Author: Tom Droege Date: 960117 Revision: #0 960117 Key Words: instrumentation, CCD
First some calibration data.
Kodak Lists the following pixel use:
1-10 Empty 11-14 Non Imaging (masked to prevent seeing light) 15-782 Imaging 783-794 Non Imaging 795-796 Empty
It would appear from looking at the data that 9-11 and 13-15 are dark pixels. Pixel 12 must have some special meaning as it seems to have some charge in it. But I note that the rms value of 9-11, 13-15 is larger than the pixels exposed to the sky. This makes no sense at all to me. I will write Kodak about it. Please let me know if you understand what Kodak is doing here. Pixel 16 would appear to be partially masked. It is possible that my count is one off from the Kodak label. It is hard to interpret the Kodak clock diagrams. So possibly add 1 to the Kodak table to get the tass pixel number.
Using tass pixels 14 and 15, I get a mean value of -28663 and a sigma of 117 ADU (ADC counts). From dark runs, I get a mean value consistent with the masked pixels but with an rms value of 4-5 ADU. Taking pixels 17-20 from line 19 and 20 of Orion3.dat, I get a mean value of -26203 and an rms value of 12 ADU. It is curious that the non imaging pixels have a larger rms value than the ones exposed to the sky. These sky pixels are from the edge of the CCD with the least sensitivity.
Preliminary tests after the gain change indicate a dark pixel value of -28600 ADU and an rms value of 4-5 ADU at -18 C. From Orion3.dat, bright stars saturate around -8100 ADU. This gives a full scale range of 18700 ADU. Using Kodak's 10 uv per electron and the gain of the system of 54 uv per ADU gives 5.4 e- per ADU. Using the 18700 full scale counts determines a full well capacity of 100,980 electrons. We will do a better calibration later.
Some of you will wonder why the large gain change when a much smaller one would have used more of the full scale range of the ADC. First, I was confused by the large signals under high light conditions. These give approximately 4x the signal above because the horizontal shift register has 2x the capacity of the vertical shift register, and the output stage has another 2x capacity. So one can see 4x the signals. This is to accommodate 2x binning in the horizontal and vertical. The present gain setting almost allows 2x horizontal and vertical binning without saturating the ADC. We may want to try it as it is a simple software change.
To summarize:
Looking at the Orion3.dat file, I find a nice star at:
RA = 5h32m49.6s Dec = -0 42 47
It is listed as 7.96V in my catalog.
Having none of your powerful tools, and not having time to learn to use them yet if I did, I went after this star with a pencil and clip board. I just wrote down a 6 by 9 array of all the pixels around this star.
First we just use the peak value in the 6x9 box of pixels. This is - 11729 ADU. Using the pixels far from the star, I get a mean value of -25700 ADU and a sigma of 42 counts. The 42 is probably a high number. It probably has some stars in it. There are much better schemes. Thus the peak pixel is 15771 ADU above the mean value away from the star. Divide by 42 gives the peak 375x above the noise. This is mag 6.44 over the noise for a 7.96 mag star or a 1 sigma noise at mag 14.4.
Next we try a sliding 4x4 pixel box. When we do this, the first thing noticed is that we could do better on N-S alignment and on time. The peak value is actually skewed over two diagonal pixels. The rotation mechanism on this unit is not quite right and it is hard to adjust. We will do better on the next one. Summing the 16 pixels gives a count 77657 ADU above the background. This is mag 8.17 over the noise for a one sigma noise at 16.1. Please tell me if this is not a valid scheme.
I don't know what sort of schemes are used in detail to dig signal out of the noise, but I hope they are better than the ones above. I do not really expect to be able to scale dim stars from bright stars. I don't know what else to do for now. There are also other problems. For example, I read that charge transfer efficiency may not be the same with a few electrons in the well as with a large number. This is a problem for the experts. But it is not much of a problem for us since all that sky background insures that we never have a near empty well. So we only have to worry about the statistical noise due to all those electrons.
All in all, the above seems to indicate that we should be able to make pretty good measurements down to mag 13 at 3 sigma or so above noise. In this biz, when someone says they have a sensitivity of mag 20 what do they mean?? We should be able to measure over 50,000 stars over the year, and should find a hundred or so variable stars of various types in this group. I am encouraged that we have a machine that might find some interesting stuff.
Note that just in Orionx.dat and Equatorx.dat we should find a few new variable stars. At least according to my advisor, what do I know? It should be fun to find them. Remember, I am mostly too busy building stuff to do more than evaluate the data. So the data is fair game. If you find something you can feel free to report it.