Re: The Joys of Cooling

[Tom Droege <tdroege2@earthlink.net>]

Just about everything below has a simple explanation, I think.

>   Last night, by mistake, I forgot to turn on the pump which forces
>water through the tubing.  I ran the cameras with the TEC units turned
>on to an intermediate voltage ("3", according to Tom's control
>program).  The ambient air temperature was around 15-20 C, and
>the cameras cooled to about -5 degrees.

The real clue here is to watch the CCD current.  If it is less than the 
maximum that is possible - typically 1.5 amps with a good TEC, less with a 
broken one.

If it is less than hard on - you can tell if the current first goes up to 
some limit and then backs off - then the servo is regulating.  The 
temperature you see is what you have asked for.

>   Then I realized that the pump wasn't on.  So, I turned on the
>pump, and watched as:
>
>         - the V camera dipped by about 1 degree C for five minutes,
>                 then returned to its previous temperature
>
>         - the I camera dipped by about 3 degrees C for maybe 20 minutes,
>                 the gradually returned to its previous temperature

This is exactly what you expect to see if the temperature loop is 
regulating.  Suddenly the water side is colder since hot stagnant water is 
now being replaced by cool circulating water.  The loop cannot instantly 
correct, since it very slowly changes the TEC current.  So the temperature 
drops.  Then the servo backs off the current and the eventually (In an hour 
or so) you get to the command temperature.  I wish this were all faster, 
but temperature servos with thermal masses are tough to move fast.  If you 
want to guarantee good temperature regulation, then the trick is to have a 
lot of water in the cooling system.  Then it's temperature cannot change 
very fast.  I can explain more if anyone cares.

>   Later that night, I tried to drop the temperature of the fluid.
>The garbage can contained about 15 gallons of water and antifreeze.
>I placed 14 pounds of ice into the water, while the fluid was
>circulating, and watched the CCD camera temperatures.  Over
>the next 20 minutes or so, the CCD temperatures gradually dropped
>by about 1 degree C (both of them), and then they both rose
>back to their previous levels.  The ice hadn't fully melted
>(or even close to it) when the CCD temperatures were back to their
>pre-ice levels.

Yep, you can't keep a good system down.  You have to command a lower 
temperature to get it if it is possible (the system can drive enough 
current through the TEC to get to the commanded temparature.)

Now if you had circulated anti-freeze cooled to below -5 C then the CCDs 
would have tracked the temperature down once the cooling liquid got below 
-5 C.  The reason, you can only turn the TEC cooling to zero, it will not 
heat.  As it was, as you made the coolant colder, the servo just backed off 
the TEC current to hold the temperature you had commanded.  (With a TEC and 
with a  bi-directional drive you can both heat and cool.  If I had it to do 
over, I would drive it bi-directionally and get a better control.  This one 
can only cool. )

>   My guess is that the thermal contact between the block through
>which the water flows, and the finger running to the CCDs (or the
>finger-CCD connection), isn't as strong as it ought to be.
>If I'd been smart, I would have monitored the "Water Temperature"
>reported by the Mark IV, which I _think_ is the temperature of the
>block through which the coolant flows.  All I know is that
>the coolant
>
>       - started the night at 23 degrees C,
>       - was at 21 degrees C about 90 minutes after adding ice
>       - was at 19 degrees C about 130 minutes after adding ice,
>                  when I gave up and went home
>
>   It may be that the flow of coolant is too small to cool the
>block and finger effectively.  Because I'm using lots and lots
>of narrow tubing, only a trickle moves through the block --
>I measured it last January at something like 4-7 gallons per
>hour.

OK, I will save this and read it as a humility check when ever I write some 
stupid code and you (Michael) are kind to me about it.

Some day when the sun is up, you can learn about the temperature control 
servo by running TemPlt and watching all the nice curves.  OK, you will 
have to run DOS, but you don't need the Windows interface unless you are 
taking data.

Start with the dacs set at some number - where they are now is fine 
(3?).  Turn on with everything at ambient, and watch the plots as the servo 
regulates.   Particularly watch the currents.  See the maximum that they 
reach, and watch to see if they eventually get to some value less than maximum.

Now you can change the value where they are set by hitting LdADAC.  This 
will ask for a DAC number, and a value.  14 is the I DAC, 15 is the V 
dack.  Let's say you select 15 and set it to 2.  This will cause the TEC to 
regulate to a cooler temperature.  It is instructive to watch what happens 
when you do this.  You will notice that I am running the servo 
"underdamped" to get as fast a response as possible.  I am hoping that 
someone will do the work of running a calibration curve on the control 
voltage vs temperature reached.  Then fix up the code to make it so you can 
command the temperatue.

You can command lower temperatures - I normally run at -3 which gets me to 
-15 C or so.  I have stuck to this number all summer and now I can just 
make it.  The cooling water is set to not go below the dew point, so on 
warm humid nights there can be a problem.

>   Tom, can you tell me exactly where the thermistors for the
>"Water Temperature" and "CCD Temperature" are connected?

The CCD thermister is in a block between the TEC and the CCD.  Since there 
is not a big heat load on the CCD, it is close to the back plate 
temperature of the CCD chip.

The water temperature thermister is in a block between the V and Ic 
cameras.  So it measures too hot for one and too cold for the other.  As 
long as there is enough flow, there will not be much difference.

>   Oh, one more thing: the tracking is fine when the telescope
>points to the East or South.  Once it moves about 45 minutes
>West of South, it starts to stick, so that 60-second exposures
>show short trails, often with "dots" (meaning that the telescope
>tracked well for a while, then stuck for a moment, then tracked
>well again afterwards).  Perhaps this is due to dirt on the
>portion of the threaded rod which touches the drive mechanism
>when the telescope points West.  I plan to clean the rod, and
>_maybe_ to lubricate it, before the next night's run.

This sort of problem is usually caused by something hitting something, or a 
piece of tubing catching on something, or a bad RA drive chamois.  You take 
your choice.  Nothing to do but sit there and watch what is going on.  Note 
that there is an "exercise" mode for the RA drive which moves it back and 
forth.  By increasing the time on for each mode you can get it to cover the 
full range.  It takes a while, but it is much faster than just driving at 
RA speed.

This is all good news for me.  I is just possible that your camera TECs are 
OK.  The clue is to do a plot of TEC current vs temperature done with (if 
possible) constant cooling water temperature.  Note that I would move the 
command voltage only about once an hour.  One volt steps are fine for a 
start.  I think a volt change in the DAC will change the control point 
about 3C (just a guess).  So a good start is to change from +3 to -3 in one 
volt steps and record V and Ic currents and temperatures after an hour or 
so of operation at each point.  Quit it the current hits a limit and stays 
there.

OK, enough.  I will be happy to answer questions.  Obviously this needs to 
be discussed more.

Tom Droege