Temperature regulation of the Cookbook CCD - an example of a low tech. approach (Addendum is at bottom of page.)

We will take a building block approach on this page, and start with simply measuring the temperature of the cold finger. I chose to use a Radio Shack thermistor (the colder it is, the higher its resistance), and this is one way to use it in a temperature measuring circuit:

In the above diagram R-therm is the thermistor. R-multi is an adjustable resistance - in my case I used a multi-position rotary switch to select various resistances (more on that later). R1 and R2 are identical in value, and in my case I made them both 220K ohms. This circuit (if I remember correctly), is called a wheatstone bridge. The voltmeter can be any type (analog, digital, new, old), but it must be able to measure both positive and negative voltages (in the old days I think they called this type of pos/neg voltmeter a galvanometer). This bridge/voltmeter setup is very sensitive, and I use it to determine if R-therm is equal to R-multi. If those two resistances are equal...the output of the voltmeter is zero. If R-therm is smaller than R-multi, then the cold finger is warmer than my 'target temperature' determined by R-multi. (Make the polarity of your voltmeter such that it shows a positive voltage in this case.) If R-therm is larger than R-multi, then the cold finger is colder than my 'target temperature' determined by R-multi. (And your voltmeter will show a negative value.) Using this approach, you can adjust your peltier voltage manually to stay close to your target temperature (a zero reading on the voltmeter), and make occasional adjustments as night air temp and water bucket temperatures gradually drift. How sensitive is this setup? With the Radio Shack thermistor, at about -40C, and using a voltmeter that measures to 0.001 volts...each change of 0.001 volts is roughly equivalent to 0.001 degrees Celsius....plenty sensitive for our application. You may want to set up your voltmeter for a high sensitivity range, and also a low sensitivity range. My voltmeter in my temp control box has a switch to select a low sensitivity range that's 1/100th as sensitive as the high sensitivity range - in other words, in the low sensitivity range, the least significant digit represents roughly 0.1 degree C, while the high sensitivity range's least significant digit represents roughly 0.001 degree C.

OK, here's the section that talks more on the values of R-multi you may want to use. With the Radio Shack thermistor the colder it is, the higher its resistance. There is a table of temp. versus resistance on the back of the package, but here are two handy formulae you can plug directly into a spreadsheet:

1. =188.4*(2^((-40-A2)/12.5)) this formula takes the value of spreadsheet cell A2 (as a temperature, in celsius), and outputs the resistance of the thermistor in Kohms.

2. =-12.5*(LOG(D2/188.4)/LOG(2))-40 this formula takes the value of spreadsheet cell D2 (as a resistance, in Kohms), and outputs the temperature in celsius.

Don't be put off by these formulae. Here's the basic idea behind them: at -40C the thermistor has a resistance of 188.4 Kohms. The resistance of the thermistor doubles (or halves) for every 12.5 deg C that it gets colder (or warmer). These formulae are pretty accurate for temps between -20C and -50C...close enough for this application.

What values did I use for R-multi, and what temperatures do they represent?

Kohms---Temp (deg C)

60------- -19.4

80------- -24.6

100------ -28.6

120------ -31.9

150------ -35.9

188.4---- -40.0

220------ -42.8

250------ -45.1

300----- -48.4

350----- -51.2

400----- -53.6

500----- -57.6

The value of resistances you choose for your camera depend on your cooling system's performance. (Mine is a two-stage cooler, and I've reached -59C with an ice water bath...your mileage may vary.)

At this point, you now have a way to measure the camera's temperature, and if that's all you want to do, then you can stop here. If you want to 'close the loop' and make an automatic temperature regulator, then read on....

The above circuit takes the output of the wheatstone bridge we spoke about earlier, and connects it to an op-amp. R-therm, R-multi, R1, and R2 are the same as discussed before. R-feedback is used to set the gain (amplification factor) of this circuit. Make R-feedback larger, and the gain gets larger...and vice versa. This circuit is a proportional controller...it merely amplifies the 'error signal' by a constant. We could stop here, and use the output of this op-amp to vary the peltier's voltage supply, but I recommend adding one more stage to this circuit...the integrating stage. (As a side note, if we remove R-feedback, gain gets very, very high...essentially the op-amp will behave as a comparator and take one of two output values: Vmax or Vmin, with no intermediate values. This temperature controller circuit is a 'bang-bang' design that runs full-on, or full off, switching back and forth between the two states. An example of this design is at: http://www.astroguy.com/tempreg.html... which also shows how to connect the output of your temperature controller to the 'stock' Cookbook' peltier power supply...keep that in mind at the end of this article.)

The figure above shows the integrating stage. The component that does the integrating is the capacitor, C-integ. In my case I used a 22uF capacitor. R3 in my case is 220Kohms, and I use a 500Kohm potentiometer for R-feedback. By changing R-feedback, I can adjust the overall gain of the system in the field. (I tend to adjust gain to the point where the system temperature starts to oscillate, then I back the gain off by lowering the resistance a bit.) I also use a 500K ohm potentiometer for R-adj. This is an offset adjust that sets the voltage output of the controller when the error signal is zero...in my case I set the controller for approx. 5 volts, tucked that potentiometer inside the box, and make no more adjustments to it in the field.

NEW Addendum: (Oct 2002) This temperature regulator was so successful, that it changed the way I took calibration frames. Because repeatability from night to night was so consistent, at the beginning of an observing 'season' I would take the following calibration frames on a cloudy night: 199 x bias, 199 x flat dark, 25 x dark (these darks were approx. 15 minutes long...that's about six hours of CCD dark current being measured for a very high quality calibration frame). All of these frames would be median combined to make master frames. This meant that during clear nights I only had to shoot flat frames, which took only a few minutes. Thanks to use of a flat light box, I could take flats any time I wanted to. BOTTOM LINE: Every available second of dark time was available for imaging...no time had to be spent making calibration frames.

Don't take my word for it. If your camera is temperature regulated (commercial or home made), try this for yourself and you may find that your calibration frames remain consistent from night to night (except for flats). If that's the case, then you can make your imaging time more productive. Also, do some QC of your master calibration frames. Make your master bias from median and average combinations...then subtract one from the other. Are the differences very small? Does the same apply to your flat-darks? With your darks, use a cross histogram plot to compare each individual dark to the master (median combine) dark. Is there a linear relationship? The occasional cosmic ray hit will cause some pixels to deviate, but all the rest of the pixels should cross plot in a linear fashion. Also on this cross plot...do the maximum/hottest pixel values for each individual frame fall within 1-2% of the max/hottest pixel in the master frame? If yes, then you have good long term stability of your temperature regulation. If not, how many of your individual frames are outside this 1-2% criteria? If only one or two, then discard these oddballs and make a new master dark from the remaining good frames. If there is lots of scatter in many of your dark frames, then stability/repeatability of temperature regulation is suspect.

All comments are welcome, and will make this page more informative for Cookbook CCD camera builders!

Tom Krajci

email: t-k-r-a-j-c-i-@-s-a-n-.-o-s-d-.-m-i-l (remove the dashes)

Last updated 20 Oct 02