Very stable temperature control

[HendriXML]

I build a apparatus to do a temperature controlled measurement on a simple diode:
https://www.eevblog.com/forum/testgear/using-a-awg-and-a-scope-do-a-low-voltage-level-characterization-of-a-1n4005/msg2514567/#msg2514567

The measurement takes more than 20 min. So I needed a way to stabilize a certain temperature for at least that period. The accuracy of the temperature was less of a concern.

Using this setup the temperature will fluctuate less than 0.1 deg. This is done by:

• Using a relatively large thermal mass (heatsink and +/- 400 ml water)
• Good thermal capacity (water)
• Good thermal conduction (heatsink and water)
• Good thermal isolation (Isolating can and covering the setup with foam)
• Controlled heating (2 x TIP120) symmetrically near the diode under test
• Measuring near the diode under test (LM335Z)

The best way to use this setup is to start with boiling water and let it then cool down. The regulation will than eventually kick in. To keep the temperature stabile it needs about 2 W, but the setup can deliver 16 W if needed. Without any heating, the temperature would drop with some 1 deg per 3 minutes. So very slowly.
The regulation oscillates only a little (which is inevitable) but this is in the range of 1 hz. (So a minor climb/drop every second)

[HendriXML]

Here's the schematic. The working is as follows:

• One part is responsible for dividing the voltage, thus power evenly across the two TIP 120's (Q1/Q2)
• One part amplifies sensed current
• One part amplifies the difference between the measured temperature (LM335Z) and a reference voltage (BT1), I use my signal generator for the reference voltage
• One part compares that difference against the delivered current/power, and "requests" more power is the temperature difference becomes higher. Zenerdiode D2 limits this power request to 16W

[dzseki]

If you'd add an integrating part to your control loop, then you could get rid off the small ripple as well.

[blackdog]

Hi HendriXML, 

You cant say that you build a good themperature controler and also say the loop is not stable! 

The regulation oscillates only a little (which is inevitable) Ooooh...
I think you have to work on the loop control, it must be stable, don't fool yourself.

Normaly you need only a reference voltage, a bridge circuit and a power stage and one opamp to maak a good controler.(and ofcource a tuned loop)
Search for oven controlers and jim williams from LT, and you can also search for oven controlers and blackdog 

It never have come to my mind to use a plastic isolated cup, i wil buy one en wil test its performance.

Kind regards,
Bram

[HendriXML]

If you'd add an integrating part to your control loop, then you could get rid off the small ripple as well.

Thanks for your response. In a controlled (no external temperature fluctuations) environment (which this setup is) that would even be better, I agree. But when there're would be fluctuations, it would respond a bit slower to changes. I choose for mainly for heat buffering. If without any heating (0 W) the temp drops 1 deg / 180 sec. Then this would be less than 0.01 deg per second. However I can see the power delivery oscillate between 2.4W and 3W.  So my rough guess is that the temperature fluctuation is even 5x smaller.
I keep it this way because one of the driving questions was whether this method would be stable enough.
It seems to work quite well. If one influences the temperature (by blowing at the sensor for example) it does have a little overshoot -as one expects- but still very little.

[HendriXML]

Hi HendriXML, 

You cant say that you build a good themperature controler and also say the loop is not stable! 

The regulation oscillates only a little (which is inevitable) Ooooh...
I think you have to work on the loop control, it must be stable, don't fool yourself.

Normaly you need only a reference voltage, a bridge circuit and a power stage and one opamp to maak a good controler.(and ofcource a tuned loop)
Search for oven controlers and jim williams from LT, and you can also search for oven controlers and blackdog 

It never have come to my mind to use a plastic isolated cup, i wil buy one en wil test its performance.

Kind regards,
Bram

Thanks for your response!
Heat traveling takes time, so whenever there's a distances between heater and sensor. There're will be some feedback time. One can eliminate that by using "prediction", thus using an average (integrator/electronic buffer) but when the temperature does change, it will be slower/more off.
This setup is used in a stable environment, so I could make the loop entirely stable (slow). But that would also give it a less interesting behavior. The circuit is designed to give 50W of power / delta T (deg C).
So in a way it even needs some compensation at the reference voltage. Which is thus dependent on the average needed power to sustain a temperature. (Heat leakage)
If I where to improve this setup, I would target the total thermal resistance between the TIP120's and the sensor.
The used silicone pads are rubbish, and the parts of the heatsinks where the TIP120 are mounted on are somewhat hollow. The heatsink was sold as high tech back in the days, but I think the used materials is rubbish regarding to thermal conductivity. It was never a good CPU heatsink  .
This means while controlling extra heat will build up, which will flow eventually and give a bit of overshoot.
Using a better heatsink, more transistors and better pads, better mounted sensor, would be a nice experiment.

[dzseki]

If you'd add an integrating part to your control loop, then you could get rid off the small ripple as well.

Thanks for your response. In a controlled (no external temperature fluctuations) environment (which this setup is) that would even be better, I agree. But when there're would be fluctuations, it would respond a bit slower to changes. I choose for mainly for heat buffering. If without any heating (0 W) the temp drops 1 deg / 180 sec. Then this would be less than 0.01 deg per second. However I can see the power delivery oscillate between 2.4W and 3W.  So my rough guess is that the temperature fluctuation is even 5x smaller.
I keep it this way because one of the driving questions was whether this method would be stable enough.
It seems to work quite well. If one influences the temperature (by blowing at the sensor for example) it does have a little overshoot -as one expects- but still very little.

So you could look for a conventional PID control loop, with proper tunig you could have the fast response with the stable output at the same time. One can even realize a PID control with a single opamp, you'd not even have to overcomplicate your design.
But of course it is up to you, and if you have already satisfactory results, all good.

[HendriXML]

If you'd add an integrating part to your control loop, then you could get rid off the small ripple as well.

Thanks for your response. In a controlled (no external temperature fluctuations) environment (which this setup is) that would even be better, I agree. But when there're would be fluctuations, it would respond a bit slower to changes. I choose for mainly for heat buffering. If without any heating (0 W) the temp drops 1 deg / 180 sec. Then this would be less than 0.01 deg per second. However I can see the power delivery oscillate between 2.4W and 3W.  So my rough guess is that the temperature fluctuation is even 5x smaller.
I keep it this way because one of the driving questions was whether this method would be stable enough.
It seems to work quite well. If one influences the temperature (by blowing at the sensor for example) it does have a little overshoot -as one expects- but still very little.

So you could look for a conventional PID control loop, with proper tunig you could have the fast response with the stable output at the same time. One can even realize a PID control with a single opamp, you'd not even have to overcomplicate your design.
But of course it is up to you, and if you have already satisfactory results, all good.

I used what I had laying around. The TIP120 as heaters for example. Because they cannot dissipate a lot of power, it had to have a current limiter anyway  .
I"m pleased with the results. I'm a beginner, so the more components/complexity, the more there's to learn.

[HendriXML]

Search for oven controlers and jim williams from LT, and you can also search for oven controlers and blackdog 

Found your project, with the inner and outer oven, very nice to read. Targetting 40 deg, so you could also have use transistors  as heaters. 

[HendriXML]

I always try to keep track of where the values and underlying information of the schema comes from. For this purpose I use a scripting tool.

I then try to separate the values I've chosen, the values from datasheets, the values that where measured and the values that are calculated. The calculations are done running the script so it is easy to recalculate stuff.

But because I know what information is provided or consumed by what I also know what the effect of design changes are. Using this modular approach much larger project can be kept under control.

To know how the values are used, one needs to check the script. But that could also have been reported.

The script fails when assertions haven't been met.


Highly stable temperature control
Version 1.0.34

Choosen values
provides
  voltVCC                  : 24,00 V
  degHeaterMax             : 80,00 °C
  degHeaterMin             : 10,00 °C
  ohmR3                    : 100 mΩ
  ohmR4                    : 100 mΩ
  ohmR6                    : 214 Ω
  ohmR7                    : 5,1 kΩ
  ohmR11                   : 1 kΩ
  ohmR12                   : 1 kΩ
  ohmRV1                   : 10 kΩ
  wkControlPower           : 50,00 W/K

Datasheet values TIP120
provides
  kwThermalResistance      : 1,92 K/W
  degJunctionMax           : 150,00 °C
  facHFE                   : 1000,00
  voltBEOnMax              : 2,50 V

Emperical values max power
consumes
  Choosen values
  degHeaterMax             : 80,00 °C
  Datasheet values TIP120
  kwThermalResistance      : 1,92 K/W
  degJunctionMax           : 150,00 °C
intermediate
  degJunctionSafeValue     : 140,00 °C
  degMeasuredHeatSink      : 80,00 °C
  degMeasuredTransistor    : 122,00 °C
  wattMeasuredWatts        : 17,00 W
  wattSafeValue            : 18,75 W
provides
  wattPowerQ1Q2            : 17,00 W

Max current determination
consumes
  Choosen values
  voltVCC                  : 24,00 V
  ohmR3                    : 100,00 mΩ
  ohmR4                    : 100,00 mΩ
  Emperical values max power
  wattPowerQ1Q2            : 17,00 W
intermediate
  ohmR3sR4                 : 200,00 mΩ
provides
  ampMaxCurrent            : 712,56 mA
  wattMaxPowerQ1Q2         : 17,00 W
  voltMaxSenseVoltage      : 142,51 mV

Emperical values 1N4728A
provides
  ampCurrent               : 5,00 mA
  voltReverseVoltage       : 2,20 V

Sense voltage amplification
consumes
  Choosen values
  ohmR6                    : 214,00 Ω
  Max current determination
  voltMaxSenseVoltage      : 142,51 mV
  wattMaxPowerQ1Q2         : 17,00 W
  Emperical values 1N4728A
  voltReverseVoltage       : 2,20 V
intermediate
  facAmplification         : 15,43
  ampCalcR6R7              : 665,95 μA
  ohmCalcR7                : 3,30 kΩ
provides
  vwPowerPerVolt           : 7,73 W/V
  ohmR7                    : 3,3 kΩ

Datasheet values LM335Z
provides
  vkSensitivity            : 10,00 mV/°C
  ampForwardCurrentNominal : 1,00 mA

Temperature deviation amplification
consumes
  Choosen values
  wkControlPower           : 50,00 W/K
  ohmR11                   : 1,00 kΩ
  ohmR12                   : 1,00 kΩ
  Sense voltage amplification
  vwPowerPerVolt           : 7,73 W/V
  Datasheet values LM335Z
  vkSensitivity            : 10,00 mV/K
intermediate
  vkControl                : 6,47 V/K
  facAmplification         : 646,76
  ohmCalcR13               : 646,76 kΩ
  ohmCalcR14               : 646,76 kΩ
provides
  ohmR8                    : 18 kΩ
  ohmR9                    : 1 kΩ
  ohmR13                   : 680 kΩ
  ohmR14                   : 680 kΩ

D1 series resistor (R10)
consumes
  Choosen values
  voltVCC                  : 24,00 V
  degHeaterMax             : 80,00 °C
  ohmRV1                   : 10,00 kΩ
  Datasheet values LM335Z
  vkSensitivity            : 10,00 mV/K
  ampForwardCurrentNominal : 1,00 mA
intermediate
  maxSensorVoltage         : 3,53 V
  ampRV1                   : 353,15 μA
  ampRV1pD1                : 1,35 mA
  voltCalcR10              : 20,47 V
  ohmCalcR10               : 15,13 kΩ
provides
  ohmR10                   : 16 kΩ

Datasheet values LM358
provides
  voltPositiveOutToRailVoltage1mA: 1,40 V
  voltPositiveOutToRailVoltage5mA: 1,50 V

D2 series resistor (R16)
consumes
  Choosen values
  voltVCC                  : 24,00 V
  Emperical values 1N4728A
  ampCurrent               : 5,00 mA
  voltReverseVoltage       : 2,20 V
  Datasheet values LM358
  voltPositiveOutToRailVoltage5mA: 1,50 V
intermediate
  voltR16                  : 20,30 V
  ohmCalcR16               : 4,06 kΩ
provides
  ohmR16                   : 4,3 kΩ

Q1 (R18) and Q2 (R19) base resistor calculation
consumes
  Choosen values
  voltVCC                  : 24,00 V
  ohmR3                    : 100,00 mΩ
  Datasheet values TIP120
  facHFE                   : 1000,00
  voltBEOnMax              : 2,50 V
  Datasheet values LM358
  voltPositiveOutToRailVoltage1mA: 1,40 V
  Max current determination
  ampMaxCurrent            : 712,56 mA
intermediate
  ampBaseQ1_2              : 712,56 μA
  voltCalcR18              : 8,03 V
  voltCalcR19              : 20,10 V
  ohmCalcR18               : 11,27 kΩ
  ohmCalcR19               : 28,21 kΩ
provides
  ohmR18                   : 11 kΩ
  ohmR19                   : 27 kΩ

[Rerouter]

Its a bit of a weird circuit, just thrown it into a simulator,

I would say simplify back to a simple output constant current load style circuit for the output, and some basic differential circuits for the regulation.

So something like this, It heats up quickly but backs off the heating significantly when the heatsink starts getting too much hotter than the measurement point, and vice versa turns back up the heat a little when it gets too far below it.

The 100K under the second op amp from the left adjusts how aggressive this back off strategy is, the lower the value the more careful it is to not dump too much heat into the heatsink,

And the 100K under the first op amp from the left controls how agressive the heating rate is for the measurement point vs the set temperature, lower is slower and more careful. the higher the value the hotter the heatsink will generally get on startup.

Final note, your heatsink temperature would ideally be something like direct to the heat source,

[HendriXML]

Maybe the current amplification is less conventional, I found it in the LM358 data sheet.

But otherwise I don't think I went to crazy.

I like that it can be tuned with scientific meaningful parameters. The power is linear with the error temp. My guess that the setup would lose less than 5W of heat, resulted in matching values right away.

It works how it supposed to, with not much deviation from calculated values so I'm confident that it is not a "lucky shot circuit". Using the current limiting zenerdiode is a bit sensitive to tolerances. But if it was off too much I would than upped or lowered the supply voltage to get a 17 W max output.

On the first iteration I used two TL071 to be able to do some input offset. But those opamps cannot have inputs near ground (V-), so it didn't work.

[joeqsmith]

I assume you are within +/- 0.1 Celsius.   I would normally want a well insulated container with some decent airflow inside.   It looks like you actually mount the parts outside in open air and try to heat the mass to a temp.   Beach towels work well. 

When I have matched transistors and such, I have used cardboard boxes with foam and a resistor for a heater.   I am not sure how tight the current version of this is but I suspect it's under +/- 0.005 degrees C.
https://www.eevblog.com/forum/projects/poor-mans-thermal-chamber-meat-packing-box-ii/msg2454591/#msg2454591

This chamber was just for playing with some ICs. 
https://www.eevblog.com/forum/projects/waveforms-in-a-74ls04-ring-oscillator/msg646512/#msg646512

The internals on this project are controlled to about a tenth, which dominates the error.     
https://www.eevblog.com/forum/testgear/playing-with-delta-sigma-converter-poor-mans-meter/msg2041111/#msg2041111

[HendriXML]

I assume you are within +/- 0.1 Celsius.   I would normally want a well insulated container with some decent airflow inside.   It looks like you actually mount the parts outside in open air and try to heat the mass to a temp.   Beach towels work well.

I don't have the MM to do high resolution measurements. But looking how the power fluctuates I think the temperature  fluctuates around 0.002 deg. This mainly because of the thermal capacity/heat loss ratio. Water is cheap and does its work well. Much better than alu.
The active setup is with a 10 cm foam on top. Not the best, but insulating enough. The most stable setup is the one that doesn't lose any heat and needs almost no heater.

[Damianos]

In the past, I played with a circuit very similar to the one at page 16 here: http://www.ti.com/general/docs/lit/getliterature.tsp?baseLiteratureNumber=snoa748&fileType=pdf&keyMatch=AN-460&tisearch=Search-EN-Everything
I did also a variation, using a second sensor, to make a differential controller for a setting "above the environment temperature" (for testing heatsinks).
Except of the "original" parts (LM35, LM10, LM395), I tried also other types with similar results.
Unfortunately I have not access to the lab yet, so I can't give more information.

[max_torque]

This sort of thing is really best done open loop tbh. The total heat flux drives the temperature, so just put a known and fixed amount of heat in, and wait for it to stabilise!

Do it in a large room, where the thermal mass of the room and air is large, and there is nothing to drive any oscillation...

[HendriXML]

This sort of thing is really best done open loop tbh. The total heat flux drives the temperature, so just put a known and fixed amount of heat in, and wait for it to stabilise!

Do it in a large room, where the thermal mass of the room and air is large, and there is nothing to drive any oscillation...

I understand your reasoning, but the thermal mass and conductivity of air is extremely low. (That's why we use heatsinks.)

[joeqsmith]

My room temperature varies way to much to be useful.   Simple beach towels over the device is a huge improvement.

I don't have the MM to do high resolution measurements. But looking how the power fluctuates I think the temperature  fluctuates around 0.002 deg. This mainly because of the thermal capacity/heat loss ratio. Water is cheap and does its work well. Much better than alu.
The active setup is with a 10 cm foam on top. Not the best, but insulating enough. The most stable setup is the one that doesn't lose any heat and needs almost no heater.

It may be worth investing in a decent used DMM.  If you are looking at characteristics of diodes,  I would have thought you would have been all set with this.    In the case of the meat packing box, I want to be able to run at different temperatures.    Using the PC to control and record the data as much as possible just makes things easier.   

If you like watching paint dry and grass grow,  this is about 10 minutes of data collected real time as the box is starting to settle at 23.0C with the stacked Peltiers. 
https://youtu.be/j9YcKQ-2uU8

[HendriXML]

In the past, I played with a circuit very similar to the one at page 16 here: http://www.ti.com/general/docs/lit/getliterature.tsp?baseLiteratureNumber=snoa748&fileType=pdf&keyMatch=AN-460&tisearch=Search-EN-Everything
I did also a variation, using a second sensor, to make a differential controller for a setting "above the environment temperature" (for testing heatsinks).
Except of the "original" parts (LM35, LM10, LM395), I tried also other types with similar results.
Unfortunately I have not access to the lab yet, so I can't give more information.

From what I've understood is that K-type sensors are in fact differential temperature sensors. With a hot and cold junction. It should than be possible to use just a pair of those wires to do differential measurements, by reading only one voltage. This might eliminate some errors. Don't know whether this is used in practice.

[HendriXML]

My room temperature varies way to much to be useful.   Simple beach towels over the device is a huge improvement.

I don't have the MM to do high resolution measurements. But looking how the power fluctuates I think the temperature  fluctuates around 0.002 deg. This mainly because of the thermal capacity/heat loss ratio. Water is cheap and does its work well. Much better than alu.
The active setup is with a 10 cm foam on top. Not the best, but insulating enough. The most stable setup is the one that doesn't lose any heat and needs almost no heater.

It may be worth investing in a decent used DMM.  If you are looking at characteristics of diodes,  I would have thought you would have been all set with this.    In the case of the meat packing box, I want to be able to run at different temperatures.    Using the PC to control and record the data as much as possible just makes things easier.   

If you like watching paint dry and grass grow,  this is about 10 minutes of data collected real time as the box is starting to settle at 23.0C with the stacked Peltiers. 
https://youtu.be/j9YcKQ-2uU8

I've gone to so much effort precise measurements with a scope and awg that I indeed earned my right to buy me a bench MM. I will explain this right in the upcoming years to my spouse.  To her all these boxes on the bench are probably all the same and some of them are quite recent. 

Until than I've to be creative.

It is also very nice to see how others build comparable "contraptions".

[HendriXML]

If you'd add an integrating part to your control loop, then you could get rid off the small ripple as well.

I think I will add a capacitor, and implement this. Using a much lower power / error temp ratio. The situation is well suited for it and I know now already how it behaves without. It will probably add some ringing, but that will than slowly die out. (I guess)

It is also nice to lose the needed offset.

[HendriXML]

Here's is then the new schematic.

I'll try to analyse how to get a good value for C3.

As said, in this setup it is most easy to start with a too high temperature. This ensures everything is in balance, when the sensor measures the target temperature. Before that no power is delivered, so no influences of the heater.
Until that time the opamp U2A outputs 0V. However I know the stable power is around 2.5W, thus 0.32V at the U2A output. The circuit is setup to deliver 7.73 W/V. (Limited at 17W, zener voltage 2.2V)
This means that C3 will need drop some of its voltage (0.32V) , this will be done while the amplification is in "slow mode" (undergoing a strong, time related C3 feedback). In slow mode a temp difference of 0.5 grad would be needed to get the required difference of 2.5W power. (That would be the very max temperature it would drop, if C3 was infinite)
If the time it takes to have C3 drop half of its voltage is much shorter than the time it takes to loose half of 0.5 grad, I think the system will stabilize just fine.

I know the temp drops roughly, 1 deg per 180 sec. So taking 30 sec for the 0.25 grad is on the safe side.
This would be at 43 RC. When using a 680K resistor, this will lead to a 63uF capacitor. So with a 10 uF there's nothing to worry about. Thats is the largest bipolar capacitor I've lying around.
Whether it is too little? It might if the power is high compared to the thermal mass, but with the turned down reactivity? We'll see.
Maybe it is a good experiment to try no amplification feedback at all, and see how that oscillates.

(I know the calculations are very rough, discarding some meanwhile changes in voltages)

[joeqsmith]

I havn't had an eBay account for at least 10 years now but thought I would have a look on eBay for cheap used bench meters.   Now I'm glad I closed my account as I would have been in a world of trouble buying some this old vintage gear.   May have had to join that TEA thread to recover.       There are some decent meters out there, and some very expensive junk. 

[David Hess]

I have hacked together small temperature controllers a few times and always got by with a bridge circuit using a diode as the temperature sensor (1) and an operational amplifier implementing just the integration part of a PID configuration.  Frequency compensation using the normal methods is only time consuming because it takes a couple minutes to run each test.

Temperature regulation of better than 0.1C is easily achievable this way but actual accuracy will depend on physical construction and how it limits frequency response.

(1) One of the various PTAT IC temperature sensors will work just as well as a diode.

[max_torque]

My room temperature varies way to much to be useful. 

Does it? OK, if you have a room that gets lots of sun, or leave the windows open, but the OP asked for "20min" stability, and over that time i suspect the actually air temperature doesn't change that much.  Find a shady corner of your house, put the device there, and i see no reason to chase complex PID controllers etc, just put on a couple of watts of heating, leave it on till the temperature reaches stability and you're done!