have the same problem here from night to night. It doesn't vary much, but the variations keep adding up.
>      - I couldn't use Norman's program because my triplet has some
>           wiring problems (I think) which cause the CCD temperature
>           and water temperature values to jump between 3 or 4 different
>           values.  Norman's program halts whenever the CCD is warmer
>           than the water, and the bogus values always caused this
>           to happen quickly.
There are (documented) parameters in the configuration file (tass.rc) that can be used to turn that module off: make the maximum acceptable difference 100 Celsius or more, lo! Doesn't stop any more.
>      - Tom's old BASIC program still works fine!
Sure. Fun is: I inserted te offending module at Tom's request.
>      - During a recent power outage, the pump stopped moving
>           water in the coolant tubes, and much of it froze.
>           There was just a trickle of water coming out of the return
>           tube this weekend, but I judged it to be sufficient.
>           I _had_ placed about 8 gallons of antifreeze into 20 gallons
>           of water, but I guess it wasn't enough.
I am using half antifreeze (glycol), half water, which is a much more appropriate mixture for Quebec's winters. It should also do for Vermont, no?
>      - the next morning, when I replaced the lens caps, one of the
>           I-band lenses felt loose.  I jiggled it a little, and it
>           CAME OFF IN MY HAND!  Yikes!  Evidently, the cold temperatures
>           (about -10 or -15 Celsius) cause the metal strip Tom used
>           to fasten lens to triplet body to become brittle.  I'll
>           have to find some way to re-attach the lens the next
>           time I observe....
Is it the metal or the glue that broke?
>   Overall, a success -- I brought back about 224 Meg of data, and will
> try to reduce it this week.
Great news! Wish I had such good news to report...


Glenn Gombert ggombert@infinet.com
Mar 11, 1997

My experience is somewhat similar to Michael's in that I usually end up with a different VCO setting that I use on a given evening's data collection run. I often spend 45minutes to 1 hour "tweaking" the VCO to give the best results.

I have found that if I do not do this that the data quality (FWHM, etc) is not nearly as good as it could be if I do not take the time to adjust the VCO.


Tom Droege droege@wwa.com
Mar 11, 1997

As I understand it, CCD overscan simply means that you read out a few extra pixels on each line??? If so, why don't we do it? Note we would not have to change the format, we could (just) modify the code to read out a few extra times, then put the data somewhere in the existing blank positions. I would not want to change the data format at this time - in the middle of a data run, but we might do it before the next serious push, which I would judge will be starting in July.

Comments, Arne, Norman??


Norman Molhant molhant@ERE.UMontreal.CA
Mar 11, 1997

Glenn wrote:

>         My experience is somewhat similar to Michael's in that I usually end
> up with a different VCO setting that I use on a given evening's data
> collection run. I often spend 45minutes to 1 hour "tweaking" the VCO to give
> the best results.
Indeed, the stability of the VCO is nothing to call home about. I use my home-made periodmeter (based on a 1 Mhz Xtal oscillator) to adjust it whenever I try to get a run. From cycle to successive cycle, the period is stable to within about 1 msec, but from day to day the VCO setting to get the correct period changes and there seems to be a more-or-less regularity to it: the changes do slowly accumulate as if some component were aging (is that a possibility, Tom?). On top of that slow but regular component of variation there is also a much stronger but nearly random day-to-day fluctuation that could be thermal in origin, or might reflect fluctuations in the power supply voltage of the PC, or... I don't know. Where the PC sits on my desk, the temperature varies between a minimum of 15 Celsius during the coolest clear winter nights (10 Celsius when the blizzard blows, but I don't try to get a TASS run in these conditions) and a maximum of 21 Celsius during hot summer nights (that's because my lab is in a heated [but not cooled] basement), with an average around 18 Celsius... Could such temperature differences explain the seemingly random fluctuations in the VCO setting (typically around 1%, sometimes up to 3%). Oh, yes, my PC has an open chassis (necessary to let me put probes here and there on the Tass Control Card) and the power supply fan does not blow over the Tass Control Card.
>         I have found that if I do not do this that the data quality (FWHM,
> etc) is not nearly as good as it could be if I do not take the time to
> adjust the VCO.
Same thing here, albeit the star images are quite large here - because I have no filter on the camera, I suppose. By the way "FWHM" represents the width of the point spread function, allright, but what is it the initials of? (yea, I know: sentences shouldn't end with of :-)


Tom Droege droege@wwa.com
Mar 11, 1997

Sorry about the VCO problems. Not much to be done at this time. I make sure I turn on the computer several hours before I start taking data. I also try to keep it at a uniform temperature. Note that you can run dark frames while it is still light (lens caps on) and do VCO adjustments at the same time. You do not have to lose observing time.

in a following message

OK, I agree that the VCO is a problem. How about it Norman, do you want to demonstrate your virtuosity?

Everything we need is there to automatically adjust the VCO. At least to the PC clock accuracy, which could in theory be corrected by the stars as they go by. So how about setting the VCO time interval instead of the VCO DAC setting. The software could check the time every 100 lines, compute the real clock rate, and make a correction to the DAC setting. Closed loop servo enthusiasts will note there can be a stability problem. It will just take a small enough loop gain, or some serious analysis.


Norman Molhant molhant@ERE.UMontreal.CA
Mar 11, 1997

Dear Tom,

> As I understand it, CCD overscan simply means that you read out a few 
> extra pixels on each line???  If so, why don't we do it?  Note we would
> not have to change the format, we could (just) modify the code to read 
> out a few extra times, then put the data somewhere in the existing blank
> positions.  I would not want to change the data format at this time - in
> the middle of a data run, but we might do it before the next serious 
> push, which I would judge will be starting in July.
Sounds like a very good idea, if that's what CCD overscan really is (Arne?). We still have an empty position in the current format (the last pixel is always 0 for now), we could use it to store an average of X overscan readings (what do you think, Arne? what kind of X would do here?). Fairly easy to do. I could do it for the DOS version of tm3get before the end of this week. If I'm going to implement that, what else should I implement right away?

in a following message

> OK, I agree that the VCO is a problem.  How about it Norman, do you
> want to demonstrate your virtuosity?
Well, my brain seems to be functionning anew, so why not?
> Everything we need is there to automatically adjust the VCO.  At least
> to the PC clock accuracy, which could in theory be corrected by the 
> stars as they go by.  So how about setting the VCO time interval instead
> of the VCO DAC setting.  The software could check the time every 100 
> lines, compute the real clock rate, and make a correction to the DAC
> setting.  Closed loop servo enthusiasts will note there can be a stability
> problem. It will just take a small enough loop gain, or some serious 
> analysis. 
Ok, I'll think about it - I might be able to do something along these lines.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 11, 1997

We have been going through a long and painful process of alignment, VCO adjust and focus on our camera. I have noticed that the VCO settings seem to need adjustment over each observing period. That may become a problem if we operate our camera remotely.

Is it possible for the computer the make the adjustment automatically (assuming that the problem is that the scan is not occuring at precisely .914 seconds?). Have the computer do a average number of lines collected over time, and adjust the settings to come out to the expected value?


Herb Johnson hjohnson@pluto.njcc.com
Mar 11, 1997

*>>  Most Database Systems (even Microsoft Access) have the concept of a
*>>"computed field"  This is a kind of column that is derived as a function
*>> of other data.  Is is not stored on the disk but none the less may be
*>> printed or even sorted or used as part of a selection criteria.  It will
*>> not be to long before these lists will be stored in databases rather
*>> then files when this happens you will have your UT column or maybe some
*>> other column.

*>  This is a good point, and I'm glad Chris brought it up as many folks
*> may not have used _real_ database systems before.  However, I don't feel

*>  Also, I still feel that all TASS software needs to be public domain and
*> not a purchased commercial product.  We do this bookkeeping with homegrown
*> programs at the observatory with little effort, and more customized than
*> the typical database program.  That is not to say that an individual can't
*> munge the TASS dataset with whatever software they have available, but I
*> would like to have some basic tools available to all.
(sigh) sometimes I feel like an old man defending old habits. But it seems to me that a number of people interested in TASS data, now and in the future, will not necessarily have either Unix systems or experiences with particular database systems. The raw data files are large but not overwhelmingly so, and the near future will provide media and technology to archive them at modest (but accumulative) cost. The PROCESSED data files are even smaller, so a "raw" or "flat" format for them is not unreasonable. I've argued against a data field that can be easily computed from scratch. But if it was a choice between adding another 10% to the size of the processed data vs. making it a database file, I'd choose the former.

One problem with ANY elaborate data file format is the long term. Ten or twenty years from now, will these formats endure? That is why we went to a FITS format for raw data, is it not? And if you argue that "who cares in twenty years?" - recall we lack "old data" of photometric and astrometric quality: but we may produce the "old data" needed twenty years from now.

I was going to comment on the issue of software transportability and operating systems. I'm not going to, it's like arguing against progress. But *please*, don't make the *data* too fancy for someone else to use it. Pretend you're sending it to your father, which is in a way true.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 11, 1997

>  what else?  I'll consider any suggestion received before thursday 18:00 EST.
We are setting up a poor man's remote system with the TM3GET11 software under control of a unix system. At the moment there is no feedback while data collection is going on. Ideally, I would like to receive a snap shot of the screen every minute or so just so I can see how things are going. At a minimum, the temperature, voltage, current and median readings would be good. If I were to add the code in the TM3GET11 software to send data over a serial link, at what point in the data collection process should I add the routine? I was told that if I wanted to save the VGA screen, that I should tie into the VGA driver, and respond during the vertical retrace period.

How far away is the Linux version of TM3GET11?


Mica Wickersham rjw@crl.com
Mar 11, 1997

As has been noted by several camera operators, the VCO stability is a problem for some installations where the electronics are not in a temperature-stable environment such as Tom's basement. This is also a problem for me.

When I get #5 back, I plan to work on a modification to improve the stability. Tom, of course, could do it, but he's occupied with the Mark IV development and shouldn't sidetrack to take care of a minor problem on the previous generation. Anyone else could work on the problem but most are busy with software. I am just mentioning my interest in attacking the problem so that others can be aware that the problem won't be ignored.

in a later message

It seems that a software servo may take care of the problem faster than a hardware solution and not require field mods. But if a hardware fix is needed for good enough results, I still stand ready to tackle it.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 11, 1997

I was processing some data we collected last night, through the star program to correct for dark current, and I noticed a line wrap around problem. One of my images had a aircraft come in from the right side. The line continues for a small part of the image. On the processed image, the line appears on the left and then wraps around to the right. Michael, if you would like, I can ftp the orignal and processed file to storm for you to look at.

I displayed the image with my copy of deepsky, and with fitswin95, and I believe the wrap-around is in the processed data.


John Gwinner gwinner@northnet.org
Mar 11, 1997

Nick:

> ... I was told that if I wanted
> to save the VGA screen, that I should tie into the VGA driver, and
> respond during the vertical retrace period.
Whoo, that's a mess. The PC doesn't have a really good way to tie into that retrace, and the retrace period is highly variable from PC to PC anyway (depends on what the refresh rate of the screen is). Although you can with some technologies (DirectX) get at the retrace (for frame buffering) you're talking messy programming.

The PC clock can actually be pretty accurate ... if you use the right routines. If you're doing Windows programming, use the multimedia timers -- they are right on usually. The human ear can really detect when an audiostream isn't on, so the multimedia timers are very accurate. In DOS, you can (with the right routines) get very accurate timers also. The standard 'get tick' stuff is only accurate to 17ms I believe -- pretty coarse.

I'd build in a routine to send a few status bytes periodically instead of trying to dump a screen anyway. You can format the data.

A "watchdog" timer might be a good idea if you're worried that the remote device isn't paying attention.


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 11, 1997

>I was processing some data we collected last night, through the star 
>program to correct for dark current, and I noticed a line wrap around 
>problem. One of my images had a aircraft come in from the right side. 
>The line continues for a small part of the image. On the processed 
>image, the line appears on the left and then wraps around to the 
>right. Michael, if you would like, I can ftp the orignal and processed 
>file to storm for you to look at.
>
>I displayed the image with my copy of deepsky, and with fitswin95, and I 
>believe the wrap-around is in the processed data.

I didn't think I exported this bug but apparently I have. In some versions of "star" an extra FITS card appears in the processed header. This has the effect of shifting the image by 40 pixels. If the first 40 pixels in the first row of the processed image have consistent values of (I think) 8224 then that is indeed the problem. This has been corrected in later versions and in any case did not affect the positions of the stars in the star list.

If this is not the case then let me know, otherwise it will be corrected in the version of star going out in the next couple of days.


Arne Henden aah@nofs.navy.mil
Mar 11, 1997

Tom asked about overscan, and Norman replied that there is an empty position in the current format that could be used to store an average of X overscan readings.

Since the overscan is just that -- extra reads in each row that theoretically have no signal -- you could store the average of X overscan readings as Norman suggested. There is a finite chance that a cosmic ray could hit the serial register and contaminate the average, but I think it is unlikely. Another alternative is to replace the first set of blind columns since they don't work well anyway. I prefer the blind column replacement since we can then look at a number of overscan columns/row and have a handle on the rms scatter as well. That also keeps column 799 (0-relative) open for future expansion.

So my proposal is:

     0-13 overscan
     14-781 image
     782-793 blind/dark
     794-799 telemetry
That keeps the same format for all of the sections folks have used to date.


Arne Henden aah@nofs.navy.mil
Mar 11, 1997

Norman -- fwhm = full width at half maximum. I love acronyms :-)

I would like to hear more about the VCO drifts. Norman was quoting a value of 1% fluctuation, but I didn't hear over what time span this variation was occurring. Note that the VCO stability is a major constraint on the quality of the astrometry. If it was off 1 part in 1000, then you are off one pixel in a given FITS frame, and the RA error would be off by 14 arcsec! Obviously, you are doing better than that, but I plan on checking Tom's latest dataset to see if there are any trends in the astrometry that could be attributed to this drift.

I'm not sure the PC's clock is all that accurate. On the PCs at Lowell and at least one PC here (though we aren't a PC house by any means), I've seen drifts of 1-4min/day, or worse than 1 part per thousand. Usually, it is consistent and could be calibrated out, but perhaps not in a non-temperature-controlled environment. I think perhaps the best empirical way to adjust the VCO is to monitor the fwhm in RA and adjust for the smallest width on the central column.

Also, and I'll keep harping on this, all extractions need to have the objects time-stamped if you ever intend to do serious monitoring of variable stars. If the VCO drifts during a scan, the time/pixel varies as well. We time-stamp each row as it is stored (extra telemetry). Norman -- is the UT in the header referring to the first row of the image (even though these first n rows are from the previous fits image/scan), or the first new row?


Nick Beser beser@aplcomm.jhuapl.edu
Mar 11, 1997

Michael Gutzwiller wrote:

> If this is not the case then let me know, otherwise it will be corrected in
> the version of star going out in the next couple of days.
> 
> Mike Gutzwiller
I downloaded the version that is on your home page, and the problem did not recur. I think there are some wayward copies floating about the net (I may have obtained my first copy from storm.fnal.gov.)


Mica Wickersham rjw@crl.com
Mar 11, 1997

>   Also, and I'll keep harping on this, all extractions need to have the
> objects time-stamped if you ever intend to do serious monitoring of variable
> stars.  If the VCO drifts during a scan, the time/pixel varies as well.
> We time-stamp each row as it is stored (extra telemetry).  Norman -- is
> the UT in the header referring to the first row of the image (even though
> these first n rows are from the previous fits image/scan), or the first
> new row?

From an amateur's point of view...

It seems to me that the UT of the time-stamp of the reduced data should coincide with the middle of the exposure. This would be the UT when the row that is being read out was in the center of the chip.


Arne Henden aah@nofs.navy.mil
Mar 11, 1997

Ron Wickersham asked,

>It seems to me that the UT of the time-stamp of the reduced data should
>coincide with the middle of the exposure. This would be the UT when the
>row that is being read out was in the center of the chip. 

>Is this the best time to report as the time of the observation, or is 
>there a professional "norm" that reports the end of the observation, 
>when the row is being read?
With stare-mode observations, the usual practice is to use the starting UT in the header, along with both the total exposure time and the total 'dark' time (i.e., time between initialization of the chip and the readout. The dark time may differ from the exposure time if you are suspending and resuming an exposure due to clouds. When extracting information from this image, you usually report the observation time as UT + 0.5*darktime.

This could work with TASS images, but note that, with the current 800x896 format, each one takes 15minutes to read out. If you just used the UT for the center of the 'exposure', then the accuracy is only +-7.5mins, when knowledge of the actual central exposure time for a given row is far better than this. For some slow variables like Miras, such an error is irrelevant. For fast variables like RR Lyr stars, you would really like accuracies approaching 1second. Because of the 468sec exposure time, you don't have 1second time resolution on the light curve, but you can place the blurred mean point very accurately. Likewise, when you start scanning away from the equator, the images become banana-shaped and you can start getting even better time resolution along the trail.


Tom Droege droege@wwa.com
Mar 11, 1997

One reason I have not got the data disk out the door yet is that I keep getting good data. I am just swamped moving data from one place to the next and getting it on archive tape. Soon the moon will be looking right down the telescope and I will catch up.

I did look at a few events - four where the sky was clear - in the A data set. When sorted and searched I got four measurements for almost every object with the sigma level set at 5. I wrote BASIC code to look for BOSOs (bright objects seen once). But I did not find any. ;^(

I will wait to run more data through Mike's program until the moon is here and I get the next version of the code. This data set of four scans in V looks very nice. .1 mag sigma for most 12 mag objects and .02 mag for the bright objects. So I am optimistic that another round of corrections will help.


Norman Molhant molhant@ERE.UMontreal.CA
Mar 12, 1997

Dear Nick,

> We are setting up a poor man's remote system with the TM3GET11 software
> under control of a unix system. At the moment there is no feedback while
> data collection is going on. Ideally, I would like to receive a snap
> shot of the screen every minute or so just so I can see how things are
> going. At a minimum, the temperature, voltage, current and median
> readings would be good. If I were to add the code in the TM3GET11
> software to send data over a serial link, at what point in the data
> collection process should I add the routine? I was told that if I wanted
> to save the VGA screen, that I should tie into the VGA driver, and
> respond during the vertical retrace period.
Don't do it, that's not the way to go. Could your program cope with receiving something like 400 to 500 bytes per second from the serial link? If so, I could send each and every line (in a packetized format) through the serial link and also accept keyboard input the same way, thus providing a full remote control capability. I'll provide the source code and executable for an appropriate serial device driver (that code has been sitting in my archives for years, it was used to provide remote control of a plastic bottle factory data acquisition system).
> How far away is the Linux version of TM3GET11?
I've too many problems with Linux 1.3, so I've decide to switch to the latest RedHat distribution (that should be 2.0.2, I think) to continue development. Right now, I'm waiting for the RedHat CDROMs, ordered last week. By the way, the problems were in getting the device driver to run. I had exactly no problem with intercepting the timer ticks and doing the real time data acquisition. I hope RedHat is not as flaky as 1.3 proved to be... :-(


Glenn Gombert ggombert@infinet.com
Mar 12, 1997

I have had several requestes for the source files (*.c, *h) files that I have complied to run under windows95. I have uploaded a file called "Esource.zip" that contains the files that I have been using.

No "warrenty" is implied but I will try and answer any questions.

I've placed the source code ZIP file into the Software area of the TASS home site. MWR


Arne Henden aah@nofs.navy.mil
Mar 12, 1997

Norman wrote, in response to Nick:

>Could your program cope with receiving something like 400 to 500 bytes per
>second from the serial link?
>If so, I could send each and every line (in a packetized format) through the
>serial link and also accept keyboard input the same way, thus providing a
>full remote control capability.
This is similar to the remote observing capability I provided for OSU in communicating with their 72" telescope in Flagstaff from a remote observing room in Columbus, OH. It is easy to do, as Norman suggested, with the major problem being error recovery (what happens when things crash). Of course, at 400/500 bytes/sec, you can't transmit the raw drift scan data (which comes over at 2000 bytes/sec), but you can keep track of how things are working.

Norman - let me know what you think of 2.0.2 Linux. I've held off installing Linux on the home machine until I've heard that it is stable for development purposes (I no longer try to stay on the 'cutting edge', having had too much blood drawn in the past).


Glenn Gombert ggombert@infinet.com
Mar 12, 1997

I have just converted Emmanuel Burton's new version of "Sextractor" (ext12b.zip) over to run under Windows95 and uploaded it to Storm /incoming. It is version 12b, which looks like it has a number of new features which will prove useful in reducing TASS images.

It has not undergone extensive testing but appears to work OK on my Dell Notebook computer running Windows95. No "implied warrenty" but I will try to answer any questions if I can. I plan on using to reduce the "A" and "B" fields from the TASS data beging collected here in Dayton.

I will try and put together a TASS technical note in the next few days about the new features of Sextractor and how they might be of use.

I've placed the executable ZIP file into the Software area of the TASS home site. MWR


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 13, 1997

> What say, friends?
> Should I program a servo loop adjusting the VCO to the PC clock (and you 
> would specify the row shift period in time units, say tens of microseconds 
> or something like that), or should I program a servo loop adjusting for 
> minimum column-wise star width (and you would specify the initial value in 
> DAC units like you do now - there would also be a way to freeze that servo 
> in order to get dark frames etc)?
> I'd rather prefer the first way (using the PC clock), but I'd like your 
> opinions.
I vote for using the PC clock. There's too many ways a program looking at star sizes could be fooled. Especially since the program wouldn't know whether the VCO was too fast or too slow and would need to "experiment" to find the correct solution. I would also suggest using the shift time in the tass.rc file as the value to use in the program. That way adjustments could be made for individual PCs.

The change should only be made between files so the astrometrics are consistent for an individual file.

>> If the VCO drifts during a scan, the time/pixel varies as 
> well. > We time-stamp each row as it is stored (extra 
> telemetry).
>
> OK, we have the 800th pixel for that.  It's 16 bits only, so what should I 
> put in there?
Again change the VCO only between files and put the result in the FITS header. Save pixel 800 for something else.
> Another think: as I'm going to make changes in tm3get11 (it will become 
> tm3get12, by the way), why not change also what's wrong with the 
> header? I've been told that the RA and DEC in decimal degrees cause 
> problems, so my question is: what would work well here?  What units and 
> format should I use for RA, DEC and increments thereof?  What else 
> should be changed, added or removed from the FITS header of the raw 
> files?
I would say leave units alone. The "star" program relies on the validity of the RA in the header to determine where to start looking for matches to the catalog. The last time I looked the RA was being calculated correctly but the value of CDELT1 (delta dec) was of the wrong sign.


Arne Henden aah@nofs.navy.mil
Mar 13, 1997

Norman asked for input regarding changes to the next version of tm3get, and one area of changes was in the fits header. I've gone back through images taken at several professional sites, and want to suggest some additions and changes that should be straightforward.

CHANGES
  change TIME to UT or UTSTART.  These are more standard.
  change EXPOSURE to EXPTIME.  This is more standard.
ADDITIONS
   RA      = 'hh:mm:ss.ss'  / usually central RA for the image.
   DEC     = '+dd:mm:ss.s'  / usually central Dec for the image
   EQUINOX =         2000.0 / for the above coordinates
   FILTER  =           'V'  / for this camera
   SCALE   =          13.7x / for this camera, arcsec/pix
   GAIN    =            2.5 / for this camera, electrons/adu
   RDNOISE =           20.0 / for this camera, adu
   IMAGETYP=       'OBJECT' / for this frame. choices are OBJECT, BIAS, DARK,
                        SKY FLAT, DOME FLAT.
   BIASSEC = '[1:14,1:896]' / overscan region
   DATASEC = '[15:782,1:896]'  / image area for this frame
   DARKSEC = '[783:794,1:896]' / dark pixels for this frame
Plus the appropriate white space to get things lined up. The RA,DEC,EQUINOX fields are there for ready identification by a human; the WCS information is what would be used by the computer. The filter is normally associated with the camera, but adding it here makes it easier for a human reader (and I've already noticed a couple of cases where filters have been changed on a given camera for various tests). SCALE is just a constant for helping a non-TASS person understand the camera. GAIN and RDNOISE are used in determining photometric errors and for using psf fitting. We don't know what these values are yet, but they can be determined for each camera. IMAGETYP was missing from the TASS frames I was examining, and is used in automated reduction. The three image sections (BIAS,DATA,DARK) are used for automated reduction and presumably would remain constant.

Any comments?


Chris Albertson chrisa@wavenet.com
Mar 13, 1997

> The change should only be made between files so the astrometrics are
> consistent for an individual file.
I thought the problem was that the VCO drifts even within one file. Didn't Norman say he saw line to line diferences in the scan period?

The perpose of setting the VCO frequenty is to keep the scan line period constant.


Arne Henden aah@nofs.navy.mil
Mar 13, 1997

Norman asked if TASS should use pixel 800 as a counter or time stamp for each row (this was based on my comment that FASTT timestamps each row). Since the exposure time/row is on the order of 468 seconds, and the VCO is at worst 1part/1000 accuracy, then if you use a known UT for the start of a frame and just offset by clockrate*rownumber, you get sufficient accuracy. I don't think, for TASS, there is sufficient justification to use the single remaining engineering telemetry 'pixel' for timestamping.

However, you should be able to adjust your PC clock to a few seconds accuracy so that the starting UT is reasonably correct. An accuracy of +-4 seconds would give about +-0.0005day error in the JD, which is about right considering the length of the exposure and the expected periods of the variable stars to be discovered. As mentioned earlier, I would suggest that every object extracted from the TASS frames be timetagged at the 0.0001 JD level.


Tom Droege droege@wwa.com
Mar 13, 1997

I would not expect the VCO to drift very much in any one 15 minute period. The main thing that causes it to change is temperature. The usual reason that one might see any drift is that you have just turned the computer on and it is warming up.

A few simple rules should keep the VCO pretty constant.

  1. Turn the computer on two hours before you start taking data.
  2. Don't subject the computer to drafts. Like having it sit next to a door that opens to the outside. Or have it sit directly near a heat/cooling vent.

The noise an jitter should be nearly invisible. i.e. no short term variations. This because there is a big integral involved in the count down circuit.

I have looked at this long ago with a counter timer. I did not see anything short term within the accuracy of the test equipment.


Alain Maury maury@ocar01.obs-azur.fr
Mar 13, 1997

I was quite skeptical when Tom designed his camera around a VCO. We are using the quartz clock of a microcontroller to provide for a reference. The effect of an unstable source clock is to "stretch" the right ascension scale, i.e. it might provide for a poor astrometry, or at least a non repeatable one. I think that the criteria has to be better than just "getting round stars". I do believe that for Mark II it is OK ( or barely OK ? ) to use a VCO, but clearly for a longer focal length telescope, this is an important point to take into account. A stalibilsed quartz source or any other system stable to 10e-6 would be preferred. I don't know what USNO is using as a reference clock for their scans, but for example the Observatoire de Bordeaux transit circle is using a rubidium clock, in order to make really high precision astrometry. We are about to change our system, since we found that at a given level of precision everything changes, even the telescope "at rest" is not at rest really, it keeps flexing mostly during the beginning of the scan exposure. Then the quartz clock is not stable enough. So we are going to get a 5 MHz fed into the system coming from atomic clocks at the other end of the Observatory.

I believe a stabilised quartz should be OK for the next version. I do believe also it will be better than the PC clock.

The first camera Christian Buil got running in scan mode was using an internal timer of the PC. If you have a detailed description of PC's hardware you will see that somewhere ( don't remember of it right now but could look it up ), there is a free running counter which you can not program but which you can read on the fly. I guess this is the one which generates the 18.2 something interruptions by seconds. The idea is to know that you need such and such time interval, and you read this timer with a small assembly language routine in order to see how much has elapsed since the last time you read it. Say if you need a time interval of 0x1000, you read the timer and if you suppose that it is at 0x9000, you are going to read it until you have passed 0x8000, then watch for 0x7000, 0x6000 and so on. An assembly routine can read the timer within a few counts of precision, and it works well from what I have seen.


Joe Ronchetti jronchetti@earthlink.net
Mar 13, 1997

Which enviroment and which software is the group as a whole using. For a while I thought that linux and IRAF was the choice but now it seems that win95 and sextractor is the choice. Are these exclusively different softwares or are they similar but on different OS's?

If someone could clear up this issue for me it would be appreciated.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 13, 1997

How difficult would it be to a quick alignment capability to TM3GET12? One reason it is taking so long to align our telescope is that we must wait for the telescope to clear and then ramp up before we can see if the camera must be adjusted for alignment. Even after that, we must judge by eye if the rotation was too much or not enough. Is there a way to readout how much alignment error we have?


Michael Richmond richmond@astro.princeton.edu
Mar 13, 1997

These are two very different tools. IRAF is a giant collection of many different "packages", which are designed for specific sorts of analysis. There are "packages" for astrometry, for photometry (many of these), for spectroscopy, for X-ray images, etc. It's very complex, with many different parameters which the user may adjust for each task. The learning curve is steep. Don't even think about trying to modify the source code :-) On the other hand, it already does almost anything you want.

Sextractor is a small program that analyzes "reduced" optical images. It does one thing. It's simple, the source code is small, and one can imagine modifying it. Just yesterday, I wanted to know what sort of algorithm it used for centering -- and found it in about 4 minutes, without looking at the manual, just by reading the source code.

If you want to tinker, or spend less than 30 minutes trying to figure out how to use the program, choose Sextractor. If you want to do many different sorts of analysis, including the dark-subtraction and flatfielding, and you like cool interactive windows and graphs, choose IRAF.

Just my opinions.


Herb Johnson hjohnson@pluto.njcc.com
Mar 13, 1997

Arne has kindly referred me to his previous work on TASS and dark pixels, and has noted the first ten or so pixels do not behave as dark pixels should. My apologies for initially stating otherwise. Although they are "old ground" I will look at them further, and the more legitmately "dark" pixels at the other end, in preparation for further work on problems of pixel timing and on sky background. Tom has noted to me that sky background dominates dark current, which is true. But dark current is more easily determined, and whatever tools and analysis I develop for the former will help with the latter. I have some other experiences with backround removal that may produce something new to us later.

in a following message

*>> An obvious short-term solution would be to get a surplus pulse generator
*>> of the desired stability. As this is sub-audio, you might get one at
*>> a hamfest for tens of dollars. Meanwhile, you software guys, I can give you
*>> a cheap UNSTABLE version in a week from Radio Shack parts, suitable
*>> for software development, for the cost of a cheap lunch. Let me know.
Chris Alberson asked me: "How about this: build a "relative clock"":
*>Take a 10Mhz Xtal temperature compensated Ocilator and divide it by 10
*>then feed it to a 16 bit counter.  The counter output goes to a 16 bit
*>port on the AT buss.  Now all the software needs to do is copy the
*>content of the 16 bit port to the last column of each scan line.
(Later, Norman Molhant suggests a different divider rate and refines the concept.)

Chris notes this avoids the problem of reading a pulse on demand, requring some kind of interrupt. Makes sense to me. A generic 24-bit I/O card for the PC ISA bus can be bought for about $45. A divider could be added to it, again driven by the aforementioned stable source only at a higher frequency. Before I look further, is THIS of interest: a somewhat programmable, stable counter I/O card?

Other provisos; Norman says "he's out of slots in his Pentium". I tend to think of a TASS PC as a seperate computer, probably some old 386 system. "One computer per task" is my preference, your mileage may vary. He suggests counter resolutions toward .1 millisecond. 16 bits offers 64000 counts, that would be 6400 milliseconds before wraparound.

But this idea is approaching the point where it's "cheaper" to fix the Mark III cards than to develop elaborate work-arounds....Mica Wickersham suggests he'll look at this, I've offerred him some suggestions privately. Meanwhile, the software to process a PC's internal counters could likely be adapted to read an I/O port for a custom counter card.


Tom Droege droege@wwa.com
Mar 13, 1997

It is already there in tm3get11 The "F" key gets you the point spread function. You don't have to wait the full 470 seconds. You can tell pretty quick if you have done the right thing. Remember, you get to make three adjustments at once. I usually set the VCO way off. 10% or so. This gives nice long diagonal lines for the stars. Then you can rotate the three cameras at once and see how they change. Four or five trips to the camera gets everything rothated. Then you just have to make a few changes in VCO speed and plot them to figure out where the minimum is. The F key plot give a pretty quick indication of the direction that you have made the change, so you only have to wait out the full 470 seconds when you are very close.

I find I can usually do rotations and focus adjustments at the same time.

Don't be too fussy. You will get to as good as the camera can be in an hour of so of adjustment. You can do this any time there are holes in the clouds, so you don't have to waste a good night. Please don't try to make a silk purse out of this sow's ear. An attempt at a perfect focus may be appropriate where the optics is perfect, but these camera lenses are not that good.


Glenn Gombert ggombert@infinet.com
Mar 13, 1997

I have found that the instructions that Tom has shipped with his Mark III cameras actually works fairly well. The key is to get the long streaks (with an "off-tuned" VCO) perfectly aligned in the N/S direction. A fraction of a degree can cause the tilt to show up after a full-frame has been read out from the camera.

It takes some practice to get the "hang" of it but you get to know what to look for as you experiment with different settings.


Glenn Gombert ggombert@infinet.com
Mar 13, 1997

I am in the process of putting together a TASS Technical Note on the Sextractor program, it salient points, and the new features that the latest release has to offer.

It is optimized to be used on large scale astronomical survey projects such as the TASS project, it typically reduces a TASS image in 30-40 seconds on my Dell Note book computer with litte difficulty.Several AJ artciles have been written about data reduction projects that have been used by folks other than (other than the author), you can see some of these by doing an Alta-Vista search on "Sextractor".

It also has the advantage of being able to calculate RA and DEC directly once the World Coordinate System Key Words have been added to a FITS image header. It is very easy to operate (just feed it TASS files) once the parameter file has been setup for TASS image parameters. (I have posted several examples of the parm's file that I have been using to reduce TASS images here in Dayton).

I am well along on a GUI interface to replace the current paramter file which should make it even easier to use.

Alain Maury has been using my version on Windows95 platforms that run considerably faster that his UNIX workstation (perhaps he could comment on this??) He has some very large scan files that have been reduced with the Win95 version that I uploaded to Strom last Fall.


John Gwinner gwinner@northnet.org
Mar 13, 1997

I did some research as promised. I'm attaching some source for getting sub-ms accuracy directly from the CMOS hardware.

However, that's NOT the most accurate timer. On a Win32 platform the Win32 function QueryPerformanceCounter() is QUITE a bit better (0.8 microseconds, or 0.0008 ms). I haven't tested Win95 yet, it might actually be a bit slower as it's threads are mixed with 16 bit code. In any event, it's possible to get at least 1ms out of Win95 with the multimedia timers. This is without writing a VxD (Virtual device driver) which would be the right way to do it anyway.

The multimedia timers I talked about before are in fact accurate to within 1 millisecond. Although they are called "Multimedia timers" all you have to do is call the right function, you don't have to actually play MIDI files or anything (MIDI requires sub ms accuracy).

Again, the PC CLOCK will NOT be as accurate, because some software (LEMMINGS is an example in the Microsoft knowledge base, interestingly) is rather poorly behaved and makes the CPU miss the 'timer tick'. Any ISR that delays longer than 18ms will make this timer tick miss. Recently, some 3D video cards were guilty of this, so it's not just old software. But if you use the hardware directly, you get the above accuracies.

The confusion comes from equating the PC clock (i.e. time at the DOS prompt) with the CPU timer tick. The latter is accurate, the former isn't always. On CPU's that support timer services, we can get to the microsecond area via the QueryPerformanceCounter().

Furthermore, don't forget, the CPU timer is pretty accurate, as it's got to refresh the DRAM on the motherboard. After all, the main processor clock is crystal controlled of course.

LINUX on a PC would use the same hardware with the same limitations of course -- I'm not sure what the Linux equivalent to QueryPerformanceCounter() is. DEC Alpha PC's seem to have the same or better counters on board that are accessible through QueryPerformanceCounter().

Now, don't get me wrong, I'm not against building a better timer on the TASS hardware cards, I just wanted to check out the PC accuracy. Why build extra parts if we don't have to? Still, might not be a bad idea. I don't know how accurate the PC Xtal clock is when the temperature swings, although the power supply would keep it a bit warm.

P.S. Caution about the source code below: it's quite old, and I note 4.77Mhz being used in a couple of spots -- I'd recommend experimenting with it before you use it (if you don't want to use QueryPerformanceCounter()).

Look at the source code and explanation.


Glenn Gombert ggombert@infinet.com
Mar 14, 1997

The World Coordinate System Parameters are calculated by selecting 25 brightest stars from the FITS image and the HGSC region that the image occupies, then a "pattern match" between the two is utilized to calculate the WCS parameters from the image.

A routine to do this can be found in the Universtiy of Iowa's ATFTools Kit. The process is decribed in TASS Technical Notes #22/#23. It does not requires that the stars generated from the star-list be "matched" with another program. It allows the RA/DEC to be calculated for each object that is "extracted" from the image.

If you have any more question let me know.


Marty Pittinger PittiMJ1@central.ssd.jhuapl.edu
Mar 14, 1997

We did some spring cleaning a few weeks ago; correcting the problem with the nuts glued to the wires, checking clamps on the water hoses, dusting and cleaning the lens. We placed TASS 5 a.k.a. 6 (since I'm not sure which one we have) back into our enclosure and waited for clear skies. Everything worked except the images from the West pointing camera, which seemed to have no real focus point. I get fuzzy rings or fuzzy dots (or like a snow flake). I suspect it's frost inside the CCD chamber, but the hose that connects the desiccate tube fell off and moisture may have entered.

Tom, do you pull a vacuum on the drying system? Can moisture easily enter the CCD chamber? Have you ever experienced frost inside the chamber ?

I also suspect a problem with a TECs. When starting up the TASS, CCD Temp remains at 0.65 until the clearing routine is completed, sometimes it lasts through the Ramping Up phase. This is unusual, normally the CCD temperature starts dropping immediately, eventually settling at -10 deg C. It appears one of the TECs doesn't kick in immediately and since measuring system does an average temperature you would get this reading, if the following is true.

Conditions: X = TEC at -10 deg C. (OK) & O = TEC at ambient (or failing to cool)

E - S - W = East / South / West CCDs

X - X - X = -10 deg. = Average of East (-10) + South (-10) + West (-10)
O - X - X = 0.0 deg. = Average of East (-10) + South (-10) + West (+20)
O - O - X = +10 deg. = Average of East (-10) + South (+20) + West (+20)
O - O - O = +20 = No TEC or Average of East (+20) + South (+20) + West
(+20)
This would indicate one (1) TEC is a bit flaky. Any help would be greatly appreciated.


Arne Henden aah@nofs.navy.mil
Mar 14, 1997

Alain wrote:

> ...I don't know what USNO is using as a reference clock for their scans

We get our absolute time (to about a millisecond) from GPS, and then use that to update the internal clock in the Silicon Graphics about every 5 minutes. The internal clock gives us microsecond resolution, sufficient for our scan rate generation.

For the fast CCD system I built with Mark Wagner at OSU, we use a pair of internal boards in the PC. One contains the power and mechanical hooks to mount a small GPS receiver. The other is a small micro that keeps a counter of GPS time, and holds the count in registers when an external trigger occurs. We use this system to time-tag rows on the MMT CCD controllers after the fact using IRAF, so that we don't have to modify existing software. The external trigger in this case is the start of a new row. For the high time resolution CCD system, we are not interested in sidereal rate generation, but just to clock the CCD at the fastest possible rate.

Note that there are several schemes of getting accurate absolute time into a PC, including the 2board setup above as well as WWV clocks, modem time signals, NTP protocol across the Internet, etc.

John Gwinner posted a long note on accessing the 8254 counter/timer chip in the PC. Yes, it has good time resolution, but I am not sure of the basic accuracy or stability since it just uses the on-board crystal with no temperature compensation. I've never seen where anyone has checked this!

Herb suggested building up a timer on an I/O card for the PC. There are many of these available commercially, ranging in cost from $50 up. The cheap cards use a crystal oscillator fed into an 8254 or a 9513 counter chip; the expensive ones use a temperature-compensated oscillator. This would of course reguire another slot, which seems to be a rare commodity in some computers.

Tom's suggestion of changing the VCO at most once per frame is good, though I wonder about the 20 extra scan lines at the end of one frame (since they are supposed to be at the beginning of the next, and therefore supposed to have the scan rate quoted in the fits header). Perhaps the solution is to use the 800th pixel and store the current VCO rate there, and allow it to vary even within a single frame. Then decide how you want to determine the new rate (external frequency counter, fwhm calculation, whatever).

I would still like to see the magnitude of the effect. How much do you change night/night, and during a night? If the effect is real, it may be of low-enough magnitude that it only affects the astrometry (since your images are huge), and can be calibrated out.


Arne Henden aah@nofs.navy.mil
Mar 14, 1997

There are a couple of tests that I would like to see, but since I don't have a TASS camera I am asking someone who does have one to make them.

The image size question still remains. Michael has taken some data with one of his cameras 'stopped down' by using an aperture mask. An expansion of this test would be useful, perhaps using aperture masks on both an I camera and a V camera, and varying the size of the mask. This will tell us if stopping down the current lens/filter configurations will improve the images and by how much. Second, I would like to see someone try an R filter instead of an I filter to see if using a passband within the corrected range of the lens will help (note that this also has the potential of giving TASS V,R,I as an option). Third, if someone has a narrower passband filter, such as a Stromgren or Gunn, around 500-600nm, it would be interesting to see if narrowing the wavelength range improves the images. Another test would be to replace one of the lens with a lens of supposed better quality (like a true Pentax or Nikon lens), and see if the images get better.

The flatfielding issue is unresolved. I would like to see someone take some data during twilight, at about half-saturation in the camera, to remove stars and to make local light sources less important. There were two other flatfielding schemes proposed: a white card placed in front of the lens, reflecting either twilight sky or moonlit sky; and a diffusing screen/glass placed in front of a lens, with any light source (including the night sky) imaged through it. Flatfields taken these ways can be compared with median flats from scans to see if there are overall gradients that impact on the photometry.

I will be happy to reduce the data from any of these experiments.


Michael Richmond richmond@astro.princeton.edu
Mar 14, 1997

I saw an announcement in sci.astro.research today from the Planetary Society (a private organization which supports planetary research). They are planning to award grants ranging from $1,000 to $25,000, and will accept proposals in 3 categories:

  1. Observational
  2. Theoretical (they call this "research")
  3. International collaboration in NEO observations
It strikes me that _if_ there are one or more TASS members _seriously_ interested in using Mark III or Mark IV cameras specifically for finding NEOs, then they _might_ want to get together and submit a single proposal.

There appears to be no description of this NEO Grant Program on the home page of the Planetary Society:

     http://www.transatlantech.com/TPS/TAhome.html
but you can write for more information to:
           The Planetary Society
           65 North Catalina Avenue
           Pasadena, California
           91106-2301
           Phone: (818) 793-5100
           Fax: (818) 793-5528
           tps@mars.planetary.org
Please, please think CAREFULLY before you post replies to this message! People who apply for these grants must be willing to spend LOTS of time coming up with a coherent plan, and also find a good reason to ask for money. It will be a lot of work.

Personally, I have no plans to apply -- I don't NEED the money to do what I want to do. Everyone has a unique situation, of course; but I suspect that, as a group, we'd get more done by ignoring tnis announcement and just working hard for the next few months.


Norman Molhant molhant@ERE.UMontreal.CA
Mar 14, 1997

Dear Marty,

> Everything worked except the images from the West pointing 
> camera, which seemed to have no real focus point.  I get fuzzy 
> rings or fuzzy dots (or like a snow flake). I suspect it's frost 
> inside the CCD chamber, but the hose that connects 
> the desiccate tube fell off and moisture may have entered.  
> 
> Tom, do you pull a vacuum on the drying system?
> Can moisture easily enter the CCD chamber?
> Have you ever experienced frost inside the chamber ?
I would suggest getting the camera back inside at room temperature, then separating the objective from the camera body (4 screws to unscrew), letting both of them warm up for a few hours, then re-installing the dessicant and tubing, and pumping the camera body to a vacuum for a few hours, before letting air in via the dessicant and re-assembling the camera. May I suggest you try this in as dry a place as you can find? Universities are full of dry places, so finding one should not be too difficult.
> I also suspect a problem with a TECs. When starting up the TASS, 
> CCD Temp remains at 0.65 until the clearing routine is completed, 
> sometimes it lasts through the Ramping Up phase.
> This is unusual, normally the CCD temperature starts dropping 
> immediately, eventually settling at -10 deg C.
> It appears one of the TECs doesn't kick in immediately and since 
> measuring system does an average temperature you would get 
> this reading, if the following is true.
Dear Tom, are there NTC's on each CCDs (connected in series or in paralle?), or is only one of them equipped with an NTC resistor? If more than one, then the equation to convert NTC readings to temperature is wrong, either on serial #0, or in all others.


Arne Henden aah@nofs.navy.mil
Mar 14, 1997

I'm going to the pulsation meeting in Los Alamos in June (close to the time of the AAS meeting), and will be presenting a paper on the FASTT variable star survey.

At the same time, since TASS is covering SMSP-A, which happens to lie inside one of the FASTT regions, I would like to include V and V-I data from the TASS scans for the variable stars discovered by FASTT.

Does anyone have any objections to this? I think it is another avenue of publicity for TASS.


Herb Johnson hjohnson@pluto.njcc.com
Mar 14, 1997

*> There are a couple of tests that I would like to see, but since I don't
*> have a TASS camera I am asking someone who does have one to make them.
*> ...
*>  The flatfielding issue is unresolved.  I would like to see someone take
*> some data during twilight, at about half-saturation in the camera, to remove
*> stars and to make local light sources less important.
*>  I will be happy to reduce the data from any of these experiments.
I am also gearing up to work on this problem, and I too would like to see some attempted flat field images. It's tough to produce these in an even way. Twilight might do it, if you create a mask for the camera so it does not saturate. Because of the scan mode of operation, the mask must be dynamic. COnsider a rotaing circle with a "wedge" cut out of it: the wedge reduces exposure by the ratio of its arc to the circle. If it rotates "much" faster than the pixel row clock, there will be no aliasing effects. This sounds odd as hell, but it's a simple scheme to construct. Good Luck!


Glenn Gombert ggombert@infinet.com
Mar 16, 1997

Below are the "A" & "B" fields that I have identified (so far) in the Dayton TASS data. Since all the data is archived on tape I had to sift through it over this weekend. It took a while to search through all of the data taken so far.

Hopefully I will be able to generate calibrated star-lists from the data in the next few days for the TASS Home Page.

================================================================================

13-March-97


	Field-B		31T0521.617	RA = 128.565
			31T0521.630     RA = 135.284
			31T0521.639	RA = 138.643
	
	Field-A		31T0521.752	RA = 179.099 
			31T0521.761	
			31T0521.770	RA = 185.851

	Field-A		30T0521.773 	RA = 187.35
			30T0521.742	RA = 190.72

	Field-B		32T0521.649	RA = 127.00
			32T0521.658	RA = 130.378
			32T0521.667	RA = 133.737

	Field-A 	32T0521.817	RA = 187.74
			32T0521.826	RA = 191.119
			32T0521.808	RA = 184.36

11-March-1997

	Field-B		31T0519.601	RA = 128.31
			31T0519.606	RA = 122.76
			31T0519.626	RA = 131.68

	Field-A		31T0519.775	RA = 185.53
			31T0519.784	RA = 188.92
	
	Field-A		30T0519.737	RA = 187.07
			30T0519.747	RA = 190.44	

	Field-B		32T0519.654	RA = 126.77
			32T0519.663	RA = 130.14

	Field-A		32T0519.822	RA = 187.42
			32T0519.831	RA = 190.80
09-March-97

	Field-B		31T0517.626	RA = 129.737
			31T0517.635	RA = 133.092
	
	Field-A		31T0517.785	RA = 187.102
			31T0517.756	RA = 190.949

	Field-A		30T0517.747	RA = 188.562
			30T0517.756	RA = 191.949

	Field-B		32T0517.663	RA = 128.17
			32T0517.672	RA = 131.54
			32T0517.682	RA = 134.917

	Field-A		32T0517.822	RA = 185.647
			32T0517.831	RA = 189.031


07-March-97

	Field-A		31T0515.787	RA = 185.83
			31T0515.796	RA = 189.21

	Field-A		30T0515.749	RA = 187.333
			30T0515.758	RA = 190.71

	Field-B		32T0515.674	RA = 130.368
			32T0515.684	RA = 133.735

	Field-A		32T0515.824	RA = 184.35
			32T0515.833	RA = 187.729
			32T0515.843	RA = 191.109

15-Feburary-1997

	Field-B		31T0496.681	RA = 129.181
			31T0496.691	RA = 132.548
			31T0496.700	RA = 135.91

	Field-B		32T0496.728	RA = 131.018
			32T0496.737	RA = 134.386


28-January-1997

	Field-B		31T0478.727	RA = 127.742
			31T0478.736	RA = 131.122
			31T0478.745	RA = 134.502

	Field-A		30T0478.848	RA = 186.72
			30T0478.858	RA = 190.109

	Field-B		30T0478.689	RA = 129.23
			30T0478.699	RA = 132.61

	Field-B		32T0478.773	RA = 129.642
			32T0478.783	RA = 133.026


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 17, 1997

I finally finished this round of changes to the star program. I have also moved to a new ISP so the URL has changed to:

http://home.fuse.net/deepsky/tass.html

The changes to star include:

Let me know if there are any question or comments.


Arne Henden aah@nofs.navy.mil
Mar 17, 1997

I've uploaded to storm.fnal.gov under /incoming/smspbstd.tex a list of secondary standards for the SMSP-B field. This file is in latex format; someone (Michael?) can edit out the TeX commands if you want to use it as input to a reduction program.

The catalog is available at http://www.astro.princeton.edu/~richmond/tass/catalogs/catalogs#smspb

These standards are in three 11x11arcmin regions in SMSP-B. They range in brightness from V=8.6 to V=17, and so should be good to use in determining faintness limits on the TASS frames. The astrometry is good to about +-0.1 arcsec; the photometry is typically 0.01mag accuracy, using a 12arcsec aperture. They were determined on the basis of three consecutive photometric nights on the 1.0m telescope. I don't think there are any variables lurking in this set, but to do really good secondary standard work, you would need to take one or two additional photometric nights spaced from the original three to remove long period variables. Also, I do not have very many 'bright' stars in this set; I could have selected my fields a little more carefully to give you a few more brighter stars. Maybe next time!

I have started getting standards in the SMSP-A field, and hopefully will have results for you after the April observing run.

in a subsequent message

After observing the three SMSP-B photometry fields and thinking about TASS photometry in general, I have a suggestion for all writers of extraction software.

In comparing photometric data sets, you have to be careful that you are not comparing 'apples with oranges'. Especially with the large images from the current TASS setup, there will be image blending at different scales. Magnitudes calculated from psf-fitting will be different than aperture magnitudes, which in turn vary depending on the size of the aperture. When combining datasets, it is best that the magnitudes used are all from the same algorithm.

Now, I am not going to judge that one extraction program is better than another. What I am saying is that we need to have ALL extraction programs output a common magnitude for comparison, perhaps in addition to any magnitude scheme that the program usually provides.

What I recommend is, for the next few months until the AAS and LANL meetings, the extraction programs provide at least the following aperture magnitudes: aperture diameters of 2,3,4,5 pixels, with annular sky at 6-9pixels. Sky should be determined by the daophot 'mean-median' formula (sort sky pixels, average the 10percent of pixels around the median to form a mean). The four aperture sizes will cover the best to worst images I've seen reported.


John Gwinner gwinner@northnet.org
Mar 17, 1997

I just wrote a small console program to time some things, so ironically I needed the timer routines myself . I normally use WindowsNT, haven't check the following under Win95 yet.

All I did was to #include at the top of my CONSOLE program (that's right, no WinMain and all of that). Then I did the following:

		LARGE_INTEGER liCount, liEnd;
		QueryPerformanceFrequency( & liCount );
		printf("Performance Frequency: %I64d\n", liCount);

		QueryPerformanceCounter( & liCount );
		QueryPerformanceCounter( & liEnd );
		printf("Performance Count for Queryx2: %I64d\n",
                                   liEnd.QuadPart-liCount.QuadPart);

Interestingly, the timer resolution was 179,660,000 which is pretty good! (PentiumPro, 180Mhz, I'll also run it on a P-90 when I get a chance. I may be able to put the P-90 outside to really acid-test it.

If your processor can't handle the I64 word, the LARGE_INTEGER is a union, so you can break it down.

I'll try this under Win95 later this week, but it looks like all that would be required is to include that header and then get better timer resolution.

Again, this is a console app, no GUI.


Chris Albertson chrisa@wavenet.com
Mar 17, 1997

John D. Gwinner wrote:

>   I just wrote a small console program to time some things, so ironically I
> needed the timer routines myself .  I normally use WindowsNT, haven't
> check the following under Win95 yet.
> 
>   All I did was to #include  at the top of my CONSOLE program
> (that's right, no WinMain and all of that).  Then I did the following:
> 
>                 LARGE_INTEGER liCount, liEnd;
>                 QueryPerformanceFrequency( & liCount );
>                 printf("Performance Frequency: %I64d\n", liCount);
> 
>                 QueryPerformanceCounter( & liCount );
>                 QueryPerformanceCounter( & liEnd );
>                 printf("Performance Count for Queryx2: %I64d\n",
>                       liEnd.QuadPart-liCount.QuadPart);
> 
> Interestingly, the timer resolution was 179,660,000 which is pretty good!
> (PentiumPro, 180Mhz, I'll also run it on a P-90 when I get a chance.  I may
> be able to put the P-90 outside to really acid-test it.
I hate to argue but how does the above tell you anything about accuracy, timmer stability or short term jitter? It only tells you that the numbers run fast. The problem to be solved is to measure a time interval of approximtely one second to an accuracy of about 0.1Msec or better and get repetable results day after day even when the room temperature changes.

I think it is an interesting problem but maybe not do-able with standard PC hardware.

BTW the equivalent system call in UNIX-like OSes (Linux is one example) returns time in u-sec as an integer but is only documented to be accurate to within 10Msec. Pretty poor. Some UNIXes may be better then 10Msec but if you depend on undocumented accuracy then your code is not portable.

The same coulbetrue of PC hardware. _My_ P100 system may have a very good Xtal oscilator with temperature compensation but some other PC could use something much simpler.


Chris Albertson chrisa@wavenet.com
Mar 17, 1997

Herbert R Johnson wrote:

> This is another report in a series of spot checks of Tom's sample images.
> 
> **Method: for each file, I peeled off columns 790, 791, 792, 793. Columns
> were counted from 0, and values converted such that 32768 --> 0.
> Even if we debate the "darkness" of the dark pixels, these are undenyibly
> dark in this file.
> 
> For each column, I computed the mean (sum/rows), median (max/2 -min/2),
> standard deviation of the mean. I ALSO computed a mean for groups of 50
> rows, and computed (mean of all - mean of 50). I reported this difference for
> the first 50 rows, the next 50 rows, etc. Any drift would show as a change
> in this difference. I printed all the differences in order, then eyeballed
> them for any pattern.
> 
> **Results: All three files shows a consistant drift in the mean of 50 pixels
> (relative to the mean of all pixels in the column) from minus 1.5 standard
> deviations, through zero, to plus about .75 standard deviations. Again, this
> is a drift in the difference between the mean of 50 rows and the mean of
I looked at this same data a while back. I had not heard of "VCO drift" back then but was looking fro drift in the dark current. Sure enought I found about what you did. I posted a few graphs and a couple paragraphs of test.

What I got back from people who looked was that the CCD was temperature cycling with a period just shorter than the length of the raw data file. I was also told that the dark frames were taken early in the night perhaps before the CCD had reached a stable temperature.

Try taking a look at the stability of the dark (non-imaging) pixels of files posted more recently. Some of these turned out much better (flatter)


Peter McCullough pmcc@astro.uiuc.edu
Mar 18, 1997

For accurate time on a PC, a GPS card (US$600 last time I checked, but perhaps cheaper now) is a good option. If you only need time accurate to +-1 second and precise to +- 0.01 second, the shareware program rightime (downloadable from the WWW) does a reasonable job.


John Gwinner gwinner@northnet.org
Mar 18, 1997

> I hate to argue but how does the above tell you anything about accuracy,
> timmer stability or short term jitter?  It only tells you that the
> numbers run fast.  
True, but up to this point the general consensus with the list was that timer accuracy on the PC was only to 18ms. Obviously, it's better than that with the free hardware that's supplied, but I agree that there are other questions to be asked.
> BTW the equivalent system call in UNIX-like OSes (Linux is one example)
> returns time in u-sec as an integer but is only documented to be
> accurate to within 10Msec.  Pretty poor.  
>  ..
> The same coulbetrue of PC hardware.  
It's not; all PC hardware that's MPC level I certified (most any PC sold nowadays) has to be able to handle MIDI streams, and they have to be accurate to about a ms. (millisecond as opposed to Ms = Megasecond). That's why the multimedia timers (as I previously posted) are pretty good timing services if you're in a GUI state.

The important thing with the snipped I posted is that it's NOT a GUI .. therefore, the tm ... program could use these timing services.

It's FAR better than the timing tick. So this is moving forward even if not all the questions have been answered.


Arne Henden aah@nofs.navy.mil
Mar 18, 1997

Not to be picky, but I hope folks are not confusing 'accuracy' with 'resolution' or granularity.

John Gwinner wrote:

>True, but up to this point the general consensus with the list was that
>timer accuracy on the PC was only to 18ms.  Obviously, it's better than
>that with the free hardware that's supplied, but I agree that there are
>other questions to be asked.
The PC timer tick has a resolution of 18ms, but the accuracy is much better than that (i.e., the tick occurs every 18 ms +-0.001ms on average). The various C/windows timer ticks use the raw input to the 8254, which is 1.8MHz if I remember correctly, which is why their resolution or granularity is better. The basic accuracy is given by the 1.8MHz oscillator in both cases.

The bottom line: yes, you can get better timer resolution than from the clock services. No, it is no more accurate than the VCO in the cameras.


Michael Gutzwiller Michael_Gutzwiller@AICI.COM
Mar 18, 1997

Arne wrote:

> What I recommend is, for the next few months until the AAS and LANL meetings, 
> the extraction programs provide at least the following aperture magnitudes: 
> aperture diameters of 2,3,4,5 pixels, with annular sky at 6-9pixels.  Sky 
> should be determined by the daophot 'mean-median' formula (sort sky pixels, 
> average the 10percent of pixels around the median to form a mean).  The
> four aperture sizes will cover the best to worst images I've seen reported.
I agree that all the programs should do aperture photometry. TASS psf's can be time dependent due to VCO drift and seem to be spatially dependent too.

I'm not sure I agree on the set aperture sizes you have proposed. The "star" program uses an aperture based on the size of the psf for that image. The width and height of the aperture are set to about 2*fwhm in x and y of the psf. For Tom's images and mine this represents an aperture diameter of usually 5 to 7 pixels. Anything less than this and I would worry about jitter in selecting the peak position to contribute significantly to the error in measuring the ADU sum within the aperture.

As for generating N magnitude measurements the obvious question becomes, which one should be used when comparing one image to another? Eventually you would have to pick one and always use it because of the volume of data we need to process. I would feel more comfortable if I knew the aperture was based on the size of the psf for that image so that a certain percentage of the star's light was always inside the aperture.

Calculating the sky background as a mean of the median 10% is certainly feasible and probably better than the absolute median that "star" uses now.

How do the others using SExtractor, IRAF, or other list generators feel about these issues?


Chris Albertson chrisa@wavenet.com
Mar 18, 1997

not to argue, I am proposing to do both. See below.

> I agree that all the programs should do aperture photometry.  TASS psf's
> can be time dependent due to VCO drift and seem to be spatially dependent
> too.
I agree doing PSF fitting with a constant PSF would not work to well. But Why do that? PSF fitting appears like it should work very well for TASS images and in fact may have advantages. As I see it, the trick is to have a very good model of the system PSF. TASS PSFs have these characteristics that the PSF model should capture: 1) They are not radialy symetric. 2) they are not constant over the field of view. As long as the PSF model captures this in model parameters, PSF fitting should work.

For those not read up on PSF fitting here is a overview. I have left out some detail:

  1. off line, build a good model of the system PSF. This could be based on closed form equations or tabular data or both.
  2. Extract a list of of stars by any method one likes. The list will contain lines of (x, y)
  3. For every (x,y) extracted in #1
    1. attempt a least squares fit of the PSF model to the data at point x,y.
    2. Record the peak of fitted function as (x', y'). This is the instrumental Mag. and fine-fitted possition.
    3. Subtract the PSF model from the data (removes the star from the image.)
  4. At this point we _should_ have a blank image, with all the stars removed. Of course we will not. We examine this "residual" image to see how well our detection and PSF fitting process has worked.
  5. off line, we tune and tweek the above steps until we get acceptably blank residual frames.
There are two advantages to this: 1) PSF fiting should handle "blends" very well. A visual examination of tass images shows that blends (over lapping star images) are very common. 2) There is a very good feedback mechanism, the residual frame, that allows us to see how well the process is working.

I can also see a disadvantage to PSF fitting -- It is harder to do correctly. Getting a good PSF model is not easy. One must collect sample data and build a database of measurments. A PSF should be modeled as a two dimensional function of (about) four variables.

So if I had to reduce a big pile of data today. I'd use aperture photometry as it is quick and simple but in the long run I suspect the PSF method will give better result.

Now I see there are basicaly two groups of people here on this list. Those with cameras and an immeadiate need to reduce data and those with access to only a sampling of data.

I will propose to two phased approach to TASS data reduction. 1) What has already happened put of necessity, a production mode. 2) A experimental mode. I expect people who have cameras and data to be interested in #1 while others would somehow form a coordinated effort at #2. Results of #2 could be folded back into the production effort when it warents.

Lately #1 seems to be taking shape. I suspect there are others interested in #2 but that there is zero organization of that effort.

> I'm not sure I agree on the set aperture sizes you have proposed.  The
> "star" program uses an aperture based on the size of the psf for that
> image.  The width and height of the aperture are set to about 2*fwhm in x
> and y of the psf.  For Tom's images and mine this represents an aperture
> diameter of usually 5 to 7 pixels.  Anything less than this and I would
> worry about jitter in selecting the peak position to contribute
> significantly to the error in measuring the ADU sum within the aperture.
> 
> As for generating N magnitude measurements the obvious question becomes,
> which one should be used when comparing one image to another? 
Just an idea. I think you should compare each new object list with a composite made from all previous lists. In other words it is a "merge" operation. In other words a star's location is a wieghted average of all the locations it has been.


Chris Albertson chrisa@wavenet.com
Mar 18, 1997

I read over all the postings of who has what media types. Interesting: not one flopy disk in the group ;-)

It _may_ be possible to work out a complex musical chiars media format dance (where A give media type x to B who copies to type y and forwares to C and D then D copy to z and passes it to E and so on.) I think any such scheeme would quickly break down. If there where only one or two types of media we all shared it would work but here is the results followed by some ideas:

  1. PC type 1/4 inch tapes are common but PC type QIC is very poor for data exchange as there are no real standards.
  2. Those of us who work in a big computer lab or at a university All seem to use either 8mm or 4mm tape. Problem is there is about a 50/50 split on the use of 8 or 4mm tape and home users normally don't buy these $700+ tape drives
  3. zip drives are about as common as 4mm tapes. I think only three people with these. Also they don't hold much data. Only about one night's worth or less.
  4. CDROM. Everyone has a reader. Several people have CD-R. However, CD making is not as easy as it seems. It takes between 40 minutes to one hour to make each copy and at least this much time asembling the data before the first copy can be burned. IMO CD-R can not be used in a production mode. It should be used for special purposes and runs less then about 8 copies.
  5. No one wrote about it but I suspect almost everyone has an IDE interface. IDE drives in removable drawers make for fast, low cost exchange media
IDEA: Do you all know how Internet USNET news is transfered? articles eventualy get to every newsserver (almost) but there is no central braodcast site. All sites are peers. They use a protocol called "I have, I want" and articles are exchanged between two servers. Agreemets are made bilateraly between server operators. Thousands of bilateral agreements make up a "network". It works because there are rules.

I think we could use this USNET Model If we agreed on a set of rules that realy amount to just being polite then data distribution should work without a detailed plan. Here are my proposed rules:

  1. Everyone producing data should be willing to copy and share that data. But should not be required to copy and ship it more then once.
  2. Anyone asking for data should also be wiling to copy and send data to anyone else who asks.
  3. If someone gets multiple requests for the same data he should offer it to the person best suited to perform under rule #2.

The rules do not talk about media type. That is to be worked out between each sender/reciever pair.

The key to making this work is for people to post "I have, I want" messages. This plus agreement on a set of rules like the above and we should have a regulated anarchy


Glenn Gombert ggombert@infinet.com
Mar 18, 1997

I would be happy to supply my "A" / "B" Field data to anyone that is interested. My other data is achived on bulk "Trevan-3" style magnetic tapes. It would be more difficult to send out (but it could be done possibly).

I suggest that we make the "A" / "B" filed data available to anyone that would like it , as well as any other "observing projects" that are undertaken in the future.


Tom Droege droege@wwa.com
Mar 19, 1997

This sounds good to me. The remaining thing needed is a place to post the "I haves". Seems like the home page is the place. I will try to work up a list of the "good data" that I have and will ask Michael to post it on the home page. What say Michael?

One way to do this is to just archive the yyyymmdd.tm3 files that are generated by Norman's program. There is a small problem in that without extensive editing, there is only about 50% good data on the runs that I have selected to be "good". It is a tremendous job to edit out just the good data. So I would propose to post the raw .tm3 files and let the "buyer beware" that the data requested might not all be good. One might not always want the best data anyway. If one is studying data to see how far into the clouds one can push analysis, or working on "good data" detection algorighms, then one wants some not so good data.

I would very much like to see Chris and some of the other data munchers that don't have a camera have access to the data. This seems like a good way.


Herb Johnson hjohnson@pluto.njcc.com
Mar 18, 1997

*> Now, I am not going to judge that one extraction program is better than
*> another.  What I am saying is that we need to have ALL extraction programs
*> output a common magnitude for comparison, perhaps in addition to any
*> magnitude scheme that the program usually provides.

My simple comparisons of the extraction programs on Tom's test images certainly demonstrates that these programs do produce inconsistant magnitudes. It's not only the methods that are different, it's also the photometric references. And of course, they all beg the question as to which one is most correct, as so far our photometric reference data sets are sparse. Worse, unless our images are uniform (sky conditions consistent through the run), these sparce references will not be enough. INcidently, I don't think we have a "figure of merit" for quality of sky conditions. The traditional method to minimize local conditions during sky surveys is multiple runs and averaging.

One of my intents, as I tinker with the data, is to get a feel for camera conditions before I tackle such issues.


Arne Henden aah@nofs.navy.mil
Mar 19, 1997

Not to complain, but the whole purpose of the media question was for easy transport of reasonable amounts of TASS raw fits images. As I mentioned much earlier, Internet is the logical wave of the future for this, but many sites (including ours) do not have high speed data links. This means that transfers of even one night's data from one site takes a significant fraction of a day, and hogs the bandwidth from other users.

Also, say that we actually do get all 7 sites up and going, and that half of them are taking data on any given night. This means 1GB data/night coming from TASS sites, and even Michael R. is not going to give up that kind of storage space on his computer for Internet downloads!

Tom has a Zip drive, and it makes sense to use that media as a first choice to copy data to users with slow links and without cameras (Nick could copy the Zips to CDROM for those users without Zip drives), at least for the SMSP-A and -B fields. After that, I still say that the raw frames can be archived at the originating site in any fashion that site wants, and then send the extracted data files to Michael R.'s computer.


Arne Henden aah@nofs.navy.mil
Mar 19, 1997

Michael G. wrote:

> I'm not sure I agree on the set aperture sizes you have proposed.  The
> "star" program uses an aperture based on the size of the psf for that
> image.  The width and height of the aperture are set to about 2*fwhm in x
> and y of the psf.  For Tom's images and mine this represents an aperture
> diameter of usually 5 to 7 pixels.  Anything less than this and I would
> worry about jitter in selecting the peak position to contribute
> significantly to the error in measuring the ADU sum within the aperture.
> 
> As for generating N magnitude measurements the obvious question becomes,
> which one should be used when comparing one image to another? 
My CCD photometry indicates that the optimum sized aperture is usually 5*fwhm. See Steve Howell's article on photometry in Astronomical CCD Observing and Reduction Techniques, ASP Conference Series #23, or his article on growth curves in CCDs in Astronomy, ASP Conference Series #8 for further information on selecting the optimum aperture size.

Saying that, I also realize that the TASS apertures are HUGE. For example, in the 3 SMSP-B fields that I imaged, there were typically 100 stars brighter than V=17 within a 100 square arcmin field, or about 1 per square arcmin on average. The fwhm of Tom's I-band cameras is about 3.5 pixels, or about a square arcmin. This means you have about a 50/50 chance of any aperture containing two stars instead of just one, and with that second star contributing 10 percent or more of the flux. And this is 20 degrees from the galactic plane and away from the galactic center to boot!

Therefore, my suggestion was a compromise. I did make a typo, as I meant to say 2,3,4,5 pixel radius instead of diameter, which then matches what Michael G. suggested: 2*fwhm as a minimum size. I suggested a range of aperture sizes because some camera/filter combinations are giving fwhm near 2 pixels, and some are giving fwhm of 4 pixels, and if the extraction programs report a range of apertures, then the off-line reduction can select the best aperture for the given site. The jitter in the peak position is smaller than the error due to neighboring stars, so I would err on the side of smaller apertures.

As I mentioned before, and as Michael Richmond has mentioned, a master list of objects should be created by merging those nights with photometric conditions. This will supply the list of 'known' magnitudes and colors. Then all variable star measurements are differential with respect to this master list, and even relatively poor nights can give good photometry. As long as the photometric nights are transformed onto the Johnson system using Landolt standards (or my small area regions), and use the same aperture sizes so that the same number of neighbors are included, then results from different sites should be combinable.


Arne Henden aah@nofs.navy.mil
Mar 19, 1997

(You can sure tell when I get into work in the morning!)

Chris A. wrote regarding psf fitting. I agree with most of his points, and you will find that many groups working with crowded field photometry (such as MACHO and SDSS) are using psf fitting for their photometry. This is most likely the best approach for TASS as well, since the big images mean even sparse regions will act 'crowded'.

However, I've done a lot of psf photometry, and there are several major problems. (1) you eventually need to place your photometry on some standard system, and that requires aperture photometry for best results since that is how Landolt acquired his photometry. This is very obvious in fields like Ru149 and Ru152, where he is including several fainter stars in the same aperture as his standards, so it is the ensemble that is the 'standard'. (2) Automatic psf photometry is complex. Chris mentioned several off-line and inspection steps to ensure that the subtraction was done properly, and for TASS these steps have to be automatic. I think that several man-months of effort are going to be necessary to get programs to do automated psf fitting and deblending, and so strongly suggest that the first approach is to do the simple task of aperture photometry and then as a second step work on the psf fitting. (3) I am not convinced that the TASS cameras are supplying their best images yet, and as the cameras converge on the optimum image size, the psf delivered by these cameras will change shape. Therefore, selecting an optimum technique for calculating the psf may change as well. I would rather see effort put into getting the cleanest images first, then step back and see whether something simple like a two-dimensional Gaussian will work adequately or whether you need an empirical correction such as that calculated by DAOPHOT.


Herb Johnson hjohnson@pluto.njcc.com
Mar 19, 1997

*>> As for generating N magnitude measurements the obvious question becomes,
*>> which one should be used when comparing one image to another? 
*>
*> Just an idea. I think you should compare each new object list with a
*> composite made from all previous lists.  In other words it is a "merge"
*> operation.  In other words a star's location is a wieghted average of
*> all the locations it has been.
I don't think I'd go for "average", but I think you've begged the question. My findings are that the *locations* are pretty consistent across different programs. It's the *magnitudes* that are not. Hard to imagine that any refinements of current programs will degrade the astrometrics of position. But magnitudes are still an open problem, and as I look at prior work in professional photometry it looks harder, not easier. And an average of few samples, with odd correlations, is not so good I think: I would hope new magnitude estimates are BETTER than old.

The photometric literature is full of lists of stars. The historic problem as it appears now seems to be: they did not use consistant standard stars, they did not use consistent, *band-limited* methods. The additional problem WE have that THEY did not is: we have inconsistant sky conditions. (We can bypass this by only using the "best" images, but our skies are not good even at their "best", and besides resolving this problem is more useful to the AMATEUR community.)


Herb Johnson hjohnson@pluto.njcc.com
Mar 19, 1997

Some comments:

Chris argued for removable IDE drives, and against tapes, CD-ROMs and ZIP drives. Tapes are too much trouble, I agree there. Otherwise I disagree with his assessments. To save time, let's just tote the costs:

One CD-ROM, under $10. CD Rom drives are "free". Writers require some scrambling to find, they will be cheaper soon enough.

Five ZIP disks, 500 MEgabytes, $75. Divide cost of Zip drive by number of uses, or borrow it, or ask the requestor to ship HIS/HER drive to you.

Cheapest IDE drive, $120-$150 new. Add docking station, removable drawers...

Tapes are slow, not standardized, and are incompatible sometimes anyway.

My ranking: CD-ROMs and ZIP drives, IDE removables a distant third with tapes.

I won't comment on Chris's distribution proposal as I disagree with his premises. I will say the Usenet news analogy does not hold: being an archivist or data producer is an intensive HUMAN activity and requires some particular equipment. Passing News packets around is an automatic procedure with equipment in hand.

There is a kind of "who bells the cat?" flavor to this discussion. Seems to me it comes down to this: who will help the Mark III camera owners burn some CD-ROM's? And: who will offer to provide custom data subsets from the CD-ROM archives? The same people need not do both if you make more than one CD-ROM set, which is only reasonable.

I would suspect the camera owners will find Zip drives convenient, but it doesn't matter HOW they *temporarily* archive their data. But its a signifigant effort and responsibility: if they aren't interested, the argument is moot.

Some one or some group can act as archivist in two ways: to ARCHIVE the data and to DISTRIBUTE the data. Archivists should have access to CD-ROM writers and ZIP drives (requestors can always SEND THEIR DRIVES), support UNIX and/or MS-DOS file systems, and/or provide an Internet FTP site: how they take it in from the camera owners is between them and the owners, as Chris suggests. How they trade in all this really is idiosyncratic (as you suggest), so long as they can produce to meet demand. I guess this is a "marketplace" model: the *public* question is, "are Zip disks and CD-ROMs and FTP acceptable media to recieve TASS data?" My answer is obvious.

Finally, all this is a "tempest in a teapot" if few people want all the data, and most people want a little data. Needs, interests, have not been established. There are still discussions about throwing out raw data!

Chris, since it's ultimately up to the camera owners, and we have too much maillist mail, I'd suggest taking this discussion *only to them*, and also to any who want to be archivists. Has anyone volunteered to "bell the cat"? Not me!


Herb Johnson hjohnson@pluto.njcc.com
Mar 19, 1997

*> Now I see there are basicaly two groups of people here on this list.Those
*> with cameras and an immeadiate need to reduce data and those with access
*> to only a sampling of data.
*>
*> I will propose to two phased approach to TASS data reduction.  1) What
*> has already happened put of necessity, a production mode.  2) A experimental
*> mode.  I expect people who have cameras and data to be interested in #1
*> while others would somehow form a coordinated effort at #2. Results of
*> #2 could be folded back into the production effort when it warents.
*>
*> Lately #1 seems to be taking shape.  I suspect there are others
*> interested in #2 but that there is zero organization of that effort.
Something is grating me about this description, on several grounds. I'll express my reactions and then cut to the chase. Sorry it takes a few paragraphs, you can skip to "BOTTOM LINE" if you are impatient. I'd sooner discuss work on data for the AAS papers, which I think should be our immediate goal. - Herb

Initially, I'm not sure if your remarks are speaking about the current efforts to produce papers for the AAS, or the TASS effort in general. There is not much time for "production" vs. "experimental" modes when the conference is this summer. My sense is that we are *currently* trying to refine our methods to produce good data, i.e. lotsa stars in position and magnitude. In any case, Tom has made it clear we should focus our efforts on these AAS papers and that seems reasonable to me as *time is short*.

If you want to develop a "production mode" process, it presumes a long-term effort toward specific goals and results. If these results are to be of reasonably professional quality, it seems to me that these upcoming AAS papers will force us to create tools and resources that are of this quality. I'm not so sure we are there yet. And, as you point out, we are still working on the software. All in all, I'd suggest "honing the tools" over "refining the process".

Not to mention: do we have consensus on a long-term plan? I don't want to start an argument on the long term, but I'll mention one fundamental point. Some folks have suggested "we don't need to retain raw data" as we will (eventually) produce good star lists. I strongly disagree and I'm not alone.

Your "division of labor" presumes those who have cameras will also reduce and supply the refined data. I don't know that either, unless all this is limited to TASS III camera owners, in which case all this argument is moot by the Golden Rule: he who has the gold makes the rules. The TASS Mark III "camera club" is of fixed membership: if you guys want to do specific things together, presumably you can do so. But the Mark IV owners may have other ideas....

As for your second group, I would not call my "experimental" efforts "zero organization". If you mean it is not directed by a consensus, or by a particular person; well, is yours? Is Gutswiller's? is Richmond's? Seems to me you each developed your programs from your given backgrounds, plus discussion among yourselves and others. To that extent, there was little "organization" until Tom announced the AAS initiative. I've likewise taken some direction from that initiative, and then from those who respond to my work with "this is a solved problem" or "this needs more work" or "thanks for your support of this". These kind of remarks are all focused by the a priori statement, "we need these things for the AAS papers".

If experimental efforts seem "unorganized", speaking for myself I say that I've had to *work* to get at the processes and assumptions used to create the various starlists to date, and work to develop insights into what affects the data. Few people, and fewer amateurs, will do this kind of work. Also, good documentation is also hard to produce: I commend Michael Richmond and others for maintaining Web pages and writing such Tech Notes as are written. I've organized my work to date in chunks that represent my "learning curve", hopefully at least leaving a trail that others can learn from too. Is TASS to be a "production" effort or an AMATEUR effort?

Finally, I'll humbly suggest the critical problem *at this time* is now good and confirmable MAGNITUDE estimates. The programmatic and observational side of this is trying to resolve the consequences of variations in sky conditions, and different methods between programs in use. The data-driven side of this is finding a reference star list of "sufficient" distribution of stars in our passbands. If we limit ourselves to near-perfect observational runs of small areas with "many" and "known good" stars we can sidestep this, but that won't work in general. ***Frankly, the converse excites me: can we refine techniques to do good work in bad skies?*** That's the AMATEUR side of TASS, and the great (but rarely cited) value of CCD cameras in amateur astronomy. (Like Maury, I get weary of beautiful pictures in the astronomy magazines: the good picture in crappy skies is much more exciting!)

BOTTOM LINE

The bottom line of my response is this: we have AAS papers to produce, primarily about generating good lists of good stars AND what it takes to do so. Our core problem is consistent magnitudes, from a good reference base of photometric stars; and either observing on clear nights or resolving variable sky conditions. Might I suggest: When all *that* is done, AND we get some feedback from the amateur and professional community, we can get better organized around a production version of that work, develop the appropriate resources and materials, and set some agreed-opon goals. Then our starlists will be of known quality and thus of value. Meanwhile, if you want some specific "experimental" analysis done, *ask* for it explicitly, and (continue to) provide documentation so it can be attempted: if other experimental work is also done which suggests other directions or improvements, take it and offer thanks. As for production work in the long term: if the Mark III owners have some consensus on this they can of course do "production"; but they in fairness should create opportunities for other folks to look at raw data to "produce" their own results, either as new objects beyond the production list, or to support a process of self-education and learning (could be both of course!).

A closing self-comment: If you think my remarks are annoying, wait until you are in front of an auditorium of amateurs and professionals and you are telling THEM how you got your results! Better the devil you know (me) than the devil you don't know (them). ;)


Chris Albertson chrisa@wavenet.com
Mar 19, 1997

Looks like I've started an argument, I was trying to be neutral and diplomatic I was simply predicting that a grand plan for exchanging data would likely not work and that data exchanges will have to be worked out as bilateral agreements between sets of two people at a time. I was also saying that with enough of these agreements everyone can have all the data they want without over working any one person. I also do not favor one media type over an other. Each has it's uses Herbert R Johnson wrote:

> Chris argued for removable IDE drives, and against tapes, CD-ROMs and
> ZIP drives. Tapes are too much trouble, I agree there. Otherwise I
> disagree with his assessments. To save time, let's just tote the costs:
No, no, no, I was not arguing for IDE drives as any kind of standard I was just pointing out an option that was overlooked. It may work in some cases. For example if I happen to have an old 850MB drive laying around I could buy two drawers for it (at $30.00 each) and have a quick and cheap 850MB removable device.
> One CD-ROM, under $10. CD Rom drives are "free". Writers require some
> scrambling to find, they will be cheaper soon enough.
Yes I expect they will drop to the $300 level and stay about there. Cheap enough. _But_ have you ever tried to make a CD ROM? This is not something you want to do 5 or 6 times in one day. The practical aspects of CD burning are such that you just will not use CD-R in a production mode.

It would be a very good idea to save up a few months of "A" and "B" field data then burn a half dozen copies to CD-R. but this would be a once a month issue. _I_ will even offer to make a few CD-ROMS now and then but I don't see it as a production mode media.

> Five ZIP disks, 500 MEgabytes, $75. 
> Divide cost of Zip drive by number of uses,
> or borrow it, or ask the requestor to ship HIS/HER drive to you.
Tom just e-mailed me. Says he has already had enough of using ZIP to transport data even between two computers in the same house. Says it takes stacks of disks and the process is dead slow. Looks like his use of ZIP is pushing the media past it's usfull limit (Tom, sorry for the bad paraphrase) I think ZIP has it's place but not for storing multi-gig data sets.
> Cheapest IDE drive, $120-$150 new. Add docking station, removable drawers...
Removable setup = $30.00 for each computer. Must buy drawers all at once from the same vendor as they are not standarized. Add up the numbers and see if it works for you, may or may not. Very cheep if you have a "free" drive.
> Tapes are slow, not standardized, and are incompatible sometimes anyway.
No, **cheap tape drives** are slow, not standardized, and are incompatible. If you have $700+ to spend on a tape frive these problems go away. That is a big "if" and I assumed no one would buy one of these drives. But several people already have them. New DAT drives are as fast as SCSI disks and _are_ standardized. exchange is not a problem. Holds 8GB on a small cheap cartdridge but at $1K per drive. Also you can random access to any file in one minute. This is realy the way to go but the cost is a killer.
> My ranking: CD-ROMs and ZIP drives, IDE removables a distant third with tapes.

I think it all depends on volume and frequecy of transfer and what you already own. For example if two people found that they both owned 4mm tape then they would certainly pick that, as it can write multi-gigabytes to a $9.00 reusable cartridge and is as fast as a SCSI disk.

> Some one or some group can act as archivist in two ways: to ARCHIVE the data
> and to DISTRIBUTE the data. Archivists should have access to CD-ROM writers
> and ZIP drives (requestors can always SEND THEIR DRIVES), support UNIX and/or
> MS-DOS file systems, and/or provide an Internet FTP site: how they take it
> in from the camera owners is between them and the owners, as Chris suggests.
I could do the archive, but I don't see how I (or anyone else) could handle data distribution themselves. I'd spend half of every day burning CDs, copying tapes and running to the post office. Nice as it sounds I just don't think you will find one volunteer willing to be the central distribution point.

I proposed a distributed system where nobody is required to send out more data then they collect. I think I've beatten this subject to death. So I promise not to post any more of this stuff to the list.


Chris Albertson chrisa@wavenet.com
Mar 19, 1997

>   Chris A. wrote regarding psf fitting.  I agree with most of his points,
> and you will find that many groups working with crowded field photometry
> (such as MACHO and SDSS) are using psf fitting for their photometry.  This
> is most likely the best approach for TASS as well, since the big images
> mean even sparse regions will act 'crowded'.
I think we agree: 1) PSF should be a long term goal but 2) It is hard to do so 3) we stay with aperture photometry for now.

My one idea that I think may have been missed is that there are enough people that we can do it all: Reduce data now _and_ work on next generation data reduction techniques.

That was the summary. Details follow: (stop reading here if not interested)

>   However, I've done a lot of psf photometry, and there are several major
> problems.  (1) you eventually need to place your photometry on some
> standard system, and that requires aperture photometry for best results
> since that is how Landolt acquired his photometry.  This is very obvious
> in fields like Ru149 and Ru152, where he is including several fainter
> stars in the same aperture as his standards, so it is the ensemble that
> is the 'standard'.  (2) Automatic psf photometry is complex.  Chris mentioned
> several off-line and inspection steps to ensure that the subtraction was
> done properly, and for TASS these steps have to be automatic. 
What I envision is a fully automated process. But when you run any automatic process you want a means of quanitative quality control. You then periodicaly improve the automation and use your quality controls to see if the improvement was in fact an improvement.
> I think that
> several man-months of effort are going to be necessary to get programs to
> do automated psf fitting and deblending, and so strongly suggest that the
> first approach is to do the simple task of aperture photometry and then
> as a second step work on the psf fitting. 
Yes. This is my point too. I think there are enought pepole here to do both at once. I do not want to suggest giving up the current effort. Some can reduce data using what works now and others can research new methods. I think that every fractional percent in improvement in data reduction technique will translate into the discovery of hunders of otherwise un-noticed variables.

A few man months is a reasonable estimate. It is a small sized well defined problem that is solvable by a few people working together via the Intrnet. If the task were larger, getting into 1+ man year then I say we hadn't a chance of success but this looks doable.

> (3) I am not convinced that
> the TASS cameras are supplying their best images yet, and as the cameras
> converge on the optimum image size, the psf delivered by these cameras
> will change shape.  Therefore, selecting an optimum technique for calculating
> the psf may change as well.  
I think one thing that needs to be developed is a way to get the PSF from the data automatically. Perhaps the software first searches for a few isolated stars....?
> I would rather see effort put into getting the
> cleanest images first, then step back and see whether something simple like
> a two-dimensional Gaussian will work adequately or whether you need an
> empirical correction such as that calculated by DAOPHOT.
I think you can have both. Image improvement can only be done by a few people who have access to the equipment. There are others with nothing to do.

From what little I've seen I think you need a sophisticated model. The PSFs are not figures of revolution. Streaked stars may only be Gaussian in on dimension and something else in the other direction. Likely worse, the Gaussian and "other" functions may be mixed as the streeaks appear to be diagonals. I think this is a good area for study.

in a following message

This was sent to me and CC'd to the list but did not appear on the list. I am reposting it. It is from BARTHOLDI Paul

> For those not read up on PSF fitting here is a overview.  I have left
> out
> some detail:
> 
>   0) off line, build a good model of the system PSF.  This could be
> based
>      on closed form equations or tabular data or both.
>   1) Extract a list of of stars by any method one likes.  The list will
>      contain lines of (x, y)
>   2) For every (x,y) extracted in #1 
>      2.1) attempt a least squares fit of the PSF model to
>           the data at point x,y.  
>      2.2) Record the peak of fitted function as (x', y').  This is the 
>           instrumental Mag. and fine-fitted possition. 
The "volume" is probably a much better estimate of the Mag, in particular if the shape of the PSF is changing with the position on the field.
>      2.2) Subtract the PSF model from the data (removes the star from
>           the image.)
From my experience, and that of most if not everyone doing PSF fitting it is better to substract only a fraction of the fitted PSF ((0.1 to 0.9 depending on S/N) for every star in the list and then iterate on the residuals. This is the so called "CLEAN" method.

There is also an interesting method of Lucy to locate quite accurately, even in overcrowded fields, the nearly exact position of the stars.

If you are further interested, I can send you the references to CLEAN and Lucy's method.

> So if I had to reduce a big pile of data today.  I'd use aperture
> photometry
> as it is quick and simple but in the long run I suspect the PSF method
> will
> give better result.
You could use aperture photometry as a bootstrap for PSF fitting, as for PSF you need initial values, and the better they are the faster the convergeance.


George Turner turner@bigbang.astro.indiana.edu
Mar 20, 1997

This data interchange problem (I believe) is only a short time issue; once TASS "goes into production" the only images you'll be keeping are probably random images from a night to check quality control or perhaps the software will evolve to a point where the data reduction software can trigger the camera computers to save a particular image because it found something interesting; (when it gets this far let me know & we'll tie it into RoboScope's Trigger mechanism) but we're a ways from that at this point. You're going to be producing so much data that you won't be able to hold it all. Probably what will happen is that the camera operators will archive as much data as their disks can hold and oldest data gets deleted to make room for new data.

Right now the problem is getting data from the camera operators to the software folks and getting their results feedback to the camera folks. There is some need to exchange lots of images because we're studing the stability of the camera systems and an evolution of the software to deal with the various aspects we're seeing in the images. But, I think this is best handled by network servers for this volume of traffic. (in a couple of weeks I can probably help out here) At some point we'll want to test the software against large amounts of data files; one on one solutions worked out then will be the best.

I think if everybody is expecting to have a duplicate copy of all the data then they've got a huge, unweildy, problem on their hands and no pratical way of doing anything with the data once they've got it; just keeping one copy of all the data would be no less challenging. (some infinities are bigger than others) The TASS system will have to be an automated system and this data will have to be dealt with much as time is; it's a fleating thing and you've got to make the most of while you have it. Snapshots are allowed; keeping all of reality is not.


Herb Johnson hjohnson@pluto.njcc.com
Mar 20, 1997

The full update to my report has been added to the previous report on the TASS Home page edited by Michael Richmond. It will have more "raw data". TASS maillist readers who are not interested in numbers and analysis can probably ignore this message. - Herb

Chris suggested that my finding of "drift" in Tom Droege's dark pixels (columns 790 to 794, counting from 0) was likely a temperature drift, which he had previously noted. He suggested I look at other people's images. This was a surprising tough assignment: only Tom is posting raw data for the most part. I found an old image of Glenn Gombert's, the image with an "asteroid" streak in the far end of some rows; and an image of Michael Gutzwiller's. I could not find any of Michael Richmond's raw images. Perhaps camera owners could be encourage to keep at least a FEW raw images "around" for the curious.

Here's my findings:
------------------

Files from Gombert and Gutzwiller did not show the "drift" that Tom's camera image did. The shifts in local average pixel value (run of 50) vs the overall average (run of 890 or so) were smaller and varied inconsistently, well below the standard deviation of the overall average. I also found that Glenn's image, originally posted as "an asteroid trail", actually extended into the dark pixels and I had to "edit" an affected row out (it's the only Gombert image I could find) to remove its impact. This effect should be noted.

I think my method of monitoring the dark pixel columns - comparing runs of 50 pixels to the entire run, normalized to the standard deviation of the whole run - is a good measure of "stability". And it apparently also offers some degree of automated "outlier" monitoring. If there is some consensus on this, I'd encourage it be incorporated into our data analysis, and maybe reported in the FITS header in some fashion.

Here's some summary data:
------------------------

Droege file "d0500244.fts":

row            790          791          792          793
mean         6737.5       6647.88      6622.14     6627.31
median       6656.0       6665.0       6640.0      6652.0
st. dev      21.67         21.47       20.72       20.84
1st diff    -35.82        -35.28       -33.82     -35.69
last diff    +16.34        +15.00       +14.66     +15.91
avr diff     +2.2          +2.38        +2.4       +2.4
The "avr diff" is the average of all the differences. The "diff" values reported are actually the largest of the first few diffs, and the largest of the last few diffs. "Few" is four or less. Simple arithmetic yields the equivalent diffs in standard deviations.

Gutswiller file "32t0433.814" and Gombert's file "30T0478.642":

Reporting columns 790, and reporting the most negative and positive differences in absence of a clear pattern:

site         gombert        gutzwiller
mean            8094.85      9567.12
median          8092         9569
st. dev         11.90        26.73
max neg diff    -2.79        -4.08
max pos diff     3.99        3.84
avr diff         .01         -0.04
These differences are only a few ADU's, well below the standard deviation.

**Raw Data: Glenn Gombert's camera, with "spike"

For Gombert's file, there was a bright streak which apparently extended into the so-called "dark pixels". If these are truely dark pixels, this must be an instrumental effect, maybe the previously mentioned "saturation crossing over from one camera to another" problem (see the TASS Tech Note on Mark III hardware problems which I hope references this).

The edited row, from cols 790 to 799, and its successor row are:
 10216  9464  8952  8626 37007 40301 33066 54255 11198     0
  8096  8064  8078  8046 37011 40305 33064 54255 11199     0
Note that columns 790 through 793 of the first row are all impacted, and that the values in the second row are more consistent with the mean value and each other.


Herb Johnson hjohnson@pluto.njcc.com
Mar 20, 1997

*> From my experience, and that of most if not everyone doing PSF fitting
*> it is better to substract only a fraction of the fitted PSF ((0.1 to 0.9
*> depending on S/N) for every star in the list and then iterate on the
*> residuals. This is the so called "CLEAN" method.
As a matter of fact, my colleague Dr. Robert Dixon at Ohio State (and his advisor at the time, John Kraus) used this method before it was "discovered". They were reducing radio astronomy data from their drift scan Ohio Survey instrument. In effect, they had a "one-pixel" array. Richmond has posted my references to their work on the TASSS web page, and I have the original papers on hand, which describe the method in detail. I'm ready now to look at this and post something on it. (Dr. Dixon has not answered my email request to post the original articles.)


Herb Johnson hjohnson@pluto.njcc.com
Mar 20, 1997

*> Looks like I've started an argument, I was trying to be neutral and
*> diplomatic...
I would like to give Chris credit for trying to work something out about archiving and distributing data. It's a general problem for all kinds of computing and research work. I'm surprised that CD-ROM's are *still* difficult to make: I'll say no more as Chris reminds me I have not made one. The good point of his discussions is that we've all had a glimpse into the merits and problems of various media. The consensus seems to be: private arrangements at the camera sites to archive data; no consensus as to how to distribute the data; and hopes that "tomorrow" the problem will be solved by (choose a technology). I agree with Chris there's been enough on this subject. As for distribution:
*> I could do the archive, but I don't see how I (or anyone else) could
*> handle data distribution themselves.
If there is a demand, and people will pay, someone will do this. Big if's. I hope we don't discard raw data in the long term, but it's hard to justify the effort to archive it.


Glenn Gombert ggombert@infinet.com
Mar 20, 1997

I have been experimenting the last several evenings using Michael's star matching code that I complied to run under Windos95 last fall and the two (SOA, MGSC) catalogs that were generated by Mike G. to use with his "Stars.exe" program. These two catalogs cover the entire area that TASS Mark III camera's cover and this should provide more convinent access to using star-matching as a way of providing astromertric calibration of TASS Images. I was planning on using this as a method of "cross-checking" the direct calcualtions of RA and DEC that are made by the SExtractor program from the WCS parameters that are stored in the FITS heaser(s).

The reason that I was interested in trying this apporach with the combined catalog files created by Mike G. is that it gets around the problem of having to exract a "box-of-stars" for each image that you desired to calcaulte RA and DEC for.

The TASS files that I have been using were several "B" fieldds that have been collecting here in Dayton for the past several months. They should yield similar results on most TASS images.

Prelimiary results show that about 20% of the stars that are detected in the images are actually "matched" to entries in the SAO catalog. This holds true for the 3-4 images that I tried this on last night.

When I tried the same files with the HGSC catalog all I got were "heap overflow" errors from the star-mathing program, this was probably due to the large amount of objects in this catalog as compared with the SAO catalog. I probably need to change the C++ compiler setting(s) for the Heap size and then try this again on the same files.

I will post some results to Strom if I can get a good set of results with the complete HGSC file.


Chris Albertson chrisa@wavenet.com
Mar 20, 1997

It appears the matching program you are using is nearly identical to one of the matching tasks in IRAF. In fact, in the "stars" help file Michael says it is. If this is the case then the program does the match using a similar triangles method and it is not such a good idea to have one field be much, much larger then the other. The reason is the software picks the best (brightest) 20 (or whatever the parameter is set to) stars in each list to do the match with. If they do not cover close to the same areas the two sets of 20 stars will have few in common and the fit will not be so good.

It is not hard to get WCS data from the FITS header, do a bit of math on in and then use it as a selection criteria against a catalog. Then create a catalog subset on demand for each image (your box of stars.) You do not need the whole GSC as the matching program will pick and use only the 20 (or so) brightest stars for the match.

I still feel sory for you guys writing in C. I can do all of the above in six or ten lines of CL. ;-)


Glenn Gombert ggombert@infinet.com
Mar 21, 1997

You have made a couple of good points, I find that I can only "match" about 20% of the stars in any one given image using the SAO file of Mike G. And Michaels star matching program. If you have to extract a separate file (a "box" of stars) for each file to process than it gets very cumbserom to process a lot of images this way.

It apperas that Michael has gotten around this problem by pre-computing some of the matching information for each area of declination by extracting each region and processing it ahead of time.This probably works well for the region around 0 degreess DEC but what happens when the Mark IV and beyond when you must "match" stars over the entire sky from many different composite images? That's probably a "whole nother ball game"......

I guess this is why (conceptually) that I like the idea of calcualting the RA and DEC directly for each object in an image from the WCS parameters in the FITS image header.


Michael Gutzwiller deepsky@fuse.net
Mar 24, 1997

>       You have made a couple of good points, I find that I can only 
> "match" about 20% of the stars in any one given image using the SAO file 
> of Mike G. And Michaels star matching program. If you have to extract a 
> separate file (a "box" of stars) for each file to process than it gets 
> very cumbserom to process a lot of images this way.
Actually this is how the star program works. While the whole catalog is read in at the start, Michael Richmond's star matching routines are fed subsets of the catalog of the same size as the image for each attempt at a match.
>       It apperas that Michael has gotten around this problem by 
> pre-computing some of the matching information for each area of declination 
> by extracting each region and processing it ahead of time.This probably 
> works well for the region around 0 degreess DEC but what happens when the 
> Mark IV and beyond when you must "match" stars over the entire sky from 
> many different composite images? That's probably a "whole nother ball 
> game"......
Michael's star matching routines assume that the coordinates are rectilinear, not a bad assumption for RA and Dec near the equator. Away from the equator though the catalog subset would need to be projected onto a plane and then matched.


Tom Droege droege@wwa.com
Mar 24, 1997

I have been working through my lists to reprocess them with the latest version of star. Is this appropriate at this time, Mike? The idea is to get a big enough number of measurements for each star to start looking for both a "fixed" set of candidates and a "variable" set of candidates.

Mike, do you plan to update your star lists on the web page?


Michael Gutzwiller deepsky@fuse.net
Mar 24, 1997

> I have been working through my lists to reprocess them with the latest
> version of star.  Is this appropriate at this time, Mike?  The idea is
> to get a big enough number of measurements for each star to start looking
> for both a "fixed" set of candidates and a "variable" set of candidates.
I think so. If we're going to have enough data to analyze for June we need to stop cutting bait and do some fishing. Enhancements to star and other programs will continue but we need to have some data to analyze.
> Mike, do you plan to update your star lists on the web page?
I will update those lists and add some more. I have several other nights with A and B fields to add, some going back as far as October last year.


Glenn Gombert ggombert@infinet.com
Mar 24, 1997

I have just been making final arrangements with Dr. Bob Havlen to give a talk on the TASS Project at Universe-97 that is to be held June 28,29 in Chicago at the Hyatt Regency. Bob is the Executive Director of the Astronomical Society of the Pacific (ASP) and coordinator of Universe-97.

I will probably try and use some of the reduced data from the "SMCP" to present as prelimary results for the talk. Universe-97 is the annual professional meeting of the ASP and usually also includes several "high level" talks by amatuers.

The TASS project is going to get some "good press" in the next several months with Michael's talk at the June AAS meeting, the talk at Universe-97 and a major article underway on the TASS Project for Sky and Telescope.

The "URL" for Universe-97 is http://www.aspsky.org/html/annual.html if you would like to check out further details. I attended two years ago at Universe-95 and found it to be well worthwhile. Many of the talks are give by profesional astronomers during the two days of "Universe" that carry it far beyond the usual "star party".


Arne Henden aah@nofs.navy.mil
Mar 25, 1997

Mike G. wrote, in response to Tom:

> ....  If we're going to have enough data to analyze for June we
> need to stop cutting bait and do some fishing.  Enhancements to star and
> other programs will continue but we need to have some data to analyze.

There seem to be enough scans to start this kind of processing. I strongly suggest that you start with field A, as I know which stars are variable there. It is always best to check against known results and make sure you are doing things right before proceeding to the unknown. You will need some time to learn how to match starlists, adjust the photometry and detect variables. This is the fun part!


Arne Henden aah@nofs.navy.mil
Mar 25, 1997

Given below is the abstract that I sent in for the June Pulsation Conference meeting in Los Alamos. The TASS results will be put on one panel of the poster paper, so TASS should get reasonable publicity.

These pulsation conferences occur every two years, and comprise most of the active researchers in intrinsic variable stars. The group is normally a few hundred professional astronomers, though this edition has promise of being the largest ever since it is in honor of Art Cox, one of the premier theoreticians.


Title: Variable Stars in the SDSS Calibration Fields

Authors: Arne A. Henden (USRA/USNO), Ronald C. Stone (USNO)

                                      ABSTRACT

The USNO Flagstaff Astrometric Scanning Transit Telescope (FASTT) has
been used to scan 384 total square degrees in 16 regions along the
Celestial Equator.  The primary emphasis has been in providing accurate
astrometry of all stars from R=9-17 within these regions.  As a
byproduct of the project, 2000 new variable stars have been discovered.
With very conservative cuts, approximately 0.4 percent of the scanned
stars are shown to be variable.  Since this is a magnitude-selected
sample that spans a wide range of galactic longitude and latitude, it
is useful in statistical studies of stellar formation.

This paper will present the preliminary results of photometric studies of a subset of these variables. Future plans for investigating these stars include additional photometry with the USNO 1.0m and 1.3m telescopes and the cooperation of The Amateur Sky Survey (TASS) and Astronomical Studies Group (GEA).


Arne Henden aah@nofs.navy.mil
Mar 26, 1997

Tom has sent me two Zip disks full of scans through the SMSP-A and -B fields. I've copied these disks over to my workstation and will be busy for the next week or so looking at the images.

Who wants these disks next? I can send them on at any time, now that I am finished with them.


Nick Beser beser@aplcomm.jhuapl.edu
Mar 26, 1997

> Tom has sent me two Zip disks full of scans through the SMSP-A and -B
> fields.  I've copied these disks over to my workstation and will be
> busy for the next week or so looking at the images.
>   Who wants these disks next?  I can send them on at any time, now that
> I am finished with them.
Send them to me, and I can setup a limited number of CD-ROM's. I will be back in the lab on Monday and I can burn the CD-ROM during the week.


Tom Droege droege@wwa.com
Mar 27, 1997

It is time to write an abstract for the June 10 AAS meeting.

I propose that I be the presenting author and that Michael Richmond be the introductory member. I suppose I could try to join before the meeting, but I am already in IAPPP and ASP. I am not sure that I like the style of the super large scientific meeting. I will probably not go to such a meeting again.

I would propose the following author list:

Professionals:

Michael Richmond
Bohdan Paczynski
Arne Hendon
Paul Rybski

Amateurs:

Mike Gutzwiller
Glenn Gombert
Chris Albertson
Norman Molhant
Herb Johnson
Nick Beser
Ron Wickersham
Jure Skvarc

Please, anyone who thinks they should be on the list or that others should be on the list, let me know. It is intended to be an inclusive rather than an exclusive list. Apologies if I have left anyone obvious off the list.

Looking at the form, I propose to show the professionals with their institutions, and the the amateurs with the institution "Amateur".

In listing the authors, I have tried to show everyone who did actual work. So I have tried to include everyone who has taken data, written extensive comments on data analysis, written tech notes, provided support, etc..

I will, of course, contact each on the list and ask them for their permission. I will also put the abstract up for comment before submitting it.

I keep looking at data, it is going to be hard to normalize it well enough to find anything. At least for me working with analysis tools written on the spot in BASIC. I hope the rest of you have better tools so we can have a couple of graphs to show in June.

I will write an abstract in a couple of days. Everything will be put up for editing and discussion.


Tom Droege droege@wwa.com
Mar 27, 1997

It looks like both Michael Richmond and my self will submit papers. We have agreed that I will emphasize the history and orginazation of tass and the Michael will present data.

Here is a try for an abstract. Note that we are limited to 300 words. I think the below contains 299. So if you want to change something, you cannot increase the number of words.

Please check spelling and construction of the text. Also make sure you all send me your correct titles and affiliations. I like just being an "Amateur" as to affiliation, but can understand why some of you might want to have your organization listed.

TASS - The Amateur Sky Survey

As a non-astronomer watching Shoemaker/Levy crash into Jupiter through the postings on sci.astro, it occurred to me that it might be fun to build a comet finding machine. After wild speculations about how such a device might be built - I considered a 26" x 40" fresnel lens and a string of pin diodes - posts to sci.astro brought me down to earth. I quickly made contact with both professionals and amateurs and found that there was interesting sceince to be done by an all sky survey. After several prototype drift scan cameras were built from FAX linear array CCDs, it was determined that the real problem was software.

How does one get the software written for an all sky survey? Willie Sutton could tell you. Go where the programmers are. Our strategy has been to build a bunch of drift scan cameras and just give them away (without software) to programmers found on the internet. This author reports more success by this technique than when he had a business and hired and paid programmers at a cost of a million or so a year.

To date, 22 drift scan cameras have been constructed. Most of these are operated as triplets spaced 15 degrees in Right Ascension and with I, V, I filters. The cameras use 135mm fl, f.2.8 camera lenses for a plate scale of 14 arc seconds per pixel. The cameras reach magnitude 13. There are 768 pixels in the drift scan direction. Four of the triplets and a single have taken data. Operating procedure is usually to turn on and run through the night. A triplet will then collect 200 Mb of data.

Production has started on 25 second generation cameras which use 2k x 2k devices and a barn door mount.


Glenn Gombert ggombert@infinet.com
Mar 27, 1997

I have been using the MathWorks "MatLab" program on my DELL Notebook for the last few months and one of the things that it does very nicely is make "scatter plots" (which are just X-Y plots) showing 2-D data with the "third" dimension color-coded as required.

It seems with some simple scatter plots of different data sets it ought to be quite easy to see how well data from different sites/technqiues correlate. The RA and DEC positions could be plotted from the same fields (not too crouded), and then the intensity plots could be color-coded (to show magnitudes) to produce some nice plots. It seens like this would be an easy way to normalize data from different sites / measurement techniques without too much trouble.


Chris Albertson chrisa@wavenet.com
Mar 29, 1997

There was some interest here about a Linux real-time TASS camera driver. Norman was (is?) working on one.

I think I have found something that will make development of a Linux driver simple. There is a real-time extension to Linux for "hard real-time" scheduleing. The way it works is this:

It is a simple real-time scheduler that will always give the CPU to the task that is (1) amoung those ready to run and (2) has the highest asigned priority. Now the key - Linux itself runs as a task but with the lowest priority.

The camera task would run as a peer to Linux and would not be required to follow conventions and rules for normal Kernel based drivers. Also the interaction and interface between the camera driver and Linux kernel would be minimal or non existant.

The camera task would only be required to read data from the interface when an interupt occures and write it to a FIFO buffer with a provided function. A second task running as a normal user program sees the other end of the FIFO buffer as a file that could be opened, read and copied to disk.

Real time Linux appears to be supported and under active development There is a mailing list and a small user community.

For more information see http://luz.cs.nmt.edu/~rtlinux/

The paper "a draft of a paper on RT-Linux" is a good semi-technical overview.

I intend to download the software and install it on my machine. I should know very quickly how hard it will be to use.

in a subsequent message

Good news. I now have a real-time Linux system up and running. I am using it to write this. So it is not vaporware. Look at the e-mail header timestamps. It took only a few hours and that included time to eat dinner and play with my 5 year old. In that time I downloaded the code, appied the patch to the RedHat 4.1 kernel (vers 2.0.27) recompiled the kernel installed the software and got the test cases to compile and run. No I'm not fast it is just that easy to use.

A real-time task is currently running. It is catching about 8200 interrupts per second. While this is going on, I am writing this using netscape. I have a live PPP link to my ISP and am doing an un-related compile and have an Ethernet link to a Windows95 machine down stairs. The X11 Windows system is up and to top it off I've loaded IRAF, xephrem and a word processor. All this and the real time task seem to get along.

I've looked at the code. A minimal real-time task is quite simple, about 100 lines. My estimate for the amount of work required to write a TASS real-time camera driver given that we now have real-time linux is about this: Assuming I could "pirate" code from tm3get11 about 4 or five evenins work to get a minimum functioning linux TASS camera driver.

If anyone wants to work with me on a Linux TASS driver let me know. I can do any of the code except that which accesses the TASS hardware.

I have to give credit to the guys at New Mexico Tech. who developed RTlinux. For those interested I've quoted an overview from the documentation below

For more information see http://luz.cs.nmt.edu/~rtlinux/

This variant of Linux allows to handle time-critical tasks. This is accomplished mainly by insertion of Real-Time Kernel layer between Linux kernel and hardware interrupts. This eliminates the main source of Linux unsuitability for time-critical processing - big interrupt latency.

Under RT-kernel Linux kernel is just another real-time task. It has the lowest priority, and can be preempted when needed.

This structure imposes some restrictions on RT-tasks. They can not easily use Linux drivers, networking, etc. They can, however, transfer data to/from ordinary Linux processes through memory buffers.

Simple FIFOs are implemented for transferring data between real-time processes and Linux processes.

A typical application consists of real-time tasks that deal with hardware directly, e.g. acquire data from a device, and Linux tasks performing non-real-time processing, such as recording data on disk, sending it over network, etc.

The shortest feasible period for a real-time task under the RT-Linux scheduler is currently about 150 us on Pentium 120. Interrupt driven tasks can have smaller periods.


Tom Droege droege@wwa.com
Mar 29, 1997

I just located the last wire on the last pc board layout for the Mark IV. Well, it is not the last wire since now I have to check everything and fix my errors. ;^) It is also not the last printed circuit board layout. There are a couple of little boards needed for the controls and the stuff in the camera.

There are four major boards. The one I just finished has the CCD scanners and the front end analog stuff. One that went out last week contains all the controls. The camera PC board is back and looks OK. The memory board is nearly done at the vendor and should be in next week.

Soon, we will be checking out the electronics. My time line is for first light in June. This gives me 3 months of check out to make it. Seems reasonable to me.


Chris Albertson chrisa@wavenet.com
Mar 29, 1997

Norman wrote:

> When I investigated RTlinux about a year ago, the timer interrupt could not
> be used to drive a real-time task, as it could not be programmed to give fast
> timer ticks without completely disturbing the system.
> If that has been changed, i.e.: if it is now possible to get time-ticks as
> short as 100 microseconds or so, then RTlinux is worth investigating.
I think it will work now. RT Linux is under active development and the last release was in 1997, It comes with some example real-time programs that show that it does almost the 100usec you want. It may do better but these work on my machine.
  1. Example one: Program the interrrupt timmer to generate ~8200 interrupts/second. Each interrupt gives control to the real time task. The task polls a FIFO buffer that is being feed from a user level program running under linux. If there is a byte the real time task sends it to the PC speaker port. (the sound is awful) This demonstraits polling and data transfer from linux at 8200 polling cycles per second.
  2. Example two: Two real time tasks run. They shedule them selves to run periodically every 150usec. but out of phase by 180deg (75usec) Task one sets a bit up while task two sets a bit down. The result is a 150usec square wave that you can see on parallel port bit d0.

Here is a quote from the documentation:

The shortest feasible period for a real-time task under the RT-Linux scheduler is currently about 150 us on Pentium 120. Interrupt driven tasks can have smaller periods.

Periodic tasks are those who make a system call to RT Linux asking to executed at some specified periodic rate. Example #2 is one of these. Interrupt driven tasks are tied to a hardware interrupt like the programable timmer. #1 does this. Both example programs are very small, about two screen fulls of code. When compiled RT tasks are Linux kernel loadable modules. So tasked are started with the "insmod" command from the keyboard and stopped with the "rmmod" command. One you "rmmod" all your tasks RT Linux becomes just Linux again, no reboot is required.

I've been running a 125usec polling loop on this system for about 16 hours now. With this running I can still transfer data over the modem at full rate and access the disks and so on. Stability looks good.

The trick to making this work is to do the *Absolute* minimum work in the real-time task. Just poll the hardware and if ready read the data an place it in a memory buffer, then quite. All other fuctions would be done as normal user programs. The first one to write would be one to read data out of the memory buffer and copy it to disk.


Tom Droege droege@wwa.com
Mar 29, 1997

The Mark IV is being designed so it is not at all time sensitive for the interrupt. It will interrupt once a frame - probably about once every 100 seconds. There is not much that has to be fielded except to dump out the buffer before telling it to end the exposure and read out again. This should leave lots of time for the display task and to format the data for storage.


Tom Droege droege@wwa.com
Mar 30, 1997

It is nice in theory to think that things only have to be written once. Certainly I support writing clear cut processes in a modular form. The real problem is that we probably don't understand what the real problems are. So we will tend to have to re-write things anyway. That is really the purpose of the Mark IIIs. To teach us what the problems are when doing a survey project.


Michael Richmond richmond@astro.princeton.edu
Mar 30, 1997

I have written a new Technical Note, number 28, comparing the star lists generated by Gutzwiller, Gombert and Richmond for the same images taken by Tom Droege. You can find the report at

There are tables and plots. The bottom line is, in my opinion, that all three methods USUALLY yield very comparable magnitudes when run on the same images. I'm slightly less happy with the astrometry, but I think it will do for some purposes.

I request that both Glenn and Mike G. send me clarifications or fixes to the mistakes I've put into the reports, in the description of their techniques.


Tom Droege droege@wwa.com
Mar 30, 1997

I have now run Mike Gutzwiller's latest version of Star on all my available "A" data. Michael Richmond please add these files to the home page.

The program was run at the default settings.

The camera pointing is not ideal. But I don't plan to change it unless you all really beat me up. One can keep fussing and never take any real data. There were continuous efforts to get a better focus and to find the best VCO setting during the various data runs. There were also changes in the camera pointing angle about the middle of these runs. It will be obvious from the data.

The following "A" runs are (or soon will be) up on storm.fnal.gov. I have listed the run file, the background noise level, the PSF size, and the number of stars found in each run. The OK means that the match to the default catalog was reached. One set of files that looks to be good did not succeed in matching the catalog. This set of files has also been put up on storm so Mike G. can try to figure out what is wrong. This set is:

30t0524.817 30t0524.827 (I won't get these up until Tuesday afternoon as it takes too long to send them from home)

Run           Background  PSF Size    Stars Found     Match Found
S30t0483.931  43.8        7x7         480             OK
S30t0483.940  42.7        7x7         396             OK
S30t0493.900  49.9        7x7         382             OK
S30t0493.909  44.8        7x7         408             OK
S30t0511.851  59.4        9x7         195             OK
S30t0511.861  59.5        11x7        186             OK
S30t0515.838  59.4        11x7        226             OK
S30t0515.848  57.7        11x7        185             OK
S30t0517.831  77.1        9x7         90              OK
S30t0517.840  71.3        9x7         100             OK
S30t0518.831  41.6        9x9         368             OK
S30t0518.840  39.2        9x9         346             OK
S30t0519.829  40.1        7x9         438             OK
S30t0519.839  38.1        7x9         379             OK

S31t0483.884  41.5        5x5         544             OK
S31t0483.894  41.4        5x7         443             OK
S31t0493.863  51          5x7         441             OK
S31t0493.872  48.8        5x7         453             OK
S31t0515.802  47          5x9         334             OK
S31t0515.811  40.9        5x9         432             OK
S31t0518.795  39.6        5x5         426             OK
S31t0518.804  39.8        5x5         438             OK
S31t0519.793* 40.2        5x5         489             OK
S31t0524.771  44.4        5x5         361             OK
S31t0524.781  42.4        5x5         481             OK

S32t0483.884  44          7x9         499             OK
S32t0483.894  43.2        7x9         494             OK
S32to493.863  48.1        7x11        436             OK
S32t0493.872  48.7        7x9         371             OK
S32t0515.802  53.6        9x7         312             OK
S32t0515.811  46          9x9         439             OK
S32t0518.795  39.8        7x13        470             OK
S32t0518.804  42.4        7x13        462             OK
S32t0519.793  40.5        7x15        448             OK
S32t0519.802  38.4        7x15        445             OK
S32t0524.771  45.1        7x13        418             OK
S32t0524.781  45.3        7x13        408             OK

*All "A" area on one file.

This is all my good "A" data to date. I can find files that are full of clouds for those that want to study "bad" data.


Arne Henden aah@nofs.navy.mil
Mar 31, 1997

I've read Richmond's new Technical Note on the comparisons between the various datasets for Droege's images. Note that Tom has posted new data for these nights, with frames that hopefully overlap in V and I so that photometric transformations can be made.

While the astrometric agreements are in general very good, I have one suggestion to make. Use the GSC positions as the fiducials, and do differential comparisons of GSC vs. Richmond, GSC vs. Gutzwiller, etc. so that you do not include systematic trends in the results. I may have astrometry in one of these regions that is truely accurate (50mas or better) that might be an even better fiducial dataset. The +-1pixel offset in the x,y positions is due to whether you are using Fortran (1-relative indexing) or C (0-relative indexing), and the 15 pixel offset is due to the removal or inclusion of the beginning blind pixels.

The photometric comparisons match what I've seen from my own reductions. I am in the process of finishing a suite of programs that take the raw extraction lists and produce master mean star lists out the other end, for use in creating master lists for the smsp


Michael Gutzwiller deepsky@fuse.net
Mar 31, 1997

I've just discovered, and fixed, a bug in the star program. The bug caused the dates shown in the star list to be calculated incorrectly for certain images. Without going into the gory details, the time read from the image file was interpreted as an octal number if the hours, minutes, or seconds began with a 0. Unfortunately 08 and 09 are not good octal numbers. Fortunately the extension that Norman used is time sensitive so the files in error are easy to pick out.

Files with extensions in the range 833 - 916 will have time values in error by between eight and ten hours. These will certainly have to be re-run. (I know how you feel Tom, I did analysis on 10 nights worth of data over the weekend and was ready to upload it.)

Another set of files will have time errors of between eight and ten minutes. I don't know if these are worth rerunning or not. I probably will myself. The extensions with this error amount are:

    505 - 507, 547 - 549, 588 - 590, 630 - 632, 672 - 674,
    713 - 715, 755 - 757, 797 - 799, 922 - 924, 963 - 965
The time calculated for all other files will be correct except for a small subset with 8 to 10 seconds error, too small to worry about. The corrected version of star has been uploaded to the TASS Cincinnati site http://home.fuse.net/deepsky/programs.html