The Amateur Sky Survey Archive

Entries before the TASS listserv (the very beginning)

First I want to thank Craig Gullixson and and Bratislav Curcic
for their helpful answers to my questions in the post
"Thinking about a CCD "Device".  This has encouraged me to
tell you all what I am contemplating.

INTRODUCTION

With many others, I am sure, SL-9 caused me to drag out the 6"
reflector that was gathering dust in the garage and to
actually take it out in the yard and look at Jupiter.  This
also caused me to start lurking in this group.  Since I
have arthritis it is literally a pain in the neck to try to do
any observing.  This caused interest in the CCD devices and
started the thought process.

Next there was a post by Herbert Johnson in the thread "NASA
Creates Near-Earth Object Search Committee."  He pointed out
that the data processing requirements for sky searches are not
all that formidable and might be done with a 486 with modest
memory.

Please forgive me as I have only been reading up on astronomy
for a few weeks and so am at the really stupid mistake stage.
This is intended to be an amateur adjunct to my real project
of the moment which is building a GIANT cosmic ray telescope.
For all I know, someone is already doing what I describe
below.

GOAL

Let's say our goal is to search a hemisphere of the sky for
comets, variable stars, nova, and asteroids down to mag 15.

Using 5' cells and allowing 6 bytes per sky point this
requires 30 Mbyte to store a reference sky.  Six bytes is
enough for a magnitude and an index to other information -
such as a variable star history. With 600 Mbyte hard disks
readily available these days there is room for a considerable
history.

Thus it is a possible project for amateur computing
facilities.

THE SCHEME

Look straight up and focus the sky on a line of PIN diodes.
(CCD's, I believe, are an array of PIN diodes with a charge
bucket read out scheme built in.)  Let the earth rotate the
telescope and sweep the star images across the PIN diode line.
Read the charge accumulated on the diode every 20 seconds.  We
need only 1080 diodes to cover the 90 degress of sky, so a 10
us 16 bit converter will read them out with switching time in
20 milliseconds.  The time for readout thus introduces only a
tiny skew in the measured sky line.

THE TELESCOPE

One way to do this would be to use 6" dia f/5 mirrors.  Using
PIN diodes with a 1 mm active area, we could get 150 of them
across a 6" bar at the focus.  If we leave out a few at the
edges where the focus might not be so good (but we are
interested in light not a pin point focus) such a telescope
would cover 10 degrees and nine would be needed.  My old
Edmund Scientific catalog lists such mirrors at $166 so 9
would be within reach of a determined amateur.  You just nail
them to the roof at various angles and take data.  (I define a
determined amateur as one who drives a Yugo instead of the
Buick that he could afford and puts the difference into his
hobby.)

By my computations, such a system would deliver 6E8 photons
per second to a PIN diode from a mag 1 star.  This is based on
an un-checked calculation that estimated 3.5E6 photons per sq
cm per second reach earth from a magnitude 1 star.  Using a
factor of 2 for light loss in the mirror and quantum
efficiency of the PIN diode, and assuming 3000 electrons
sensitivity for the electronics, we should be able to see down
to mag 16 or so.  Some of you will look at that 3000 electrons
and remember numbers like 15 electrons and wonder why I am
being so conservative.  The low numbers are achieved when all
the electronics is buried in the silicon,  for this scheme
everything is out and exposed, and it will still require
cooling and will be tough electronically.  But this is
supposed to be an amateur project, and finding linear arrays
with the amplifier built in is likely to be expensive.
Numbers like $40,000 come to mind.  Individual PIN diodes are
closer to $1 each in the thousand quantity that we will need.

But there is another possible solution.  Edmund Scientific has
Fresnel lenses as are used in overhead projectors and for
solar furnaces.  They have a nice unit that is 28.5" x 37.2"
with a 54" focal length.  This is the same area as a 36" lens,
and it is a refractor.  A refractor is a big advantage for
such a set up.  One wants to have a lot of insulation around
the PIN diodes if they are to be cooled.  This is a mess for a
reflector, which would need a very long skinny diagonal
mirror.  So far, I have tested one of the 10" Fresnel lenses
and find that it can resolve 2'.  Hopefully the larger one
will be as good.  Yep, it is not color corrected.  But I could
separate a red and a green LED at 50' in my basement with the
8" focal length 10" dia lens.  Yes, it is f/0.8.  The focus was
long and fuzzy, so I think it would be OK with the PIN diodes.
Possibly one of you can help me with the calculation as to how
much the lack of color correction will spread out the focus.

This should get enough light to get to mag 20 or so with the
above assumptions.  We should be able to cover 22 1/2 degrees
with one of these lenses, so only 4 would be needed.  These
would just be big square boxes sitting on the roof,  One
would really have to remember to put the covers on when the
sun comes up or they would burn the house down.

ELECTRONICS

A typical cheap ($1.50) PIN diode has a room temperature
leakage current of 6E-11.  I would just buy 2000 for this
project and sort out the good ones (and hope that someone
else did not get to them first.)  A mag 1 star would produce a
current of 3.7E-9.  This is 60 times the leakage current and
gets to mag 5.  Cooling to -40 C would seem possible, and adds
two orders of magnitude or to mag 10.  Assuming that the PIN
diode does not saturate and the 5' resolution, a twenty second
exposure is allowed.  This adds 3 magnitudes and gets to mag
13.  Now the big guess!  How stable is the leakage current?
This is an early thing to measure, but 1% is not too much to
ask, particularly since we are cooled.  So add 5 more
magnitudes by measuring and subtracting dark current.  This
gets to mag 18, and I would be happy if that could be
achieved.

The plan is to allow each diode to accumulate charge on the
capacitance of the read out system.  A CMOS multiplexer will
dump this charge into a low noise JFET integrating amplifier.
I have built a lot of large physics data collection systems
here at Fermilab and have a good idea what can be done.  A
sixteen bit ADC will digitize the result.

Assuming that the capacitance of the PIN diode, lead, the
multiplexer switch and a storage capacitor add up to 100 pf, a
magnitude 1 star would produce 740 volts across the storage
capacitor in the 20 second exposure.  A cooled leakage current
would produce only 6 mv during the same exposure.  So the low
end is OK, but we have a problem at the high end.  We can
solve this by reading out the array 100 times during the 20
second exposure.  This is still within the range required to
scan with a single ADC but we would likely use several.  With
a scheme like this, a little noise actually helps (at the low
end - we do *not* want to read out the same bits every time)
and the dynamic range is extended two orders of magnitude over
the 16 bit ADC.  We simply add up all the charge dumps in
hardware (or software if the computer can keep up.)  There is
room for lots of tricks here to expand the range.  Log
amplifiers come to mind.  But this simple trick should give a
mag 17 dynamic range.

CONCLUSION

This is already too long for a first post on the subject.
There are clearly lots of things to discuss, like how to
normalize everything against a constantly changing sky so that
a change from the previous day's measurement can be detected.

But suppose this worked as stated, i.e. suppose we could
detect the arrival of a new light source in the observed sky
at the mag 15 level.  Would it be worth doing?

The purpose of posting this here is to get some feedback from
you dear readers.  Someone could save me a lot of work by
pointing out a glaring error.  Or by telling me that this is
already being done in some better way.  I might actually try
to do this.  It would give all those computers in my basement
(23 at last count) something to do.

As a start, this is a hand waving contest, so I have not
included the details (what few there are) of the computations.
I will be pleased to try to justify every assumption and
calculation above.

Tom Droege

This has some of the numbers revised from the earlier post.

I posted this to sci.astro and tried to post it here but only the
heading came through.  My real question it that of the movie
"Field of Dreams".  If I build it, will it be useful?  i.e. if it
can detect changes to mag 15 in a half sky survey in 5' square
cells, will it detect anything interesting?  If the answer is
yes, then comes the fun of actually trying to build it.

INTRODUCTION

With many others, I am sure, SL-9 caused me to drag out the 6"
reflector that was gathering dust in the garage and to
actually take it out in the yard and look at Jupiter.  This
also caused me to start lurking in this group.  Since I
have arthritis it is literally a pain in the neck to try to do
any observing.  This caused interest in the CCD devices and
started the thought process.

Next there was a post by Herbert Johnson in the thread "NASA
Creates Near-Earth Object Search Committee."  He pointed out
that the data processing requirements for sky searches are not
all that formidable and might be done with a 486 with modest
memory.

Please forgive me as I have only been reading up on astronomy
for a few weeks and so am at the really stupid mistake stage.
This is intended to be an amateur adjunct to my real project
of the moment which is building a GIANT cosmic ray telescope.
For all I know, someone is already doing what I describe
below.

GOAL

Let's say our goal is to search a hemisphere of the sky for
comets, variable stars, nova, and asteroids down to mag 15.

Using 5' cells and allowing 6 bytes per sky point this
requires 30 Mbyte to store a reference sky.  Six bytes is
enough for a magnitude and an index to other information -
such as a variable star history. With 600 Mbyte hard disks
readily available these days there is room for a considerable
history.

Thus it is a possible project for amateur computing
facilities.

THE SCHEME

Look straight up and focus the sky on a line of PIN diodes.
(CCD's, I believe, are an array of PIN diodes with a charge
bucket read out scheme built in.)  Let the earth rotate the
telescope and sweep the star images across the PIN diode line.
(Yes, I know this sort of thing is already being done.  In fact
talking to a programmer at Fermilab that is working on such a
survey got me to thinking about whether an amateur version was
possible.)  Read the charge accumulated on the diode every 20
seconds.  We need only 1080 diodes to cover the 90 degress of
sky, so a 10 us 16 bit converter will read them out with
switching time in 20 milliseconds.  The time for readout thus
introduces only a tiny skew in the measured sky line.

THE TELESCOPE

One way to do this would be to use 6" dia f/5 mirrors.  Using
PIN diodes with a 1 mm active area, we could get 150 of them
across a 6" bar at the focus.  If we leave out a few at the
edges where the focus might not be so good (but we are
interested in light not a pin point focus) such a telescope
would cover 10 degrees and nine would be needed.  My old
Edmund Scientific catalog lists such mirrors at $166 so 9
would be within reach of a determined amateur.  You just nail
them to the roof at various angles and take data.  (I define a
determined amateur as one who drives a Yugo instead of the
Buick that he could afford and puts the difference into his
hobby.)

By my computations, such a system would deliver 6E8 photons
per second to a PIN diode from a mag 1 star.  This is based on
an un-checked calculation that estimated 3.5E6 photons per sq
cm per second reach earth from a magnitude 1 star.  Using a
factor of 2 for light loss in the mirror and quantum
efficiency of the PIN diode, and assuming 3000 electrons
sensitivity for the electronics, we should be able to see down
to mag 16 or so.  Some of you will look at that 3000 electrons
and remember numbers like 15 electrons and wonder why I am
being so conservative.  The low numbers are achieved when all
the electronics is buried in the silicon,  for this scheme
everything is out and exposed, and it will still require
cooling and will be tough electronically.  But this is
supposed to be an amateur project, and finding linear arrays
with the amplifier built in is likely to be expensive.
Numbers like $40,000 come to mind.  Individual PIN diodes are
closer to $1 each in the thousand quantity that we will need.

But there is another possible solution.  Edmund Scientific has
Fresnel lenses as are used in overhead projectors and for
solar furnaces.  They have a nice unit that is 28.5" x 37.2"
with a 54" focal length.  This is the same area as a 36" lens,
and it is a refractor.  A refractor is a big advantage for
such a set up.  One wants to have a lot of insulation around
the PIN diodes if they are to be cooled.  This is a mess for a
reflector, which would need a very long skinny diagonal
mirror.

So far, I have tested one of the 10" Fresnel lenses and find that
it can resolve 2'.  Hopefully the larger one will be no more
than a factor of 2 worse.  Yep, it is not color corrected.  But I
could detect a red and a green LED separated 0.2" at 50' in my
basement with the 8" focal length 10" dia lens.  This is just
slightly over 1' of arc.  Yes, it is f/0.8.  The focus was long
and fuzzy, so I think it would be OK with the PIN diodes.
Possibly one of you can help me with the calculation as to how
much the lack of color correction will spread out the focus.
Looking at a star with this lens and a 10 micron pixel ccd TV
camera, the star is spread out over about 10 by 10 pixels.  This
sort of confirms the 2' measurement.


This should get enough light to get to mag 20 or so with the
above assumptions.  We should be able to cover 22 1/2 degrees
with one of these lenses, so only 4 would be needed.  These
would just be big square boxes sitting on the roof,  One
would really have to remember to put the covers on when the
sun comes up or they would burn the house down.

ELECTRONICS

A typical cheap ($1.50) PIN diode has a room temperature
leakage current of 6E-11.  I would just buy 2000 for this
project and sort out the good ones (and hope that someone
else did not get to them first.)  Using the big Fresnel lens a
mag 1 star would produce a current of 3.7E-9.  This is 60 times
the leakage current and gets to mag 5.  Cooling to -40 C would
seem possible, and adds two orders of magnitude or to mag 10.
Assuming that the PIN diode does not saturate and the 5'
resolution, a twenty second exposure is allowed.  This adds 3
magnitudes and gets to mag 13.  Now the big guess!  How stable is
the leakage current?  I am quite willing to hold the temperature
to 0.001C.  I know some of you have had trouble with temperature
controls.  This is a tough servo problem but I have done it
before on a sensitive calorimeter project.  The problem is that
the time constants are so long (one of you mentioned 2 minute
oscillation period) that the conventional analysis techniques
(bode plots, etc..) take forever.

Dark current stability with temperature is an early thing to
measure, but 1% is not too much to ask, particularly since we are
cooled.  I would be appreciative of experience that you all have
as to how dark current changes with temperature.  I know it is of
order 7% per C or double for 10 C, but what I am interested in is
how stable this value really is.  I can imagine all sorts of
creep and hysteresis and other nasty things.

So add 5 more magnitudes by measuring and subtracting dark
current.  This gets to mag 18, and I would be happy if that could
be achieved.

The plan is to allow each diode to accumulate charge on the
capacitance of the read out system.  A CMOS multiplexer will
dump this charge into a low noise JFET integrating amplifier.
I have built a lot of large physics data collection systems
here at Fermilab and have a good idea what can be done.  A
sixteen bit ADC will digitize the result.

Assuming that the capacitance of the PIN diode, lead, the
multiplexer switch and a storage capacitor add up to 40 pf, a
magnitude 1 star would produce 1850 volts across the storage
capacitor in the 20 second exposure.  A cooled leakage current
would produce only 15 mv during the same exposure.  So the low
end is OK, but we have a problem at the high end.  We can
solve this by reading out the array 500 times during the 20
second exposure.  This is still within the range required to
scan with a single ADC but we would likely use several.  With
a scheme like this, a little noise actually helps (at the low
end - we do *not* want to read out the same bits every time)
and the dynamic range is extended two orders of magnitude over
the 16 bit ADC.  We simply add up all the charge dumps in
hardware (or software if the computer can keep up.)  There is
room for lots of tricks here to expand the range.  Log
amplifiers come to mind.  But this simple trick should give a
mag 17 dynamic range.

CONCLUSION

This is already too long for a first post on the subject.
There are clearly lots of things to discuss, like how to
normalize everything against a constantly changing sky so that
a change from the previous day's measurement can be detected.

But suppose this worked as stated, i.e. suppose we could
detect the arrival of a new light source in the observed sky
at the mag 15 level.  Would it be worth doing?

The purpose of posting this here is to get some feedback from
you dear readers.  Someone could save me a lot of work by
pointing out a glaring error.  Or by telling me that this is
already being done in some better way.  I might actually try
to do this.  It would give all those computers in my basement
(23 at last count) something to do.

As a start, this is a hand waving contest, so I have not
included the details (what few there are) of the computations.
I will be pleased to try to justify every assumption and
calculation above.

Tom Droege

INTRODUCTION

First, I want to thank everyone for their contributions.  I
am buried in great ideas, and am saturated in good data.  I
will try to acknowledge everyone who has contributed (but no
today!).  That is what makes this medium work so well.  If
you think I have used one of your ideas without proper
credit, a private gentle nudge will correct the matter.  This
is a particular problem today (Memorial Day) as the print
queue is stalled and when I print to the local printer I
don't seem to get headers and ... you don't want to know!
Another problem is that I don't yet know you all well yet.

If you see that #4 above and think you have missed something,
the previous 3 notes were titled:

#1  Thinking about a CCD Device
#2  An Amateur Sky Survey Device
#3  About An Amateur Sky Survey Device

#3 above produced several favorable votes to keep up this
thread, and no flames.  So I will keep it up and use the
standard heading above with incrementing Design # so that the
notes can be collected.

I will try to repeat the end section "THE DESIGN GOAL" in
each note so that new readers will be able to understand what
is going on.  Expect the goal to change over time.

THE PROBLEM OF THE DAY

Sky Brightness

This is a compromise decision.  Going to a smaller  cell will
help find a dim object because there will be less light from
objects in the cell for comparison.  But a narrow field fills
up the disk, so 1.5 minute cells are about the limit I can
imagine for an amateur device.  At 6 bytes per cell this
requires 300 Mbytes for a hemisphere for the referance data
base, and much more space if we want to keep the last few
days of data.

OK, you all have given me a range of sky brightness of mag 17
to mag 25.5 (with a filter) per arc sec^2.  The names of the
respondents will be withheld to prevent a fist fight.  Looks
like the cheap lens below will resolve 1.5' of arc.

Pessimistic Calculation:

Starting with mag 17 per sq arc sec sky brightness, then a
1.5' lens resolution views 8100 arc seconds of sky and will
give a background of 17 - 2.5 log 8100 or mag 7.2.  A
pessimistic assumption is that we can subtract background
only to 10%.  This adds 2.5 to the above and we get to mag
9.7.

Optimistic Calculation.

Starting with mag 25.5, we have a background of 25.5 -9.7 or
15.7.  Since we are optimistic in this calculation, we
subtract background to 1%.  This adds mag 5 to the above and
we get to mag 20.7.

If we take a log mean between the two estimates, then we have
mag 15.2 and we will meet the mag 15 goal. (Grin!)

MAG 15 GOAL

Who said I knew what I was doing when I set the arbitrary
goal of a mag 15 sky survey?  I have received almost no
comment on this.  What would a good goal be?  Seems to me
that a 1.5' mag 10 survey would find a lot of comets.  This
is fun but not so good science unless we find *all* the
comets in some mag range with some error limit.

SOME LAB WORK

I built a slightly better lab set up and attempted to measure
the resolution of the two Fresnel lenses that I have.  Note
that I am using a cheap TV camera not a CCD camera designed
for astronomy to make these measurements.  I just removed the
lens and focus the image right on the CCD.  My guess is that
the camera cells are 12 micron. I cannot change the scan
rate.  It is done in little surface mount chips that I can
hardly see, so I am not too eager to carve up the PC board
and change the circuit, even if I knew what it was.

Lens #1 is Edmund Scientific cat #C32,593 ($36.75).  Note
that it is not my intention to advertise ES products.  The
numbers are stated so that others may find the item and to
give an idea of possible project cost.  This is a round lens
of 10" dia, and a focal length of 8".  It is cut with 200
Fresnel lines per inch.  I have set up some green LEDs at 50'
distance in my basement.  They are spaced at 1', 2' and 4'
and are arranged so 1/2' of arc is exposed.  Putting the TV
camera at the the focus, I can see the 2' spaced LED clearly
as two objects.  The 1' spacing is just barely separable.
The 1/2' spot size is blurred out to about 1 1/2'.  The TV
does not seem to have much dynamic range, so possibly using
ADC values and a little fitting a better measurement could be
made.  The blob I can see on the TV is about 10 pixels wide.
So this is 4.8E-3 on the chip or 2' of arc.  Again, fitting
ADC data might give a better resolution.

Lens #2 is ES cat# C32,691.($46.75)  It has 10.5"x10.5"
useful area or equal to a 11.8" dia lens.  Its focal length
is 24" and again there are 200 Fresnel lines per inch of
radius.  This of course gives a larger image.  But if
anything, the resolution is a little worse.  If I give the
above lens 1.5', I would give this one 2'.  I would have
thought that the longer focal length would be easier to make
and would thus have higher resolution.  I would appreciate
comment from a lens expert.  Also is there a better way to
make this test?  Something I can easily do without a lot of
fixturing?  Someone who is an expert might tell me which ES
lens to order next.  I am tempted to go for the 28"x37" unit
as it is equal in light gathering ability to a 36" refractor.
What amateur has one of those?

Mounting the TV camera behind lens #1 and taking it outside,
I can just make out Deneb (mag 1.3).  Again, the light is
spread out over 10x10 cells and this is running at 60 frames
a second.  So I can imagine a gain of 100 by putting all the
light in a single pixel, a gain of 60x6 or 360 by the 6
second exposure of a 1.5' line scan, and another factor of
100 in S/N for cooling.  This would be a sensitivity of
around mag 17.7 so it looks possible.

Note that I am not against using good glass lenses.  It is
just sort of a perverse sense of humor that makes me want to
do a measurement with a cheap device.  The Fresnel lenses
have a great advantage in that they are refractors, and allow
wide fields and a place outside the light space to put all
the cooling junk.

FIXING IT IN THE COMPUTER

Given that the Fresnel lenses are good to 2' and that we are
trying to measure 1.5' cells.  Is there something wonderful
we can do in the computer to sharpen up the data?  We could
easily take 30" data in each direction if we did not keep it
too long - say grind through it during the day when we have
the covers on the telescopes.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.  I am new to this game, so I may not have it
right yet.  Comments welcome on a better problem statement.

The design should be within the capabilities of a determined
amateur.  I figure this at about $2000 spread over a few
years.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

At the moment we have no concern that the world may not be
ready for the amount of data that such a project can
generate. (Hint, if good data is going down the tubes because
there is no organization, someone will step forward!)

Tom Droege

INTRODUCTION

After a lot of very positive comments to the first few posts,
they have fallen off.  That is to be expected, now I actually
have to build it.  I have succeeded in throwing several days
of mail into the trash while switching to a new machine.  So
if you expected a response to something you sent me, please
try again.

Norman Molhant described some nice tricks to obtain wide
dynamic range.  I am proud to say that I had already
independently thought of most of them.  This is one reason to
think about something before you learn too much about what
others are doing.  There is then the chance to think up
something new.

Tom Mote reminds me to not to fall off the roof and to be
sure to put on the covers when the sun is out.  I had already
worried about burning the house down in a previous note.  I
plan to locate the telescope on a little second floor balcony
off my bedroom.

Thanks to Brad for a nice discussion of CCD devices.  He says
"TI is a well known producer" of linear CCDs.  Possibly not
any more.  When I asked about the TC 103 there was a long
pause from the information center.  The part is listed as
obsolete, and they could not name any linear replacements.

I would be very interested in finding a linear CCD array with
a large cell size.  Something with 1024 cells and 25 micron
cell size would be nearly ideal.

So far I have ignored color correction.  The basement tests
seem to indicate that it is not a problem for the 1.5'
angular resolution that I seek.  Please tell me if I am
wrong.  Eventually I will get around to learning how to
compute it.

THE PROBLEM OF YESTERDAY

Sky Brightness

Looks like I have all the information that I am going to get
until I actually build something.  But a lot of numbers came
in around mag 20 for the sky background.  A 1.5' square
subtracts mag 9.8 which gives a background of mag 10.2.  A
middle of the road background subtraction of 2% adds back
4.2.  This is then the new design goal of mag 14.4.

Thanks to Benoit Shillings, Markus Buchhorn, Allen Gilchrist,
Michael Richmond, DrDarkMatter, Norman Mulhant, Herbert
Johnson, Gerhard, Brad, and others for their comments.  My
book keeping is whacko since I moved to a new computer so I
am sorry if I missed anyone.

Benoit Shillings actually went out and made a measurement, so
he wins first prize in the useful information contest.
Michael Richmond wins the doom and gloom award for his
completely pessimistic view.  I treasure a good pessimist.
They protect you from wild optimism.  But I am not discouraged.
I just know that this project is hard enough to keep down the
competition.

TODAY's PROBLEM

I am worrying about how to design the optics.  I think that I
have decided to go for Fresnel lens optics.  Others can do it
with real glass, I will go for the wide field of view and low
cost that seems possible with the Fresnel lenses and give up
on angular resolution.

1) Use the very large 28"x37"  54" fl lens and just focus it
on a string of PIN diodes.

Using 1/2 mm dia PIN diodes gives a 1.2' of arc element.  The
problem is that one cannot get them in one neat row, as the
housing is 5+ mm in diameter.  So they have to be placed in
10 or more rows.  Precision of location is not a problem, but
the curvature of the sky is. It may require a different PC
board for each Declination angle.  This is not a terrible
problem. I put it in the "pain in the neck" category.

With one of these lenses and a big square box, one could
cover 20 degrees of sky.  This requires 1000 PIN diodes and
its a strip 20" long.  One question is whether the lens will
be good over this long a strip.  Tests with the 10" lens are
encouraging.

This means that I get to design the circuits to move the
charge around and measure it.  I like this, as I know how to
do it.  There is no bucket brigade shift register.  One just
connects a CMOS switch to the diode load capacitor and sucks
off the accumulated charge.  I know how to do that, but I
also know that everything is exposed and that I will do about
an order of magnitude worse than the circuits buried in
silicon.  It is possible that the extra light from the large
lens buys back the noise loss.

2)  Use a linear array as is available from Kodak, and others
such as those used in FAX machines.

The problem here is that these devices have small cells.
They seem to range from 7 to 13 microns.  To match them to a
single lens and 1.5' of arc would require one with a focal
length of about 1".  The 10" dia lens I have has a focal
length of 8".  Thus the image is about 10x larger than I
want.

I need to figure out if I can add lenses to reduce the image
size without losing a lot of light in the process.  So I am
off to a self taught course in Optics 101.  I am not very
happy with the books I have so far on the topic.  Full of ray
diagrams for a zillion special cases.  So far I do not have
a single example of a wide angle lens for study.  I guess I
am looking for a philosophy of wide angle lens design book.
I can work out the mechanics when the time comes, what I want
to know about are the successful tricks.

I would appreciate advice from any optics experts about how
to reduce the image size without losing any light.  I notice
that the ES catalog does not have any negative Fresnel
lenses.  Besides the 28"x37" 54"fl lens, there is a 10" dia
8" fl, a 6"x6" 3" fl, a 1.3" 0.85 fl, and about 40 others.
My current plan is to try stacking them up.  The ultimate
objective is to focus 1.5' of sky onto a 10 micron square
pixel.  I have no objection to use of glass lenses if
reasonable sized ones work out better.  At least the 10" dia
8" fl lens will just about resolve 1.5'.

If 1.5' of sky can be reduced to match a 10 micron CCD cell
size, then we only need a good area of 0.32" to map 20
degrees of sky.  I would be happy with that.

I hope that one of you has a great optics design program and
wants to exercise it to keep down it's fat.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 14 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  I figure this at about $2000 spread over a few
years.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 20 degrees of sky to a with a
single set of optics.  To limit the storage to possible
values for an amateur, we will limit the resolution of the
sky map to 1.5' of arc.

Tom Droege

INTRODUCTION

Help!!  After you read the design below and quit giggling,
please offer a better solution.  For the moment, I am
determined to find a Fresnel lens solution.  I could always
be talked out of it by a better idea.  Some of my goals are
to amuse you and to make you feel superior.  It is good for
all concerned.


TODAY's PROBLEM

I am filling sheets of graph paper with ray traces.  The goal
is to use the 10" dia, 8" fl Fresnel lens as the primary
lens.  Then do something to focus 1.5' of sky onto a 10
micron CCD pixel.  Another goal is to scan 20 degrees of sky
per telescope.


Start with the 10" dia +8" fl lens and put a 3" dia, -4" fl
lens at the focus.  If I understand it correctly, this
preserves the angular sky image, and collects all the light
from 21 degrees of sky into a parallel beam.  44" further
back from the first focus, put a 3" dia +4" fl lens.  Now at
a focus point 4.4" back from the third lens place the CCD. A
rough diagram:

|
|
|       |                                          |
|       |                                          |   'Focus
|       F1                                             F2
|

10 dia  3" dia                                     3"dia
8" fl   -4" fl                                     4" fl
L#1     L#2                                        L#3

I compute that 20 degrees of sky produce a 2.9" image at F1.
The lens L#2 at F1 moves the image to infinity but preserves
the apparent size at F1.  It also focuses most of the the
light collected by L#1 and transports it to L#3.  L#3
demagnifies the image at F1 by 0.1.

Twenty degrees of sky produce a 2.9" image at F1.  This is
0.0036 for each 1.5' of sky.  The image at F2 is 0.00036" for
each 1.5" of sky or 9 microns for 1.5 minutes of sky.  This
is in the range to match available linear CCD devices.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 14 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  I figure this at about $2000 spread over a few
years.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 20 degrees of sky to a with a
single set of optics.  To limit the storage to possible
values for an amateur, we will limit the resolution of the
sky map to 1.5' of arc.

Tom Droege

INTRODUCTION

Well, there was no response to my plea for optics help.  But
I move too fast for moss to grow on a design, so I have
abandoned mucking around with optics to make the image
smaller.  The plan is now to use a smaller, shorter focal
length lens.

DrDarkMatter has described several schemes using camera
lenses for "parking lot" surveys.  I await his paper in the
fall "CCD Astronomy" (if S&K ever sends the stuff I ordered).
I am still working toward something that would be really
cheap to mass produce (well, in hundreds).  So I will try to
make do with a $16 Fresnel instead of a more expensive camera
lens, and try to use a mass produced CCD made for scanners.

I have received a lot of direct encouragement to keep making
these posts.  But there have been only a few comments to
sci.astro. Still, this seems like the right place to post.  I
will keep it up until flames start coming in.

I am a little embarrassed to present the design below.  After
thinking about designs that had the light gathering capacity
of a 36" refractor, I am now down to something that would
almost fit in your pocket.

TODAY's SOLUTION

The problem has been to match the optics of a large lens to
the of order 10 micron CCD pixel size.  OK, I now give up and
just match the optics of a small lens to a CCD.

We pick the 2" dia Fresnel lens with a 1.3" focal length from
the Edmund Scientific catalog (We will possibly find a
cheaper source).  This produces a 0.00057" image for 1.5' of
arc.  (Actually, I do not have great hope that the image will
be this small for a star.  But the 10" lens produced a mushy
0.003" star image so I am only asking that the image size
scale with the lens size.)  This is equal to 14 microns.
Looking at the pile of literature that Kodak just sent me, I
find the KLI-2103.  It is a 2098 pixel RGB linear array with
a 14 micron cell size.  The response is as given in the table
below in Volts/uj/cm^2.

A pixel is .0014 cm on a side.  So we multiply by
1/(0.0014^2) or 5.1E5 to get volts per uj per pixel. A uj is
6.2 E12 electrons volts.  Divide by 2.5 ev per photon to get
roughly 2.5 E12 the number of photons in a uj.  If done
right, we multiply Volts/uj/cm^2 by 5.1 E5 and divide it by
2.5E12 to get volts per converted photon.

---- From Kodak literature--     -Computed--
WvLength, nm   Volts/uj/cm^2     Volts/Photon
450                6              1.2E-6
550               12              2.4E-6
650               20              4.1E-6

Note these numbers are less than the KAF-0400 numbers which
is 10 E-6 volts per electron which about accounts for the
filters, the difference in size and the quantum efficiency.
At least I argue that I am in the right ball park, even if no
one is playing these days.  As always, I welcome anyone who
wants to check my numbers.

Picking the 550 nm numbers, a mag 1 star at 3.5E6 photons per
second, a six second exposure allowed from a 1.5' line scan,
and a 2" dia lens, we get 1000 volts per 6 second exposure
for a mag 1 star.  The device apparently has a one volt
dynamic range.  So we have to scan it 1000 times in 6
seconds.  Possible, but adding up the separate conversions
will be a pain for simple electronics and an order or two of
magnitude too fast for a 486 to keep up - there is always the
Pentium.  At the other end, the dynamic range claimed is 72
db or 4000 to one.  This means we should be able to measure a
range of 4000 * 1000 or 4E6.  This is mag 16.5 below mag 1 or
mag 17.5.

OK, this says I can see mag 17.5 with this scheme.  A very
rough trial program (Compiled Basic) on a 486 looks like it
will perform 100 single color scans in 6 seconds.  So a mag 1
star will be 10x saturation.  OK, a mag 3.5 just saturates
the system.  I can live with that.  The KLI-2103 does not
have anti-blooming features, so we will have to live with a
mess around mag 3.5 and up.  This gives a 400,000 to one
dynamic range if we do the ADC stuff right.

George Aumann has a nice write up in the CCD group today
where he describes what he actually achieved with a 6" lens
and a TC211 based camera with an 8 bit ADC.  He measures a
limiting magnitude of 15.1 and a SNR of 3.  The above scheme
loses mag 2.4 by the smaller lens.  Something is gained back
by the wider dynamic range ADC that I plan which uses
multiple samples per cell and a 16 bit ADC.  So there is hope
that it might work.

This device will have to be cooled for maximum sensitivity.
Looks like it will drift 1/2 full scale in a second at room
temperature.  I would cool it moderately, with tight
temperature control (0.01 C or better) in an attempt to hold
the dark current constant.  If we read out 10 times a second,
then 60 ea 16bit measurements can be accumulated per 1.5' cell.
Norman Molhant has suggested time random samples to further
improve the average scheme.  This is a nice idea, and I may
be able to implement it in software.

This is not the best match, but I happen to have some 10us 16
bit ADC that I understand very well.  So it will do for a
start.  Later we could use a 12 bit 1 us converter, a local
memory, and an adder accumulator.  This would allow 2000
samples per 6 seconds of a 1.5' sky cell and would achieve a
true 17.2 mag dynamic range with no saturation at the bright
star end.  But there is no sense in building a lot of digital
logic (which I hate to design) until the scheme is proven.
The proposed scheme can be done with a very simple PC
interface which drives everything (including the CCD clock
lines) from I/O pulses (with a few registers).

Since there are three slightly spaced rows, this will
automatically be a fast moving target detector.  The RGB
columns are spaced by 8 cells which is 12' of arc or 48
seconds of time.  Nearly simultaneous hits in the three
columns will indicate a fast moving object.  Tom Tongue wants
me to look for meteors.  Looks like this would watch a very
small piece of sky (1.5'x 50 degrees straight up) and count
meteors.

Markus Buckhorn wanted me to do a two color survey.  He said
it "would be more than twice as useful".  OK Markus, how
about three?  Comments on the RGB wavelengths wich I presume
do not match lines favored by astronomers??

One nice feature is that this device will cover 51 degrees of
sky, and gets three color samples if we choose to measure
them all.  There is a computational mess with the curved sky
if we get too close to the pole, but that is what computers
are for.  We could easily fit two of the line scanners spaced
30 sky degrees or so behind the lens.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 14 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  I figure this at about $2000 spread over a few
years.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 20 degrees of sky to a with a
single set of optics.  To limit the storage to possible
values for an amateur, we will limit the resolution of the
sky map to 1.5' of arc.

Tom Droege

INTRODUCTION

This series of notes documents the development of a design
for an amateur whole sky survey device.  See the note at the
end for a description of what this means.  I have been making
these posts to sci.astro, but this now appears to be the
appropriate group.  As you can see, this is the eighth item
in this series.

I try to cover the logic without filling the page with
detailed computations.  If any of you cannot follow, I will
be pleased to explain items.  You should not assume that I
know what I am doing unless I am covering electronics.

We continue flubbing around trying to figure out a good
design.  Norman Molhant has provided a number of good
comments so we have again made modifications. DrDarkMatter
can chuckle that I have come around to his way of thinking.

I have received a lot of direct encouragement to keep making
these posts.  But there have been only a few comments to
sci.astro. Let's see how sci.astro.amateur responds.

TODAY's SOLUTION

The problem has been to match the optics of a large lens to
the of order 10 micron CCD pixel size.  OK, I now give up and
just match the optics of a small lens to a CCD.

Everyone says a fresnel lens will not work.  I am not so
sure, but for the moment will switch to something that
everyone will agree will work in order to get on with
something that will make some measurements.

We will now consider a 50 mm focal length, F/1.9 lens.  This
seems to be a readily available item.  Using a CCD with a 14
micron pixel, we have an angle covered by the pixel of
14/50000 radian or 0.96 degree.  A 2048 pixel linear CCD
would then cover 32.8 degrees.  This linear CCD would be 28
mm long, so we would expect that a lens designed for 35mm
film would have a flat enough field that it would work fairly
well.

Looking at the pile of literature that Kodak just sent me, I
find the KLI-2103.  It is a 2098 pixel RGB linear array with
a 14 micron cell size.  The response is as given in the table
below in Volts/uj/cm^2.

A pixel is .0014 cm on a side.  So we multiply by
1/(0.0014^2) or 5.1E5 to get volts per uj per pixel. A uj is
6.2 E12 electrons volts.  Divide by 2.5 ev per photon to get
roughly 2.5 E12 the number of photons in a uj.  If done
right, we multiply Volts/uj/cm^2 by 5.1 E5 and divide it by
2.5E12 to get volts per converted photon.

---- From Kodak literature--     -Computed--
WvLength, nm   Volts/uj/cm^2     Volts/Photon
450                6              1.2E-6
550               12              2.4E-6
650               20              4.1E-6

Note these numbers are less than the KAF-0400 numbers which
is 10 E-6 volts per electron which about accounts for the
filters, the difference in size and the quantum efficiency.
At least I argue that I am in the right ball park, even if no
one is playing these days.  As always, I welcome anyone who
wants to check my numbers.

Picking the 550 nm numbers, a mag 1 star at 3.5E6 photons per
sq cm per second, a four second exposure allowed from a 1'
line scan,  and a 1" dia lens, we get 168 volts per 4 second
exposure for a mag 1 star.  The device apparently has a one
volt dynamic range.  So we have to scan it 168 times in 4
seconds.  Possible, but adding up the separate conversions
will be a pain for simple electronics and an order or so of
magnitude too fast for a 486 to keep up - there is always the
Pentium.  At the other end, the dynamic range claimed is 72
db or 4000 to one.  This means we should be able to measure a
range of 4000 * 168 or 4E6.  This is mag 14.5 below mag 1 or
mag 15.5.

OK, this says I can see mag 15.5 with this scheme.  A very
rough trial program (Compiled Basic) on a 486 looks like it
will perform 10 three color scans in 4 seconds.  So a mag 1
star will be 17x saturation.  OK, a mag 4 just saturates
the system.  I can live with that.  The KLI-2103 does not
have anti-blooming features, so we will have to live with a
mess around mag 4 and down.  This gives a 40,000 to one
dynamic range if we do the ADC stuff right.

George Aumann has a nice write up in the CCD group where he
describes what he actually achieved with a 6" lens and a
TC211 based camera with an 8 bit ADC.  He measures a limiting
magnitude of 15.1 and a SNR of 3.  The above scheme loses mag
3.9 by the smaller lens.  Mag 5.5 is gained back by the wider
dynamic range ADC that I plan which uses multiple samples per
cell and a 16 bit ADC.  So there is hope that it might work.

This device will have to be cooled for good dark current
subtraction.  It looks like it will drift 1/2 full scale in a
second at room temperature.  I would cool it moderately, with
tight temperature control (0.01 C or better) in an attempt to
hold the dark current constant.  If we read out 2.5 times a
second, then 10 ea 16bit measurements can be accumulated per
1' cell.  Norman Molhant has suggested time random samples to
further improve the average scheme.  This is a nice idea, and
I may be able to implement it in software.

This is not the best match, but I happen to have some 10us 16
bit ADC that I understand very well.  So it will do for a
start.  Later we could use a 12 bit 1 us converter, a local
memory, and an adder accumulator.  This would allow 200
samples per 4 seconds of a 1' sky cell and would achieve a
true 14.7 mag dynamic range with no saturation at the bright
star end.  But there is no sense in building a lot of digital
logic (which I hate to design) until the scheme is proven.
The proposed scheme can be done at the start with a very
simple PC interface which drives everything (including the
CCD clock lines) from I/O pulses (with a few registers).

Since there are three slightly spaced rows, this will
automatically be a fast moving target detector.  The RGB
columns are spaced by 8 cells which is 12' of arc or 48
seconds of time.  Nearly simultaneous hits in the three
columns will indicate a fast moving object.  Tom Tongue wants
me to look for meteors.  Looks like this would watch a very
small piece of sky (1.5'x 50 degrees straight up) and count
meteors.

Markus Buckhorn wanted me to do a two color survey.  He said
it "would be more than twice as useful".  OK Markus, how
about three?  Comments on the RGB wavelengths wich I presume
do not match lines favored by astronomers??

One nice feature is that this device will cover 32 degrees of
sky, and gets three color samples if we choose to measure
them all.  There is a computational mess with the curved sky
if we get too close to the pole, but that is what computers
are for.  We could easily fit two of the line scanners spaced
30 sky degrees or so behind the lens.

Note that previous posts have discussed the problem of the
light background caused by looking at such a large area of
the sky per pixel.  We cannot cover the whole sky looking at
a small area as this requires too much storage.  Various
estimates have placed the background around mag 20 per sec
sq.  Looking at 1' means that the background will be 3600
times mag 20 per sq sec or mag 11.  To be sensitive at the
mag 15 level means that we must be able to subtract
background to 4 mags or a few percent.  This seems possible.

Note that this takes 150 Mbytes of storage to keep the most
basic sky map for a 30 degree x 360 degree path.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  I figure this at about $2000 spread over a few
years.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

OK, we have settled on a design and most of the parts are
here.  I should get a good start on construction this
weekend.  I wish to thank all those who have contributed,
about a dozen.  I particularly want to thank Norman Molhant
who took the trouble to read and critique the various posts.
He added a number of things to the design that would not have
occurred to me.

We note that the design is far from where we started.  That
is what happens when you listen to a lot of good advice.  I
have not really given up on fresnel lenses, they just do not
match the available CCD components.

This will be the last Design #x post.  After this we will
label the posts "Test#" or "I Give Up#".  I have no idea as
to whether this will be a useful device or not.

This will be posted to sci.astro as it now seems to be the
appropriate group.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

OPTICS

I have purchased a f/1.4, 50mm focal length Cannon camera
lens ($98.00).  I have also purchased a KLI-2103 from Kodak
($350.00). This combination will focus 0.96' of arc onto a
single 14 micron pixel of the linear CCD.  The 2103 contains
2098 active pixels for an angular coverage of 33 degrees.
The KLI-2103 contains three rows of PIN diodes, and each has
it's own CCD read out strip.  RGB filters in front of each
strip allow three color detection.  The separate CCD read out
allows exposure of the next sky strip while the previous one
is being processed.  The dynamic range quoted is 72 db, with
a maximum well capacity of 340,000 electrons.  Given a mag
one star produces 1E7 photons per sq cm per second, and given
a 10 sq cm lens, and given a loss of a factor of 10 or so for
quantum efficiency and reflections on the optics; then the
well fills up in 1/30 of a second.  At the other end, 72db
implies a 4000/1 dynamic range.  This means a sensitivity
level of 340,000/4000 or 85 electrons. Comparing 4E7
electrons for a mag 1 to 85 electrons gives a range of mag 14
below mag 1 or just the mag 15 we seek.  But we will have to
do some tricks like variable time slices to actually get
there.

ELECTRONICS

This first version will run with electronics that is as
simple as possible.  We will just drive the various phase
lines on the CCD with pulses almost directly from the
computer I/O lines.  I have a general purpose 16 bit ADC card
that I have designed for my home projects that is easy to use
and which contains some level and pulse circuits besides the
differential input to the ADC.

The plan is to have one I/O pulse do the transfer of the
accumulated charge into the CCD output shift register, and a
second I/O pulse shift the charge to the output.  At the
output buffer we will do double correlated sampling to remove
drift.  A third I/O pulse then clears the charge.

All this looks like it will take about 0.2 second per three
color scan and with everything stepped by computer I/O.  This
is a factor of 6 too slow to keep up with the light from a
mag 1 star.  So a mag 3 star will just not saturate the CCD.

We can live with that.  There are not so many bright stars
that the blooming around them will be much of a problem.

Dark current is a problem.  It amounts to 125,000 electrons
per second in a well that can hold 340,000.  We plan to cope
with this by rapid scanning and tight temperature control.

PACKAGING

We plan to put everything in a 8" x 10" x 6" deep NEMA box.
This is the kind of box electricians use for control circuits
and they are designed to be oil proof.  I will mount the box
pointing straight up.  The plan is to put the lens and the
CCD on a three point suspension.  Use of a circular level
will then allow pointing straight up.  Unfortunately, the
trees are in the way of Polaris at my balcony location, so I
will have to guess at north.  I have not yet decided whether
the complication of a North adjustment is warranted.

I will just open up the lid on the box when the weather is
right.  When it snows, etc., the box lid makes it
weatherproof.

TEMPERATURE CONTROL

I plan to glue a strip of copper or silver (depending on
scrap size I have) on the back of the CCD package.  This will
contain a temperature sensor (AD590).  I have some of the
miniature Melcor TEC which will be lined up under the CCD and
which will dump their heat/cold into a large air cooled heat
sink glued below.  The idea is not so much to cool the CCD,
but to hold it at very constant temperature.  I am shooting
for 0.01 C.  Since good viewing here in Chicago is apt to
come in December/March, we will often have mother nature on
our side.  We will bring out a lead so we can command a
temperature that is near but above ambient to prevent
condensation.

SKY SUBTRACTION

After much discussion with Norman Molhant, I concluded that
line by line sky subtraction is desirable.  We will therefore
have a second lens looking up which sees the average
brightness of the sky around the CCD strip.  Present plan is
to use one of the small fresnel lenses, and to focus the sky
on a strip somewhat wider than what the CCD sees.  This will
focus on a ground glass screen, and we will look at it with a
few PIN diodes.  This will allow line by line sky brightness
subtraction to compensate for things like clouds moving in
front of the moon, etc..

COMPUTER

I will start by using one of the computers from the basement.
It is a sad commentary on something, when I find that I
consider a 25MHz 386SX to be almost junk. But we will start
with something like that.  Then when we start taking data and
doing all those floating point operations, I will buy an 486
or a Pentium upgrade board.

SOFTWARE

We will start by writing routines in QuickBASIC because that
is what I am familiar with.  I hear groans from all the
professional programmers out there.  But I have found that
when I do it myself, that it gets done on schedule.

Later, and if this mess works, I plan to offer to trade
hardware packages for software tasks.  The goal is still to
get a network of sky survey stations with standard hardware
and software.

DATA PROCESSING

After thinking about time for a while, I realize that
comparing one days scan to the next is non trivial.  A multi-
station amateur network cannot afford accurate enough clocks
so that one day's sky map can be compared directly to the
next.  Looks like this will require a few part per million
clock.  This is possible for someone willing to do tight
temperature control and aging, but certainly beyond the clock
in the PC.  Possibly we will build such a clock into any
procuction design.  In the meantime, we will have to face
the fact that one day's sky map does not match the next in
time and do something heroic in software.

Tom Droege

INTRODUCTION

I have been doing some tests.  Pretty good results have been
obtained by tight temperature regulation of the CCD.  There
is much more to do.

THE PROBLEM

The dark current of a CCD varies with temperature.  The KLI-
2103, for example, has a dark current of 0.02 pa/pixel at
25 C.  This works out to be 124,000 electrons per second in
a chip with a full scale of 344,000 electrons.  The dark
current doubles for a 9 C increase.  The ninth root of 2
gives an 8% change per C.  Let us consider two operating
points and the effect of temperature:

Temperature    Dark Current    Electrons/C     Noise
C              Electrons/Sec   /Second         Electrons
                               Change
25 C           124,000         9920            352
-29 C             1937          155             44

The Electrons/C/Second column, is just the effect of a 1 C
change on the dark current.  For this device the output is
8 uv per electron.  The Noise Electrons column is just the
square root of the number of dark current electrons.  Those
with more statistical knowledge than me (almost everyone) may
have a better way to characterize the noise.

OBSERVATIONS

First, it is not hard to imagine that those using the pump in
a bucket of ice scheme for cooling a flat out thermoelectric
device will see several degrees C variation in the CCD
operating temperature.  There are lots of reasons for this.
The cooling water does not take the same path through the ice
bath all the time.  If it does it melts the ice along the
path.  It is easy to imagine an ice block drifting to a
position where it suddenly supplies more cooling.  There are
also some reports that have filtered back to me that confirm
this.  I have also a lot of experience running such systems
for precision calorimetry.

So I give the ice bucket scheme +/- 1 C variation over time.
I think this errors on the low side.

Another problem with cooling below ambient is condensation.
The CCD and its window must be sealed up from moisture or
evacuated, etc.  A real pain if costs are to be kept down.

I am now running my CCD with a servoed temperature control.
It is presently good to of order +/- 0.02 C.  This is a
breadboard, and I made a lot of mistakes in the layout.  I
have done +/- 0.0005 C.  I expect to do closer to this value
on the next design.

A FIRST SOLUTION

Suppose we run the cooled device as at present, and tightly
regulate as above at room temperature.  Then the third
column is now electrons of drift over the expected
temperature variation and we have:

Temperature  Dark Current    Electrons  Noise       Combined
C            Electrons/Sec   /Second    Electrons
                             Change
25 C          124,000          396         352        530
-29 C            1937          310          44        313

The combined column is now the drift and the noise combined
in quadrature.  Again, those who understand the statistics
can suggest a better combination.

So we see that we can do almost as well regulating as
cooling.  And there are no condensation problems.  So the
system can be very simple.  I could as easily make the
uncooled system look better above by a different choice of
operating points.  Obviously one would want to run a
regulated system as cool as possible.  My simple device will
go down to zero C easily, but I could not operate there as
I am not prepared for condensation.  I can and will operate
just above the dew point.

A MESSAGE TO THOSE COOLING THEIR CCD's

Regulate the temperature.  But it is not so easy as you will
find when you try to close the loop.  My little device which
was optimized for a short thermal time constant has about a
ten minute time constant.  This requires large R and C in
the amplifier compensating components.  I will be happy to
help anyone brave enough to close their regulation loop.

Tom Droege

INTRODUCTION

The sky survey machine is now up and working, and we have
seen stars.  There is much work to do before it is a useful
device, but we have gone from talking about a design to
actually looking at stars in 50 days.  Read "THE DESIGN GOAL"
and the material following below if this does not make sense
to you.

At the moment, we have only the bare minimum stuff working to
go out and look.  The sky brightness and temperature control
circuits are not yet hooked up.  There is unlimited software
fun ahead.

Where we are presently scaled, there are 6000 ADC counts
between dark and saturation.  Looks like the Kodak device
meets its specification.  At the moment we are combining
four pixels in each direction for 4' resolution on the
screen.  This is because we only have a 640 dot VGA at the
moment, and so need to squeeze it down to watch the whole
scan at once.  With this set up, the noise is of order 1
count rms after averaging 30 scans of one CCD color over 16
seconds.  At the moment, I cannot scan at the full 486 speed
because of some electronic problem - like not allowing the
ADC input time to settle.

I have done some temperature tests.  I find that the dark
current changes of order 100 counts per C at 25 C.  This
checks out with the data sheet.  So it is really worth while
to make a temperature correction.  Even better will be the
plan to hold it at constant temperature. The indications are
that tight temperature control will allow one count ADC dark
current stability at room temperature.

With this set up, we can see stars, and a lot more of them
than are on my mag 6 sky atlas.  Until I do a little more
software work, and can see a constellation that I can
recognize, I will not be able to determine the sensitivity.
My bet at the moment is that it is somewhere around mag 8 in
full moonlight.

Focusing was a pain.  But we seem to have a fair one after
mucking around a lot.  Norman Mulhant suggested a nice method
- looking at the "noise".  A good focus will have higher
frequency content than a bad one when looking at stars.
Looks like a job for the FFT.

One would think that it would be easy to recognize the sky
when looking at a 30x20 degree slice as I do on the CRT
display.  But so far I do not recognize anything.  But I
have only taken a few good scans.  On the other hand, block
out a 30x20 degree slice on a sky chart and see how easy it
is to see where you are. But some of you will know.  Sort of
like "old" navy pilots know where they are at sea.

I am presently pointing straight up.  This means I am looking
between 35 and 65 degrees.  Cassiopeia should be in there to
see, but last night a big cloud came through to block it out.
Possibly tonight I will have better luck.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

OPTICS

I have purchased a f/1.4, 50mm focal length Cannon camera
lens ($78.00 used).  I have also purchased a KLI-2103 from
Kodak ($350.00). (I now have a sample of a $20 device that is
easier to use, and have ordered 20 of them.) This combination
will focus 0.96' of arc onto a single 14 micron pixel of the
linear CCD.  The 2103 contains 2090 active pixels for an
angular coverage of 33 degrees.  The KLI-2103 contains three
rows of PIN diodes, and each has it's own CCD read out strip.
RGB filters in front of each strip allow three color
detection.  The separate CCD read out allows exposure of the
next sky strip while the previous one is being processed.
The dynamic range quoted is 72 db (75 was measured with
double correlated sampling), with a maximum well capacity of
340,000 electrons.  Given a mag one star produces 1E7 photons
per sq cm per second (some tell me it is more like 2E5), and
given a 10 sq cm lens, and given a loss of a factor of 10 or
so for quantum efficiency and reflections on the optics; then
the well fills up in 1/30 of a second.  At the other end,
72db implies a 4000/1 dynamic range.  This means a
sensitivity level of 340,000/4000 or 85 electrons. Comparing
4E7 electrons for a mag 1 to 85 electrons gives a range of
mag 14 below mag 1 or just the mag 15 we seek.  But we will
have to do some tricks like variable time slices to actually
get there.

ELECTRONICS

This first version will run with electronics that is as
simple as possible.  We will just drive the various phase
lines on the CCD with pulses almost directly from the
computer I/O lines.  I have a general purpose 16 bit ADC card
that I have designed for my home projects that is easy to use
and which contains some level and pulse circuits besides the
differential input to the ADC.

The plan is to have one I/O pulse do the transfer of the
accumulated charge into the CCD output shift register, and a
second I/O pulse shift the charge to the output.  At the
output buffer we will do double correlated sampling to remove
drift.  A third I/O pulse then clears the charge.

All this looks like it will take about 0.2 second per three
color scan and with everything stepped by computer I/O.  This
is a factor of 6 too slow to keep up with the light from a
mag 1 star.  So a mag 3 star will just not saturate the CCD.

We can live with that.  There are not so many bright stars
that the blooming around them will be much of a problem.

Dark current is a problem.  It amounts to 125,000 electrons
per second in a well that can hold 340,000.  We plan to cope
with this by rapid scanning and tight temperature control.

PACKAGING

We plan to put everything in a 8" x 10" x 6" deep NEMA box.
This is the kind of box electricians use for control circuits
and they are designed to be oil proof.  I will mount the box
pointing straight up.  The plan is to put the lens and the
CCD on a three point suspension.  Use of a circular level
will then allow pointing straight up.  Unfortunately, the
trees are in the way of Polaris at my balcony location, so I
will have to guess at north.  I have not yet decided whether
the complication of a North adjustment is warranted.

I will just open up the lid on the box when the weather is
right.  When it snows, etc., the box lid makes it
weatherproof.

TEMPERATURE CONTROL

I plan to glue a strip of copper or silver (depending on
scrap size I have) on the back of the CCD package.  This will
contain a temperature sensor (AD590).  I have some of the
miniature Melcor TEC which will be lined up under the CCD and
which will dump their heat/cold into a large air cooled heat
sink glued below.  The idea is not so much to cool the CCD,
but to hold it at very constant temperature.  I am shooting
for 0.01 C.  Since good viewing here in Chicago is apt to
come in December/March, we will often have mother nature on
our side.  We will bring out a lead so we can command a
temperature that is near but above ambient to prevent
condensation.

SKY SUBTRACTION

After much discussion with Norman Molhant, I concluded that
line by line sky subtraction is desirable.  We will therefore
have a second lens looking up which sees the average
brightness of the sky around the CCD strip.  Present plan is
to use one of the small fresnel lenses, and to focus the sky
on a strip somewhat wider than what the CCD sees.  This will
focus on a ground glass screen, and we will look at it with a
few PIN diodes.  This will allow line by line sky brightness
subtraction to compensate for things like clouds moving in
front of the moon, etc..

COMPUTER

I will start by using one of the computers from the basement.
It is a sad commentary on something, when I find that I
consider a 25MHz 386SX to be almost junk. But we will start
with something like that.  Then when we start taking data and
doing all those floating point operations, I will buy an 486
or a Pentium upgrade board.

SOFTWARE

We will start by writing routines in QuickBASIC because that
is what I am familiar with.  I hear groans from all the
professional programmers out there.  But I have found that
when I do it myself, that it gets done on schedule.

Later, and if this mess works, I plan to offer to trade
hardware packages for software tasks.  The goal is still to
get a network of sky survey stations with standard hardware
and software.

DATA PROCESSING

After thinking about time for a while, I realize that
comparing one days scan to the next is non trivial.  A multi-
station amateur network cannot afford accurate enough clocks
so that one day's sky map can be compared directly to the
next.  Looks like this will require a few part per million
clock.  This is possible for someone willing to do tight
temperature control and aging, but certainly beyond the clock
in the PC.  Possibly we will build such a clock into any
procuction design.  In the meantime, we will have to face
the fact that one day's sky map does not match the next in
time and do something heroic in software.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

A PRODUCTIVE WEEKEND

Friday night I finally recognized a sky picture.  It turns
out that I was scanning backwards.  This mirrored the image.
I was also mistaken about the direction of "up" at the
approximate 42 degree latitude of Chicago.  For some reason
in a mistake that I repeated over and over, I thought that up
would be 90-42 degrees on the sky charts.  Finally Algol and
Mirfak came by and I could recognize the constellation
Perseus.  No wonder I could never see Cassiopeia, it was not
in my field of view!

Saturday was spent furiously adding the temperature control
circuits.  It was a race to get it done with expected clear
skys.  I will save the discussion about the reason one wants
very accurate temperature control for a CCD for the CCD
group.  Will send it to anyone interested.  I really made a
mistake and picked the wrong device out of my stock of Melcor
Corp. TEC devices.  It required really high drive current,
and soon I had an oscillator.  By the time I had this fixed,
I had made a wreck of the board, and unknowingly bent some of
the pins of the CCD chip together.  So nothing worked.  This
kept up to 3 AM.  Finally I said something that lost me my
accumulated attaboys, and went to bed.

Sunday morning was better.  I found a diode that had been
shorted out in one of the arcs and sparks of getting
something working.  CCD's are very fussy about having all
the bias voltages just right.

I am really not a very careful worker.  I should always have
a large stock of spare parts.  But then I get things done in
a hurry.  I think it is a less costly way to proceed - but
some of you will be shocked.  In the end I cleared all the
sick parts, and none of them was the $350 Kodak chip.  So
back in operation.  But without the CCD temperature control
working.

Sunday night was much better.  First I knew where I was
looking and had the scan direction right.  Immediately on
turn on I spotted Cygnus.  But then what a boring stretch of
sky!  Sure, later when I have more sensitivity it will be
interesting, but for the moment give me a mag 3 or above
every ten minutes or so.  I am looking between 23 and 55
degrees of Declination.  There is about three hours of not
very much between Deneb and Alpheratz which I missed because
I was making a software change.  Then I got a very nice
picture of everything between Mirach and Almach. Almach
was much dimmer than expected, but when I went out to close
up the scope, I saw that clouds had moved in and I was
looking at Almach through pretty thick clouds.  Most of this
last set of observations was in near full moonlight, and some
clouds.

The Edmund star chart I am using only goes to mag 6.  I could
spot almost everything on the chart.  There were about again
as many stars that were not on the chart.

So let's assume I am at mag 7.  How will I get to mag 15?
Actually, I don't think I will get there without a larger
lens.  But here is how to get there:

Focus.

It looks like a bright star is spread out over 5x5 of the
four minute pixels.  If we can improve this to 2x2 by a
better focus, then we should gain a factor of 6.  Norman
Molhant and I have cooked up a scheme using the FFT where we
look at the high frequency content of the scan line and plot
it against lens position to find the best focus.

Can someone tell me the expected intensity profile when an
out of focus lens looks at a point source (star).  A couple
of stars that I looked at showed almost flat intensity over
the large out of focus circle.  But note that my display
algorithm could have masked a variation.

Binning.

I am presently merging four 1' cells along the CCD, and
taking a 16 second exposure.  (Actually 30 shorter
exposures.)  This reduces both the noise and the signal, but
the signal is reduced by a factor of 16 over using 1' cells
and the noise is reduced by sqrt of 16 or 4.  (Happy to have
someone discuss this, I am not an expert.)  So we should gain
a factor of 4 just by unbinning.

Sampling.

We are presently taking 250 microseconds to read a cell with
a 10 microsecond converter.  Should be able to improve this a
factor of 10 to 25 microseconds.  This should gain a factor
of 3.

Temperature Control.

I am presently setting a high threshold for star
identification.  This could be lower if the flat field
subtraction was constant with time.  It is not.  In
particular, it changes with temperature.  Not only that, but
some pixels change more than others, so a constant added as a
function of temperature does not work unless a different
constant is used for each pixel.  Not too difficult for a
linear array though.  Only 2090 correction factors. If I hold
the CCD chip temperature constant, I could set a lower
threshold.  I estimate a factor of three.

OK, the above get to about mag 13.  With a 1' focus, we are
at mag 14.  By using the f/1.4 I have instead of the f/1.7
being used, we gain about .5 and get to 14.5.  Well, so much
for hand waving.  The next few weeks should tell what we can
actually do.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

TEMPERATURE CONTROL VS COOLING

I now have a fair temperature control circuit working.  It is
good to 0.02 C sigma in it's present state though I have runs
where sigma is as low as 0.0009 C.  I know how to do much
better, just a question of changing a few parts and
calibration.

It is my contention that at temperatures to which it is
possible to cool a CCD with TECs that temperature control is
as important or more important than cooling to to lowest
possible temperature without control.  For those that want to
work out the numbers, consider that about 1 C sigma is
possible if cooled flat out, and 0.02 C is easily possible
with regulation.  Another advantage of tight regulation is
that by setting the temperature just above the dew point that
one need not worry about condensation.

But it looks like I have stepped on a lot of toes, and
everyone is sending me mail justifying flat out cooling.  I
guess I will have to prove my point with pictures.

MAKING THINGS BETTER

Focus.

I think I now have a better focus.  Friday night there was
of order 50% cloud cover.  But I look at a lot of sky, and
here and there a star peeked through.  I tried a second
moment scheme for each line.  But with the clouds, it did
not make any sense at all.  In the end, I just looked at the
images where a star peeked through and judged the best focus.

Not quite so easy as looking at a star and making successive
frames with focus adjustments.  Each exposure looks at a
different star field.  Then one has to judge which is a
sharper view.  I think the math scheme will still work.  I
just have to take an appropriate set of pictures under
comparable conditions.

Binning.

I am now displaying the ccd scan as 3 ea 640 dot sweeps
across the display.  I cut off the edges.  Each sweep is then
2.5 degrees and uses 150 lines of the 480 line VGA mode.  So
I can watch 2.5 degrees in rt ass. by 30 degrees in
declination.

Sampling.

We have done nothing to improve the sampling speed.  We can
take about 2 scans a second.

Temperature Control.

The temperature control works very well.  We start the
evening by setting the temperature as low as we dare to go
without frost.  As the evening progresses, the ambient
temperature decreases, and eventually we start to heat.  All
is automatic and done by the servo.

I find that I can now use the same dark current correction
through the entire session.  Before the temperature was
controlled, after a half hour or so lines would start to
appear on the display due to pixels that drifted more or less
than their neighbors.  Note that I already subtract an
average value from sub groups of pixels.  But this does not
help if there is differential change in dark current.

I hope you all appreciate that there is about a 30 gram block
of silver under the CCD to evenly distribute the heat.  Best
possible material for temperature control.  I spare no
expense!

PICTURES

I think I have figured out how to put something up on "storm"
here so that you all can log in and get pictures.  But not
for a week or so.  What format is preferred?  I will have to
figure out how to convert pixels to the favorite format.  An
easy algorithm or a source of one would be appreciated.  Best
would be a BASIC program where I just feed it x-y and color
codes and out comes GIF or some such thing.

I think I finally identified the Andromeda galaxy.  At least
there is a bright star like thing on my scan right where the
center of M31 should be.  It is well above the noise - 20x or
so.  But no spiral arms.  I did not expect them really.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

TEST RESULTS

Every so often I get a small bare patch of sky to look
through.  I suppose this is normal weather for Chicago at
this time of year.  I even have a couple of scans left after
delete *.* on the hard disk file when I was trying to clean
up a floppy disk.

Mostly I have been trying to focus.  It is likely that the
CCD is tilted slightly with respect to the lens axis.  But I
cannot tell for sure.  Every time I try to do a series of
runs with different focus settings, clouds move in at one end
or the other of the scan.  Or they cover everything.  One
still gets star images through thin clouds.  So it is hard to
tell that the clouds are coming without a lot of running
outside.  It is also hard to tell the difference between thin
clouds causing fuzzy images at one end of the scan and one
end of the scan being out of focus.  I am only trying to do
10 minute runs.  I have not had three clear 10 minute spots
in a row for a week.

But I am not complaining.  I specifically chose this field to
retire into because it does not allow one to work all the
time!

In any case, I am now looking at a lot of stars.  I have only
very crude software working, so things may get better.  I
take a dark field which is an average of 20 line scans with
the lens cap on.  Then I subtract this from the data run.  To
remove the sky light, I also subtract a sliding average.
Because a star boosts up the sliding average the resulting
graphical presentation is such that the stars look like
little mountains with shadows caused by the sliding average.

I am sure that you all are horrified at this crude analysis,
but for the moment I am interested in optimizing the
hardware.  I am just now learning the terminology.  Before I
did not know the difference between dark and flat fields.
You will be comforted to know that I store the raw data.
In any case, with a clearly bad focus, I think I am now at
about mag 8 sensitivity by two methods:

1) Looking at a scan which contained nothing larger than mag
3, I find a star which is 950 times as large as the noise.
The noise is determined by looking for scan lines with no
obvious stars.  Then determining the sigma for the line
without the sliding average.  This to me means that a mag 3
star is 950 times the noise.  If we set 3 sigma as the
detection limit, then this places mag 3 at 300 times the
detectable limit.  This places the limit at mag 9.

2) Looking in my Sky Publishing catalog, I find a star atlas
down to mag 8 that contains 43000 stars.  Counting a 75
square degree area from the screen, I can count 80 stars
easily.  For 56000 square sky degrees, this works out to
60,000 stars or somewhat more than are in the mag 8 sky
atlas.  Don't be misled.  I am looking at near the milky way
for these scans so the count is high.

So I still have to work on focus.  After that it will be to
get a better lens system.  Before it is all over, I may have
to move in with my brother who lives in CO at 9500 feet.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

A GREAT NIGHT

Finally, last night was clear, and I got in a focusing run.

I also was able to map a 30 degree wide piece of sky for four
and one half hours.  Most of Andromeda and Perseus.  I tried
both the red and the green filter.  Some software work is
needed before I can take all three colors at once.  I am
looking between about 23 and 53 degrees of declination,
although I have not yet mapped out the edges.  The data was
taken with a 12 second exposure per pixel.  This makes the
pixels 3 minutes of arc in right ascension and 1 minute in
declination.

The real question is the sensitivity.  I found one mag 2 star
that was about 7000 times the apparent noise level.  This
would put me at about mag 11.5.  But I don't think it is that
good yet.  This was done with very simple processing.  Look
at the peak value of a star.  Look at the rms value of a
typical line and take the ratio.  There is much work to be
done on signal processing to clump together the light that is
still spread out over a number of pixels.

While the map runs were taken at my best guess for a focus, I
now have about 15 mb of data taken at 12 different focus
positions to try to plot the best value.  I am pretty sure
that the CCD is tilted with respect to the optical axis, but
only analysis will tell.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

SUCCESS SO FAR

Considering that I made the first post just trying to learn
how I might do something on August 30, I think this effort
has been wildly successful.  We have lots of pictures, and
may even reach the goal on the next pass.  When I set the mag
15 goal, I only barely knew what a magnitude was.  I had no
idea as to whether or not it was a reasonable goal.  I am
still not sure.  But I am having great fun.

MANY GREAT NIGHTS

Too much of a good thing can kill you.  There have been so
many clear nights that I am exhausted staying up to measure
Capella and my disk is full.  This is what I had hoped for.
There are lots of nice pictures taken through the various
filters.  With the luck of some bad weather, I will write
code to make color pictures.  But I wish the bad weather only
for myself. ;^)  I also need to get out and buy a tape back
up system so I can store all the data.

SUMMARY OF RESULTS

I will probably not examine this data very deeply.  At the
moment I am just trying to decide if my goal of a mag 15 sky
search can be reached with the present apparatus or some
improvement to it.

Using the data collected so far, I get of order (For those
that care, "of order" means +/- an order of magnitude.)
15,000 photo electrons for a magnitude 1 star.  This is two
orders of magnitude less than I expected (hoped for).

On the other hand, the noise is about where I computed it -or
at mag 15 within an order of magnitude.  (That is on the
assumption that a mag 1 star would produce of order 1E6 photo
exectrons.)  The noise appears to be somewhere between mag 9
and mag 13, depending on which data is examined.  The "noise"
would seem to be the system noise level, as it is the same
for a dark scan line as a "quiet (one with no stars)" sky
scan line.

The result is that with the red filter, and a few different
test stars, the noise level is about 1/4000 the signal from a
mag 2 star.  The results are not consistent, which I assume
is due to the color filter and the differences in star color.
I would put the "noise" at mag 11 +/- 2 with the present
optics and the color filtered CCD.

FOCUS

Focus turned out not to be the expected problem.  I just took
10 minute scans at a number of focus settings.  Then I
scanned through the data with a program that looked for a
star above mag 6 or so and plotted amplitude vs scan pixel.

The distributions looked just as expected.  In focus one
pixel held most of the light.  Shoulders about 20% to 30% of
the main peak.  Out of focus the peak broadened out.  Looks
like I am pretty square after all.  My belief that the chip
was tilted was due to finding many more stars in one half of
the scan.  But looking at 25 to 55 degrees means that half of
the scan (these evenings) is in the milky way, and the other
half is not.  Now I get nice sharp peaks all across the scan
line.

COLOR FILTERS

I guess they work.  I get wildly different views of the sky
through the different filters.  Any one filter gives big
differences from the star chart.  Probably the differences
are different for the different filters.  All this is
expected.  I have not yet sat down with lists of star spectra
and the results of the different scans and tried to make
sense of the data.  This is not in my direct line of
endeavor.  My goal is to find moving/changing objects.  One
could get a lot of information for the visible stars with the
present device.

I guess I need a student.

SATURATION

There is no problem with saturation.  With the present direct
computer driven electronics, I can scan the line 7.5 times a
second.  The largest signal seen so far is about 1/20 of full
scale.  So the optics can be improved 20x before mag 0 stars
start to saturate.  We have a dynamic range of 600,000 to
one with the present system for 1' of arc drift scan.

TEMPERATURE STABILITY

The decision to go for stability rather than the lowest
possible temperature was a good one.  At least in winter.
Running a degree or so above the -1 C ambient last night gave
a very steady dark current.  I remember when I started this I
was asking how stable is the dark current if the temperature
is stable.  Well it is very stable!  Will eventually produce
some numbers, but to first approximation there seem to be no
other significant terms.  So by running a little warmer than
ambient, and controlling very tightly, there are no frost or
dew problems, and I can run 5 hour exposures with no problem.
Still, my control loop is not as good as I would like.  It is
a tough problem, and I do not at the moment want to do the
(hard) work required to make it right.

WHAT TO DO NEXT?

I keep looking at the Kodak filter curve and trying to decide
how much light is lost looking through the colored filters.
I bet it is a factor of ten or so over the device without
filters.  Can anyone out there make a better estimate?  Or
does anyone know?  Eyeballing the sun's curve on the chart
would indicate that the green filter would give the largest
signals, but so far I get larger signals from the red.  It is
at least 2x the sensitivity of green filter in its region.

One possibility is to just order the KLI-2103 without the
filters.  They offered it to me that way when I ordered, but
I have no idea of what the availability might be.  But that
is another $350, so I am more tempted to use the $20 Sony
devices that I have.  They would be easier to interface, and
would get me started on the production problems.

The other thing to do is to work on the optics.  Better
optics will increase the signal and not affect the noise.

(Not quite, once we get more light, we will start seeing the
"noise" from the roughly mag 20 per sq arc sec background
stars.  This should be around mag 10 in magnitude with a
statistical noise of mag 15)

A 200 mm diameter 50 mm focal length lens would do everything
I want. Does anyone out there have one of these?  I have a
150 mm diameter, 75 mm focal length lens (Fresnel) that I
plan to try on the next round.  Anyone with a better
suggestion?

My feeling at the moment is to start building the next device
using either the f/1.4, 50 mm camera lens, or the 6" dia, 3"
focal length fresnel lens.  Actually, the plan is to design
the metal parts so I can do both.  Build this one with the
Sony chips, and give up on the color filters.

Meanwhile, try to take some color scans down to mag 10 with
the first device.  They will look nice.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

THREE COLOR EXPOSURE

The last couple of days have been spent writing software
that collects three color data.  It now sort of works and
last night with a full moon shining on my lens, I took a few
pictures.

No matter how I measure them, Rho Perseus is brighter than
Beta Perseus (Algol). If I understand how a variable star is
marked on the charts, then Algol should always be brighter.
Or do I not understand the marking system?

I also now know that I am not pointing exactly to north.
This just adds to the complication of trying to figure out
what star is coming up next.  Tonight I will kick the
mounting board and see if I can't line it up a little better.

I now have two new star atlases and could tell more about my
sensitivity except that it was not very good (about mag 6)
last night with the near full moon shining on the lens and
scattered clouds.

The three colors work as expected.  The three CCD strips are
spaced 8 pixels apart, and I could see the images appear
first in the red, then green, then blue channels as expected.
(Errr, actually, the more I think about it, the more they
appear to be in blue, green, red order.  It is always a
struggle to get the wires in the right place.  Thinking about
it too much usually *guarantees* that you get it wrong.
Eventually I will need a known red star or something to be
sure.)

I could see differences in the channels with different stars
just looking at the on line display.  There is a fair balance
between the colors.  It will take a lot of work to match up
known stars with the color channels.  With luck in a few
nights there will be no moon and a clear sky and I will get a
few three color scans.  My back up tape drive will arrive
today, so I will have a place to store the data.

A COMET TO FIND

It looks like comet Borrelly will cross my path between
November 21 and December 21.  It is supposed to be around mag
8, so I should be able to see it at my present sensitivity.
I plan to stay up to the required time and take data.  If I
can't see this one at my present sensitivity, then it will
not be too promising that I will be able to see new ones with
the improvements that I can expect to make.  There is also a
comet Machholz to see at about mag 12.  But I do not expect
to be sensitive enough to see it.  I will try.

At the moment it looks like I will be more sensitive looking
through a single filter.  This is because I have to allow
more settling time when I switch between color channels and
so can take fewer samples.  If I look at a single color for
Borrelly, what color filter should be used?  The red has more
absolute sensitivity (quantum efficiency), (RGB in ratio of
20/12/6) but looking at stars they all look about the same.

CONSTRUCTION PROJECT

I am making parts for the Mark II.  It will use the Sony chip
and the metal parts are being designed to use either the 75
mm focal length f/0.5 fresnel lens, or the 50 mm f/1.4 camera
lens that I have.  The fresnel lens will collect a lot of
light, but will resolve possibly 3' of arc.  The 75 mm focal
length will give a 20 degree field of view.  So it will have
40" pixels, and be fuzzy to 3'.  Seems to me that this is the
ideal situation for some nice computer grinding to make the
image better.  We shall see.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

IR BLOCKING

After taking three color exposures, and with a lot of help
from the CCD group, I now realize that the problem with Rho
Perseus and Beta Perseus is that Rho emits more in the IR and
since I do not block IR it looks brighter.  Thanks to this
group, I now have lots of sources for IR filters and will
order one.

THREE COLOR DATA

I now think that the three color chip was a mistake.  With a
filter it might give nice three color data, but this is not
my objective.  Why not just look wide open and take whatever
photons are available?  I am looking for moving objects.  Why
should I care what color they are?  Comments would be
appreciated.

Still, when I get a filter I will take some 3 color scans and
try to put them together.  Should make a nice mural for my
wall.

NORTH

Last night was spent with my newly acquired star book, timing
the arrival of stars.  After moving the "C" clamps a few
times I am now within 1 degree of north.  This is good enough
for now.

COMET BORRELLY

Each clear night I take a data run that includes the path of
comet Borrelly.  With a dark sky, I should just be able to
see it.  But the moon is in an unfavorable position, and
there is little hope that it will be seen in the 50+Mb files
that I am taking.  Still, they will be useful for developing
software.  There is hope that I will get a few good scans
between now and December 21 when it leaves my field of view.
Then we will see if I can write software to find it.

BACK UP TAPE

I now have a back up tape drive and am filling up tapes with
data.  No illusion that this is good data, but it is useful
to start a procedure early.

A TRIP TO CHICAGO

The adventure of the weekend was a trip to Chicago, and a
large camera store.  I came away with a 135 mm f/2.8 lens,
a star catalog, and a 10" LX-200.  It is a fun toy.  The goal
is to be able to sit in my warm room and take CCD pictures
with the LX-200 of objects found by the drift scan camera.
This means I will have to build a x-y CCD camera as it would
be against my principals to buy one.

CONSTRUCTION PROJECT

When the sky is bad, I continue working on parts for the
design that uses the Sony chip.  I am now all the more
convinced that what I need is more sensitivity.  I know about
the problem of the sky star background.  But I will continue
to attempt to make a low resolution stable system work.

The present design is a factor of 10-100 less sensitive than
I computed it should be.  This does not surprise me very much
as we are pretty close to the design on a log scale ;^).  I
think the filter loss is a big part of this.  We shall see.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

JUST TAKING DATA

Each clear night I take a data run.  I have about 600 mega
bytes so far.  The data covers roughly 27 to 57 degrees in
declination.  Mostly I have been taking 6 hour runs which
then gives me 2700 square degrees of sky and requires about
60 Mb of storage.  No effort to compress data yet.  Just take
it in real (4 byte) format. Capella give me about 14000
counts above a noise level of 1 count rms.  This through the
green filter.

I went down to the local computer store and bought a 486-66.
It uses the AMD chip and is slower for my purposes than the
486-50 that I use for taking data which uses the Intel chip.
No plug intended, but the -xx does not tell all.  I guess I
should carry a test disk with me when I buy a computer if I
care about speed.  It is usually not a big factor, but now I
have a lot of data to grind.  I now have tape back up units
on both systems and can actually move data between them.  But
there was a shuffle after I learned that IOmega is not
compatible with Colorado Data Systems drives.  If someone
knows how to make tapes interchangable don't tell me now, as
I have bought a second Colorado Data Systems unit.

Those of you that have grown up in computers recently don't
know how lucky you are.  You can just buy a basket of stuff
and plug it together and it works.  Somehow computers on the
boards that you plug into your computer all talk and figure
out how to work together.

It was not that way in 1959 when I worked on my first
interface.  Even though highly classified (the X-15 and
original astronaut programs - Redstone?), IBM would not tell
us how to interface to their IBM-650.  We had to test all the
pins on a 109 pin connector and somehow figure out what they
did.  We would have tense meetings with IBM officials and ask
questions and they would say things like "we can't answer
that."  Even pressure at the highest levels of the Navy did
not work.  Later there was a consent decree and things were
easier.  They would then answer questions, but only if we
figured out what to ask.  In those days, the digital computer
was used to run check solutions for the analog computer.

The device lives outside on the porch railing.  Last night I
had to crack the ice off the lid to open it up to take data.
Clouds moved in later but I got a few good hours.  It is
amazing that I still get a reasonable picture of the sky when
I can look up and see rows of clouds.  I just hope there is
not a sudden rain storm some evening when I am sleeping.  The
water tight box would just fill with water (or snow).  Not so
good for electronics.

The plan is to make a movie out of the repeated scans and to
look for moving objects.  I have to give credit to Benoit
Schillings for beating me to this technique.  It had always
been my plan to use this for a quick and dirty analysis, but
it looks like I did not mention it in any of my posts.

The goal of the present runs is to detect comet Borrelly.  I
know the device is not sensitive enough for the final goal.
But looking for a real comet that is fairly bright should be
good practice.  Or it will be really discouraging if I can't
find it.

CONSTRUCTION PROJECT

Work on the new design continues.  I have most of the
mechanical pieces machined.  This unit will have a larger
heat sink so I will be able to control the temperature
better.  The present unit takes about an hour to settle down
in temperature.  I also plan to use a better control circuit
and to take the time to design it properly which I did not do
in the rush to take some data.

After mulling over various fancy schemes using adder-
accumulators, I have finally settled on a simple read out
for the new unit.  It may saturate on mag 2 or above,
but that is not critical for the goal.  This weekend I should
be able to get a good jump on the wiring.  Mainly I have to
make another interconnection cable.  I hate making cables.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

STILL JUST TAKING DATA

Lots of data.  I am up over a GByte on tape.  Looks like I
am covering 24 to 54 instead of the 27 to 57 previously
stated.  But then the board it is nailed to may be warping
as it is rained on.  We shall see when I reduce more data.

I am getting pretty good recognizing the sky in this region.
But when I look up it is all different.  It does not help
the quality of the data very much that this is the holiday
season and everyone, including me, has their Christmas lights
on.  I guess that I should be pretty happy to be able to
recognize most mag 8 objects under these conditions.

CONSTRUCTION PROJECT

Mostly the next version is completed and wired up.  Just
needs a few hours testing to get it up and running.  Then
the usual software work to accommodate the changes.

DATA ANALYSIS

I have finally started looking at all that data.  When I
raise the noise threshold so that I do not get very many
noise pixels, then I get about a mag 7 star map.  I had hoped
for a little better than this.  But it depends on the IR
content of the object, I think, so some things show up much
better than others.  On my star maps are little circles with
"R" by them.  These really show up bright.  What are they?

In any case it does not bode well for finding comet Borrelly
as it is supposed to be around mag 8 where I have data.  I
have only 4 or 5 clear data runs between November 21 and
December 21 when Borrelly was in my field of view.  So I will
line up the fields, crank up the noise threshold, make a
"movie" and see what I can see.  I do not hope for much.

Still, it looks like I could go into the Mag 7 star map
business between 24 and 54 degrees of Dec.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we will limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

BORRELLY NOT SEEN

I spent the last week grinding data.  I did not find comet
Borrelly.  I only got good data on Dec 21, 24, 27 and 29.  It
was only in my field of view 21 and 24, so little chance to
spot a moving object at the very limit of my sensitivity.  I
figure this data is good to roughly Mag 7.  That is I could
spot a moving object at magnitude 7 with some statistical
level of confidence - like one sigma.

This is clearly not good enough to do anything useful.  But I
have learned a lot.  Getting only one picture a night makes
it tough to get enough data.  On the other hand, an area
device has its problems too when looking at a large field.
So the present thinking is to look at 3 or 5 sky lines in
R.A..  I should be able to make a printed circuit board that
reduces the effect of the sky curvature to the level I need.
But I have not yet done the computation.  This might give me
observations separated by 4 hours or so and I would still be
looking pretty straight up.  That would generate several
hundred megabytes an evening!  About all I could stand.

END OF A PHASE

The Mark I camera has done it's job.  I now know what is hard
and what is easy.  I guess it will soon end up in the junk
box.  There seems no sense to operate it anymore.  I have
seen enough of the stars between Andromedea and Ursa Major.
Later I hope to put up more detailed comments on the problems
listed below:

Hard

a) Doing anything with all that data
b) Figuring a good cut level for selection of a star
c) Handling the day by day hour by hour etc. sky changes.
d) Getting enough light for the desired sensitivity

Easy

a) Lining up the sky from night to night.
b) Reading the CCD at low enough noise level
c) Subtracting dark current
d) Handling the sky flat field
e) Getting a wide dynamic range
f) Sky background

SOME DATA TO LOOK AT

I have made a "movie" from the four good data days.  As soon
as I learn how, it will be put up on "Storm".  I just keyed on
Pollux and averaged the one minute data by adding 3 cells in
Dec. and 5 in R.A..  This allows getting everything from
Pollux to Theta U.Ma. on the 640 by 350 EGA screen.  This is
roughly 25 by 30 degrees of sky.

MARK II SLOWLY COMING UP

Already I have found the spot where I wired +12 directly to
ground.  One of the advantages of using low power supplies
that just barely put out the power you need is that things
seldom go up in smoke.  I keep telling the designers this at
Fermilab, but they usually do not want to worry about taking
the effort to make a tight design.  So they specify large
over capacity supplies, and then from time to time things
catch fire.  It gives me something to do on safety
committees.

This design is really simple.  The camera has only 9 chips.
It would be even simpler if I did not add all the stuff for
on line testing.  Once you put in an ADC it is worth
multiplexing in a few things so that you know what is going
on.

Yesterday I checked out the new temperature control servo.
It is a lot better than the last one, but temperature control
loops are *slow*.  So it takes a long time.  I have
sophisticated tools to work on such devices, but so far the
"seat of the pants" approach is working pretty well.  It is a
real pain to set up all the analysis to do a proper
compensation and I avoid it if possible.  This design uses
thermistors instead of the solid state device.  This allows
more sensitivity than the Mark I, but requires a calibration
so I cannot yet tell you what the temperature stability is.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we would limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

THE MARK II IS RUNNING

I would like to congratulate all the non electronics types
that have built the Cook Book Camera and made it work.
Really, the congratulations go to the Cook Book Camera
authors who have apparently designed something that non
experts could get to work.  This is not an easy task.

All this as an introduction to announce that the Mark II sky
survey camera is running.  No software yet, and clouds if
there was.  I seem to have made every possible error,
starting with wiring the +12 to ground and going down hill
from there.  The real pain was finding where I wired one
input of the differential scanner to +12.  This just blew out
the input that was supposed to go to high quality ground.
This just made the scan noisy.  Since I had no idea what the
noise level should be, it was tough to find.  Along the way
one soldered in CCD chip was sacrificed.  If I had not been
through this many times before, I would have just given up
and gone fishing.  So again congratulations to all the Cook
Bookers that did not give up and have cameras.

MEASUREMENTS

I have made a few measurements.  The temperature control
servo works great.  One test gave an rms drift of 0.0017 C.
This on the bench.  In the great out doors, I expect somewhat
less control.  At 10 C the dark current is about 100 counts
per second.  Noise at this temperature is about 5 counts with
a dynamic range of 9000 counts.  At this temperature the dark
current is changing about 10 counts per second per C.  So a
half C or so is important.  At higher temperatures, the drift
increases faster than the noise so the warmer the operating
point the more important the temperature regulation.  Not a
surprising observation.  These are just rough numbers,
don't hang me if they are not consistent.

I can cool about 20 C below ambient, though without the
current limiter it might make it to 28 or so.  No attempt was
made for large delta t.  The goal was stability.  The intent
is to operate as far below ambient as the dew point will
allow.

SOFTWARE

Now to write some software.  Current thinking is to alternate
long and short exposures to increase the dynamic range.  On
the long exposures bright stars will saturate, but the read
out noise will be minimized.  What do you think?

SOME DATA TO LOOK AT

I have made a "movie" from the four good data days.  It is
available on storm.fnal.gov.  It is in the directory pub/sky.
At the moment, the one person that tried was unable to get
the program to run.  Some bits are getting added somewhere.
Real experts can probably figure out what was changed in the
files.  I will work on this until I can read the files back
an make them work.

To prepare the data, I just keyed on Pollux and averaged the
one minute data by adding 3 cells in Dec. and 5 in R.A..
This allows getting everything from Pollux to Theta U.Ma. on
the 640 by 350 EGA screen.  This is roughly 25 by 30 degrees
of sky.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Kodak KLI - 2103 which has three rows
of 2090 pixels with Red, Green, and Blue filters in front of
each row.  The optics focuses approximately 1' of sky on each
pixel.  This gives a 30 degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover 30 degrees of sky with a single
set of optics.  To limit the storage to possible values for
an amateur, we would limit the resolution of the sky map to 1'
of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

THE MARK II IS RUNNING

Until I get some clear skies, I will polish the software.
With the current weather pattern, the software will be great.

I am trying a scheme for wide dynamic range.  Each scan line
will take 3 exposures.  Presently 4ms, 100ms, and 2562ms.
The total exposure time matches the 40" pixel of the current
optics.  Each exposure length has it's own dark line (frame).

With the long exposure, the noise is more in ADC counts but
less in terms of actual light.  The long exposure has a noise
of about 9 counts out of 9000 for a 1000 to 1 dynamic range.
The short exposure has about 3 counts of noise, with a
dynamic range of 3000.  Combined the range is 1,800,000.  For
those that care about such things, it looks like the read out
noise equals the dark current noise at 1/2 second exposure at
15 C.

Because data can now be taken as two byte integers, the
storage only increases by 50%.

This now gives three pictures of the sky taken at slightly
different times.  The plan is to use the long exposure data
as the primary sky map.  This should be just fine for
searching for moving objects like comets.  In the areas where
this map is overexposed, we can cut and paste in the the
shorter exposure data.  This would be helpful for looking for
variable stars, but here there are tough calibration
problems.  Note that we expect the fresnel lens to smear out
the data over several pixels.  We do not expect to miss
anything because all three exposures do not see an object.

So far, I have only been able to make tests using an LED
inside the camera box.  Looks good on the bench.  Indications
are that the Sony device is more sensitive than the Kodak
device with the filters.  But I have no good way to check
this.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a drift scan camera.  I.e. the camera is stationary
and measures a line across the sky.  The present
implementation uses a Sony ILX 503A.  The optics focuses
approximately 40" of sky on each pixel.  This gives a 20
degree scan across the sky.

The design should be within the capabilities of a determined
amateur.  Standard PC boards, lens specs, etc., should allow
duplication at a number of locations.  Standard software will
save total effort and allow exchange of data.

To keep down the number of devices needed to map the whole
sky, we would like to cover a wide sky angle with a single
set of optics.  To limit the storage to possible values for
an amateur, we would limit the resolution of the sky map to
of order 1' of arc.

Tom Droege

INTRODUCTION

Read "THE DESIGN GOAL" and the material following below if
this does not make sense to you.

THE MARK II SEES STARS

Last night the Mark II saw stars for the first time.  But
not until I abandoned the fresnel lens in favor of an f/1.4
Canon camera lens.

I have still not quite given up on fresnel lenses, but it is
close.

There was just barely enough time to find the focus (at least
pretty close) before the clouds moved in.

I did not see a star I could recognize before the clouds
arrived, but it is already obvious that this design is much
more sensitive.  The screen is just full of stars.

The multi-exposure scheme works great.  I plot two exposures
on different screens, and can pop back and forth between
them with a function key.  This allows sort of a blink
comparison between pictures taken with a 40/1 exposure
difference.  The low sensitivity display gives an over all
star map of the brightest stars for identification, while the
high sensitivity screen is just full of stars.

With this lens, I continue to look at about 30 degrees of
sky.

THE MARK III

Now that the Mark II is running, I have started thinking
about the Mark III.  The Mark II should be just fine for
comets if it is as sensitive as I think it is.  Now to think
about how to get more light and higher resolution.  Hopefully
the core and disk memory will keep going down so I can afford
to analyze the data.

Someone on the net just gave me a wonderful idea.  At first I
thought it was nonsense, but on reflection it solves a major
problem.  So always think about advice - sometimes it is even
good.

THE DESIGN GOAL

The goal is to design a device for an amateur whole sky
survey down to mag 15 that will detect moving objects - i.e.
comets and asteroids; and changing objects such as variable
stars and nova.

The scheme is to use a CCD designed for FAX scanning and the
like in a dr