Temperature control accuracy

[BetterSense]

Technical question:

I'm making a 5-gallon reservoir of tempered water, heated with a hot-water-heater slement. It won't be used for brewing but the application is extremely similar to the hot liquor tank...a source of stable tempered water for heat exchange at 28C.

My question is, how stable can you really control a vessel of hot water? Is 0.1C out of the question? Or is +/- 1C the best I can hope for?

 

[runningweird]

With a pid setup you can be at .01 c

 

[alien]

It's hard to calibrate thermometers closer than about 0.5°C. Also, there will be temperature variations within the vessel, unless you stir it or recirculate the water.

Do you really need 0.1°C accuracy though?

 

[kenh]

Technical question:

I'm making a 5-gallon reservoir of tempered water, heated with a hot-water-heater slement. It won't be used for brewing but the application is extremely similar to the hot liquor tank...a source of stable tempered water for heat exchange at 28C.

My question is, how stable can you really control a vessel of hot water? Is 0.1C out of the question? Or is +/- 1C the best I can hope for?

Typical PID accuracy is +/- 3% of full scale, then for a PID to work you must have some fluctuation. Temperature has to drop some for it to turn on & rise some for it to turn off. Then you have a delay from when the element turns on & your sensor "sees" the change.

The answer is "how much money do you want to spend?"

 

[BetterSense]

Money is no object.

I should not have used the word "accuracy". Whether the water is 28.0 or 27.0 or 29.0 doesn't actually matter much. What I care about is that the temperature of the output water/mash is stable, repeatable and doesn't drift in time. I want the minimum hysteresis possible. I don't want it to cool from 28.0 down to 27.5 before the control system reacts then overshoot to 28.5.

I already calculated that the "mash" is going to cool down the water at a rate of 1 degree per minute. My heating element should heat it up by 1 degree per 20 seconds or so. I think I can program a thermostatic control to kick on at 27.9C and turn off at 28.1C. The temperature should form a sawtooth right around 28C (nominal) plus or minus a tenth. I don't expect any overshoot because the heat capacity of the heating element is small compared with the reservoir. I will have multiple submersible pumps inside causing as much mixing as possible. Reasonable goals?

 

[kenh]

Money is no object.

I should not have used the word "accuracy". Whether the water is 28.0 or 27.0 or 29.0 doesn't actually matter much. What I care about is that the temperature of the output water/mash is stable, repeatable and doesn't drift in time. I want the minimum hysteresis possible. I don't want it to cool from 28.0 down to 27.5 before the control system reacts then overshoot to 28.5.

I already calculated that the "mash" is going to cool down the water at a rate of 1 degree per minute. My heating element should heat it up by 1 degree per 20 seconds or so. I think I can program a thermostatic control to kick on at 27.9C and turn off at 28.1C. The temperature should form a sawtooth right around 28C (nominal) plus or minus a tenth. I don't expect any overshoot because the heat capacity of the heating element is small compared with the reservoir. I will have multiple submersible pumps inside causing as much mixing as possible. Reasonable goals?

Maybe if you set everything up right. I can say that with my RIMS temperature stability improved when I moved the sensor from the input of the tube to the output of the tube. With the sensor on the input the temperature would over shoot a little then drop a little.

 

[dyqik]

Money is no object.

I should not have used the word "accuracy". Whether the water is 28.0 or 27.0 or 29.0 doesn't actually matter much. What I care about is that the temperature of the output water/mash is stable, repeatable and doesn't drift in time. I want the minimum hysteresis possible. I don't want it to cool from 28.0 down to 27.5 before the control system reacts then overshoot to 28.5.

I already calculated that the "mash" is going to cool down the water at a rate of 1 degree per minute. My heating element should heat it up by 1 degree per 20 seconds or so. I think I can program a thermostatic control to kick on at 27.9C and turn off at 28.1C. The temperature should form a sawtooth right around 28C (nominal) plus or minus a tenth. I don't expect any overshoot because the heat capacity of the heating element is small compared with the reservoir. I will have multiple submersible pumps inside causing as much mixing as possible. Reasonable goals?

A properly tuned fast PID controller will happily operate to keep the sensor within a 0.2F range, assuming that it has the resolution and accuracy to measure that. You don't want to be using the cheap options for this job. Controlling overshoot in PID is what the I and D parameters are for. However, the response time to changes in the set point will be slower - to maintain that tight a range, the response time has to be slowed. The bigger issue is that a pot of water will have to be rapidly stirred to maintain the whole volume within that 0.2F range. You might also need multiple sensors to average several points around the vessel, and more accurate sensors and readout on the controller - sensor drift and resolution will also be an issue.

Personally, I use a Lakeshore Model 336 for this at work - I'm maintaining 10mK temperature accuracy there (at 2.6K). You have to pick the appropriate sensors for the job though.