Temperature control?

[Callsign_Vega]

How are you guys keeping your chilled setups at desired temperatures? Obviously some of you want the coldest temperature possible so you let the compressor run continuously.

Is there some sort of off the shelf way to control temperature? I believe my window A/C unit uses a basic resistance type thermometer. It's designed to only go to 60 deg F, so obviously it won't work for sub-zero temps. Besides trial and error super insulating the thermometer, how could I control the compressor for desired temps based off of the reservoir liquid temps?

[shadow6491]

Disable the AC temperature control in some way. Then use a temperature controller with its sensor in the liquid to simply turn the AC unit on or off.

[Stewie007]

Look at a basic temp control system... Johnson Controls makes good controls with a decent variety of differentials. You just disconnect and remove the A/C controller.

[Callsign_Vega]

My A/C unit is new and it's turned on/off with soft touch buttons on a PCB so it has some relays. It also has a large capacitor. The stock temp sensor is also connected to that same PCB. I don't think the A/C unit will start and stop by engaging and disengaging the power supply by an external source with a temp probe like this:



I'd imagine a A/C unit like this measures the electrical resistance in the temp probe line. Do you think I could alter the resistance so that the stock probe could still work but at much colder temps? Open to suggestions as I am a pretty big newb at temperature control.

[Red M]

I'd imagine a A/C unit like this measures the electrical resistance in the temp probe line. Do you think I could alter the resistance so that the stock probe could still work but at much colder temps?

This should work, if you use a trim pot in series (usually) with the thermistor you could adjust the temp to suit your needs, it would just be a matter if finding the right sized pot for your thermistor. I would make sure the probe only has 2 wires coming from it as 3+ wires would suggest a digital probe.

A settable thermostat is the easiest option if you cant mod your AC controller, you might even be able to find one at your local hardware store

Another way if you are comfortable with basic programming/electronics or are willing to learn is something like a picaxe chip, they are very easy to program micro-controllers and the web site has good documentation, they would allow you to have complete control in setting things like on/off temps and minimum cycle time + more depending on what you want to do and how good your programming is.

[Callsign_Vega]

This should work, if you use a trim pot in series (usually) with the thermistor you could adjust the temp to suit your needs, it would just be a matter if finding the right sized pot for your thermistor. I would make sure the probe only has 2 wires coming from it as 3+ wires would suggest a digital probe.

A settable thermostat is the easiest option if you cant mod your AC controller, you might even be able to find one at your local hardware store

Another way if you are comfortable with basic programming/electronics or are willing to learn is something like a picaxe chip, they are very easy to program micro-controllers and the web site has good documentation, they would allow you to have complete control in setting things like on/off temps and minimum cycle time + more depending on what you want to do and how good your programming is.

It's just a two-wire thermistor. Now as I understand it there are two types of thermistor. One in which the resistance in the circuit is increased as temperature decreases and one in which resistance is decreased as temperature decreases.

Using my multi-meter, I've found that I have a negative temperature coefficient thermistor, which would appear to be the harder to modify. As temperature decreases, resistance increases. Room temperature (20 C) is at about 7.98 kilohm. At 0 C in my refrigerator it reads 10.7 kilohm. I put it in my freezer which is at -17 C and it reads 15.10 kilohm.

If I added any more resistance with trim pots or the like the controller would think the A/C is even colder and turn it off sooner, defeating the purpose of what I want to do. It seems this thermistor reading on the PCB goes right into the main-board chip. Can anyone think of a way I can decrease the resistance of the thermistor circuit so that the A/C unit thinks it's warmer and hence let the compressor run at a lower temp? Or is buying a new thermistor with resistance value appropriate for my temperature ranges my only option while keeping the stock A/C controller?

EDIT: while the unit is actually on, the resistance on the temp probe leads is 0 ohms at all temperatures. The voltage changes quickly based off of heat. At 20 C the voltage was read at about 2.5v and at about 35c the voltage was read as 1.8v. Could I use some kind of voltage regulator in conjunction with the temp probe to control this?

[[XC] gomeler]

My 30 seconds of thought yielded this solution. Your goal is to either take full control of the AC unit or trick the current electrical wizardry into running under your command. Going with the trickery perhaps the following would work?

Use a normal PID to operate a SPDT relay that would either enable a resistor in parallel to drop the resistance of the probe or a resistor in series to increase the resistance. Then with the probe at ambient air temps you can trick it into thinking it is either warmer or colder than the temp you set on the A/C control unit.

The other option is to do away with the fancy electronics and just drop in a mechanical relay that'll handle the power requirements of the unit and connect this up to your PID and put the relay in line with the common line of the compressor. Instant on/off.

This really took about 2 minutes to write and I was thinking while typing so some of it might not make sense but the idea should be correct.

[TJ TRICHEESE]

Resistance cannot be measured when a circuit is powered. just put a pot in parallel with the thermistor and adjust it for the resistance required using the equation: (1/R1)+(1/R2)=(1/Rt).

Also i would do a calibration curve for the temperature sensor so you know what resistance you want.

Changing the resistance will cause the reading on the AC to be wrong.

[shadow6491]

Are you sure its a thermistor and not a thermocouple?

It would be much easier to trick the original electronics than modify them as you really dont know what they are and use something like the device pictured.

[Red M]

As a few people have mentioned you do need to confirm the components that are in use, I would suggest it probably uses a thermistor as they are usually favored at the temp range you're looking at but by checking any markings/packaging/reaction to temp you should be able to confirm, if it isn't a thermistor it would most likely be a thermocouple (as shadow pointed out) which will generate a different voltage at different temps rather then creating a varying resistance.

If you need to raise the resistance a small to medium trim pot in series would be suitable where as if you need to lower the resistance a large trim pot in parallel would do the trick, if you're putting one in parallel some people like to use a resistor in series with the trim pot too so you cant take the resistance right down to zero as a safety measure. Again as shadow pointed out if it has a temp readout this will no longer be accurate but if your just looking for a switching solution this shouldn't be an issue.

Again I'd recommend having a look at the picaxe chips, for $10 you could have a fairly advanced controller an example of a simple program to hold the chiller between -10c to -15c on a 08M chip with a DS18B20 temp probe on pin 6 and a relay on pin 7 would be;

#picaxe 08M

let b0 = 0

low pin7

main:
readtemp 6, b0                              ;Read temp to byte 0 (750ms) NB:you can speed this up with other components/methods
if b0 < 138 then gosub chilleron            ;If temp above -10 turn chiller on
elseif b0 > 143 then gosub chilleroff       ;If temp below -15 turn chiller off
endif
goto main

chilleron:
high pin7                                   ;Turn on relay on pin 7
return

chilleroff:
low pin7                                    ;Turn off relay on pin 7
wait 60
wait 60                                     ;Two 60sec waits to prevent the compressor from turning back on for atleast 2min
return

I didn't run this through the compiler so sorry for any mistakes but it gives you an idea of how simple it is, you are welcome to use this code too if you'd like it just continuously monitors the temp and keeps it at whatever range you want (i use 138 and 143 because the first bit signifies positive or negative temp ie -10c is stored as 128+10=138 and -15 is 128+15=143).
If you want fancy you can do that too (with a bit more effort), I've run graphic LCDs with these using I2C and RS232, or for a simpler solution you could have a 7-segment display as an output, you could even have an output to your MB power pins and have the comp turn on once a certain temperature has been achieved with very little extra effort.

[Callsign_Vega]

How can I for sure tell if its a thermistor or a thermo-couple? It shows a change in resistance based off of temperature while using a multi-meter and also shows a change in voltage based off of temperature when the unit is running. It has no markings on it.

[[XC] gomeler]

You could also use a PID controller with a relay like indicated in the 4th post and just stick the temperature probe against the compressor casing. This way the system controller always sees high temperatures and runs the compressor 100% of the time.

[Red M]

How can I for sure tell if its a thermistor or a thermo-couple? It shows a change in resistance based off of temperature while using a multi-meter and also shows a change in voltage based off of temperature when the unit is running. It has no markings on it.

When the controllers off (and preferably probe unplugged/detached) attach a DMM or such to the probe, at different temps, if its a thermistor the resistance will change in the range of tens to thousands of ohms depending on the model and it will show no voltage change, if its a thermocouple the resistance should be stable but there will be a voltage change in the range of uV/mV usually so you'd need a decent scope to detect it.

Judging by your earlier post I'd guess you've got a thermistor as you get a resistance change when its powered off (cant measure resistance with the circuit on). The voltage change you see when its on will be a reference voltage fed through the thermistor to create a voltage drop relative to the probes resistance.

[NotSoCoolJ]

You need to add resistance in parallel to the NTC thermistor. Test the thermistor with a DMM at a few different temps and work up a quick temp curve. Then you can add a pot in parallel and adjust to suit your needs. You just need to know what resistance range to buy.

[TJ TRICHEESE]

You need to add resistance in parallel to the NTC thermistor. Test the thermistor with a DMM at a few different temps and work up a quick temp curve. Then you can add a pot in parallel and adjust to suit your needs. You just need to know what resistance range to buy.

That has been suggested before in this thread ^

[Callsign_Vega]

You need to add resistance in parallel to the NTC thermistor. Test the thermistor with a DMM at a few different temps and work up a quick temp curve. Then you can add a pot in parallel and adjust to suit your needs. You just need to know what resistance range to buy.

I'll post up some quick numbers of the ohm values at different temperatures and maybe you guys can help me pick the right trim pot. Would a potentiometer work or would the voltage be too high for one of those?

[Utnorris]

This is what I used in mine:

http://www.coleparmer.com/Catalog/pr...sp?sku=9352010

Basically, I pulled out the A/C controller and put this in. It keeps temps within 3 degrees Celsius. That way I can stay above the dew point and not worry about condensation. Most of the time I ran it around 10c. It would cycle about every 8 minutes for 3 minutes cooling a QX9650 @ 4.2Ghz, mb blocks and two GTX280's.

[Callsign_Vega]

This is what I used in mine:

http://www.coleparmer.com/Catalog/pr...sp?sku=9352010

Basically, I pulled out the A/C controller and put this in. It keeps temps within 3 degrees Celsius. That way I can stay above the dew point and not worry about condensation. Most of the time I ran it around 10c. It would cycle about every 8 minutes for 3 minutes cooling a QX9650 @ 4.2Ghz, mb blocks and two GTX280's.

Damn thats looks nice. Do you know if they have a higher AMP rated version? I am pretty sure my 15k BTU unit is much higher than 3/4HP compressor.

[Rubycon]

Cycling a compressor is bad. But everyone knows that, right?

There are other ways to control temperature but they are not very friendly to the wallet of the hobbyist. A compressor with a cycloconverter drive (variable speed), swash plate with unloading valves, or even using TXVs and hot gas bypass setups. Idea here is to keep the compressor motor running as long as possible, avoiding more than four start/stop cycles per sixty minutes.

Remember small PSC compressors have relatively weak starting torque even with a start kit. Cycle them too rapidly against unequalized pressure and you wind up a locked rotor and the Klixxon opening up. (meanwhile your loop temperatures keep going up!)

I suppose if your cooled water quantity is large enough and your exchanger is insulated well you would have long runs and decent off times between calls for cooling. If there is little reserve short cycling could be a real concern. In any case a compressor lockout relay should be used! Klixxons can weld shut allowing the compressor to stay energized with a locked rotor. Most pull enough amps to open the branch circuit breaker, however.

Work / play safe!

[Utnorris]

By default, don't AC's cycle and are built for that? Granted, if you are leaving it on 24/7 it will cycle way more than if it was still be used as an AC. As long as it has time in between cycles to settle, wouldn't it be ok?

[Callsign_Vega]

Cycling a compressor is bad. But everyone knows that, right?

There are other ways to control temperature but they are not very friendly to the wallet of the hobbyist. A compressor with a cycloconverter drive (variable speed), swash plate with unloading valves, or even using TXVs and hot gas bypass setups. Idea here is to keep the compressor motor running as long as possible, avoiding more than four start/stop cycles per sixty minutes.

Remember small PSC compressors have relatively weak starting torque even with a start kit. Cycle them too rapidly against unequalized pressure and you wind up a locked rotor and the Klixxon opening up. (meanwhile your loop temperatures keep going up!)

I suppose if your cooled water quantity is large enough and your exchanger is insulated well you would have long runs and decent off times between calls for cooling. If there is little reserve short cycling could be a real concern. In any case a compressor lockout relay should be used! Klixxons can weld shut allowing the compressor to stay energized with a locked rotor. Most pull enough amps to open the branch circuit breaker, however.

Work / play safe!

Hopefully the cycle times won't be called up too frequently. I will be running around 7 gallons of coolant. The liquid would get too cold when the computer is not under load (gaming) with the compressor running full time.

[Rubycon]

Yes in theory as long as your cycle time is long enough to let pressures equalize it will be fine. However run times should be longer for better equipment life.

[Callsign_Vega]

Temperature curve for thermistor: +20 C it reads 7.98 kilohm. At 0 C it is 10.7 kilohm. -17 C it reads 15.10 kilohm.

For adding a trim pot in parallel, do you think a 5k or 10k pot would work? Remember, I need to decrease the resistance slightly with a pot in parallel. When the thermistor is at -17C @ 15 kilohm, I actually need the circuit to be showing around 8 kilohm. Also, are there any worries with wattage on the circuit and shorting it out with use of a trim pot? If I ruin the A/C unit PCB controller I will be in bad shape!

Hopefully you guys can help with this as I am not very experienced with small circuits.

Is something like this what I am looking for: http://www.radioshack.com/product/in...ductId=2062356

And which of the three posts do I connect to the two wires on the thermistor?

Or would the smartest thing be to just cut off and replace the thermistor with an ohm value corresponding to the temperature I want with something like this: http://www.ussensor.com/rt%20charts/PS102J2.htm That is showing 8k ohm at -17C, and my default thermistor shows 8k ohm at +20C.

[Red M]

I would try a 25 turn 50+ohm trim pot, as this would give you a good range of temperatures while still giving you a fine adjustment at the values your wanting.
Something like this: http://jaycar.co.nz/productView.asp?...T&SUBCATID=803

Using the rule for resistors in parallel;

1 / R = 1 / R1 + 1 / R2 + ......

where R = the resistance we want/the chiller setting (ie 8 Kohm @ 20c) and R1 = the resistance of the probe at running temp (ie 15 Kohm @ -17c) and R2 = the resistance of our trim pot, we can now find the minimum resistor size needed.

1 / 8 = 1 / 15 + 1 / R2
R2 = 1 / (1 / 8 - 1 / 15)
= 17.14 Kohm

if you want flexibility you want a larger resistor as the more resistance you add the closer you will get to the original value but the more sensitive the adjustments become as a much larger range has to be moved through in the same number of turns of the trimmer (ie 25 turns 0-XX ohms).

Some examples of probe resistances when the trimmer is set to max and the controllers trying to hold 20c (8 Kohm);

R1 = 1 / (1 / 8 - 1 / R2)

25 Kohm: R1 = 11.76 Kohm
50 Kohm: R1 = 9.52 Kohm
100 Kohm: R1 = 8.7 Kohm

so you can see that if you had a 100 Kohm trim pot set to max with the controller set to 20c your temp would be about 15c, where as with the same setting and a 50 Kohm trim pot the max temp you could adjust to is about 9c, and with the 25 Kohm about -4c.

Something some people like to do (I don't bother TBH) is use a resistor in series with their pot (which are both in parallel with the probe) to act as an adjustment safety eg put a 15 Kohm resistor in series with a 50 Kohm trim pot that way your resistance range becomes 15-65Kohm (17-9Kohm on probe with cont @ 20c) giving a temp range of about -24c - +12c while preventing anyone accordantly adjusting the pot to 0 ohm shorting the controller.

Replacing the thermistor is a viable option however you are limited to the range of temperatures settable by the controller also that range might be affected but the resistance/temperature gradient of the thermistors response.

[n00b 0f l337]

You can easily be a "start delay timer" and add it to the main power loop though. That will prevent cycling by putting in a hard timer between off's to the next on.
In connection with a PID/thermocouple and an SSR or Contactor, you can have it all done for about $60.