The CCDs in the Mark IV telescope are cooled by thermoelectric coolers. The performance depends on the efficiency of the coolers, the chilled water provided to the system, and what this does inside the camera head. This note will cover some tests that have been made which point to a practical mode of operation.
For some time I have been trying to make the following test. Only recently have enough things been working to allow it.
This data is for the V camera. The I camera is similar, but always a little noisier. I don't know why.
Command Temperature Short Short Long Long Voltage V Camera ADC Sigma ADC Sigma 4 10.2 -23901 139 -21761 613 3 5.4 -24593 79 -23197 421 2 2.9 -24866 66 -24007 301 1 -1.5 -25174 39 -24665 203 0 -5.8 -25332 25 -25035 129 -1 -9.8 -25412 18 -25286 87 -2 -13.3 -25457 12 -25336 59 -3 -17.2 -25478 9 -25376 38 -4 -21.0 -25465 7 -25421 26 -5 -23.7 -25467 7 -25440 18 -6 -26.4 -25456 5 -25430 13 -7 -29.2 -25430 5 -25413 9
Command Voltage is the ADC output that drives the servo. Just arbitrary voltage to temperature.
Temperature V Camera is the temperature at the thermometer in the block below the CCD. It is probably reading colder than the actual CCD temperature. It is in a 1" square block of aluminum, 0.4" thick about 3 mm behind the face of the CCD.
Short ADC is the ADC reading for a 15 second dark exposure.
Short Sigma is the sigma of the short dark exposure in ADU.
Long ADC is the ADC reading for a 150 second dark exposure.
Long Sigma is the sigma of the long exposure in ADU.
Note that an offset is built into the electronics to provide about -25500 ADC counts for a full dark signal. The ADC is scaled for -32768 to +32767 counts. The scale factor is roughly 2.4 counts per ADU, and the full well capacity for the CCD442A is about 80,000 electrons. Thus -25500 counts is near zero, and 8000 counts is near full scale. These scale factors were picked since the noise level from the sky greatly exceeds the expected noise level from the ADC. This scale allows studying the performance beyond saturation.
The first thing that is noticed is the lack of a square relation between the sigma of the dark current accumulation and the probable value of the dark current.
For example, the 10.2 C short exposure has a value of -23901. We can assume that the zero dark current value is close to the -29.2 C value. At least it should be a factor of 2^8 less. (from 8 decreases of 5 C - the current doubles for a 5 C change) This is close enough to zero to test the statistics. If we take -25430 as a near zero dark current value, then we have 1529 counts of dark current at 10.2 C. Measurements indicate a scale factor of 2.4 electrons per count. This value closely checks with the 80,000 e- full well and the observed saturation values. This dark reading then represents 3670 electrons. This should give a sigma of 60 electrons or 25 when translated to ADU. The measured sigma is much higher, 139 ADU counts. What is wrong?
The problem is what I have been calling the Michael Richmond effect, since he first noticed it in our data. Consultation with experts at CCD-World determined that this is a well known effect. If we plot a histogram of large number of pixels, we get a multi- peaked distribution. This is caused by hot sites that generate dark current. Some pixels have no hot sites, some one, some two, some three ... I can usually see 7 or 8 such peaks on cameras operated at about 10 C. Each site generates the same amount of current within a relatively narrow range, else the statistics would smear out and the peaks could not be seen.
These hot sites tend to disappear as the temperature is decreased. At 10 C 7-9 peaks can be counted. Probably more if more careful statistics would be taken. At -17 C only one or two could be counted. At -21 C they could not be detected.
It would be nice if we could look at long exposures at low temperatures and see that the sigma is proportional to the SQRT of the count. Alas, even for the 100 second exposure, the dark current is down in the noise and no conclusions can be drawn.
For tass work in Chicago, sky brightness is typically 50 ADU. Byt the time -20 C is reached, dark current represents a 10% increase over the sky. It is probably not worth heroic effort to reduce the temperature in the suburbs. At a good location, it might be worth trying to operate at -30 C.
With 15 C water, the TEC can achieve a CCD temperatue of -30 C. This is a delta t of 45 C. Usually we can achieve a delta t of 50 C with 25 C water. The TEC are more efficient at higher temperatures. Nature conspires against us. Lowering the chilled water temperature does not help in proportion. Running at O C might get to -35 or -40 C at the CCD (this has not been measured), not the hoped for -50 C. Running cold water produces condensation problems everywhere, particularly on the camera window.
The current plan is to run with the cooling water temperatue slightly above the dew point, whatever that is for the evening's operation. This will involve "guessing" somehow the highest dew point for the evening's run.
We now have a cheap chiller that works reliably. We took a low cost de-humidifier and pulled off it's cover. We then bent the coils down to floor level, and immersed them in a bucket of cooling water. A computer controlled SCR switch just turns the de-humidifier on and off to control the water temperature. There are thermometer channels and bit control switches available in the tass hardware for this. This works well. We can easily control the chilled water temperature to 1 C. The limit to holding closer control is the overshoot caused by there being a small amount of cooling water in the system and lots of refrigerant. So cooling continues after the power is turned off. If we increase the amount of cooling water, the control is tighter.
The CCD control loop in the camera is capable of much tighter control than this if the chilled water does not change too fast. This argues for a relative large volume of coolant in an insulated tank. Later we will measure the tracking rate of the inner control loop to determine how much cooling water should be provided. With a typical de-humidifier, this might be 20 gallons if very tight control is desired. It is probably not necessary. Operating at -20 C or so the dark current is small and we will probably not be able to see the effect of the chiller switching on and off.
Since the first operation we have been plagued by ice crystals which form on the detector surface. The first plan was to have a container of "Drierite" connected to the camera volume. Ice would appear on the CCD surface at about -10 C. This with a CCD that was sealed at the factory. There is moisture inside the sealed CCD. The next try was to hold a vacuum on the camera heads. Bake out under vacuum would help, but only for a few days, then the ice crystals would re-appear. We then tried putting a second TEC in the enclosure powered so it's temperature would be lower than the CCD temperature. This was to act as a getter. Baking the camera head with the getter helped to get to about -15 C.
After consultation with CCD-World, the design was changed to a positive pressure system. A "fish tank" air pump forces air through a container of Drierite to generate a source of dry air. This air is circulated through the two camera heads, then through a trap, then to a small bubbler to prevent back flow of moist air. The trap prevents the heads from drawing the bubbler oil into the camera head volume. I recently demonstrated the need for the trap by connecting a head that I had been testing under vacuum into the system. The quick disconnect tubing connectors I am using hold a vacuum in the head when disconnected. I connected the head into the test system and heard a slurping sound. The oil from the bubbler was sucked into the trap. So the trap saved the head from a bath of oil.
The positive pressure scheme above has allowed operation without ice crystals. When we turn off the air circulation for several days, then turn on the fish tank pump air circulator and the TEC to full cold, we still get a few ice crystals. If instead we allow several hours of dry air circulation, no ice crystals form. This would seem to be an acceptable mode of operation.
During summer, we experienced condensation on the camera window. We have added heaters on the face of the camera to make sure that it is significantly warmer than the rest of the camera head. Two 10 ohm resistors are connected in series to the +12 volt TEC power supply. This provides 7.2 watts of heat to the window area. An earlier try with 1.4 watts still allowed condensation. The 7.2 watts seem to be enough with near 100% relative humidity (tested while raining) to prevent condensation. A new shell has been designed to allow a little more spacing between the CCD and the filter/window.