Peltier Temperature Controller

[Alex QW]

Hi All,

I need to design a Peltier temperature controller in pic32mx microcontroller base design. A microcontroller as a PID controller the problem is I need to control the Peltier module by its current instead of voltage due to controlling by its voltage will be low efficiency and generate a lot of noise will affect my other systems.

The only way I think is using the programmer's current Source (PCS) ic but I could not find any PCS that could be controlled until 4A. guys, do you have any idea about it? 

I need to design the same as this product (https://www.meerstetter.ch/products/tec-controllers/tec-1091)


From
Alex.

[Psi]

The problem is I need to control the Peltier module by its current instead of voltage due to controlling by its voltage will be low efficiency and generate a lot of noise will affect my other systems.

Can you explain what you mean by this a bit more.

The typical way to control something like this is voltage PWM, and that isn't low efficiency, it's pretty high.
If you want to avoid switching noise that you'd get from PWM or DCDC then you normally have to go with a linear supply, and that means you waste all the heat you don't use so it's very in-efficient

Also, let us know if this is a on-off project for you, or something you want to make 500 to sell.  There are many shortcuts you can do for making your own hobby one-off project that you cant do for volume production. Usually because you need to keep the cost super low.

[Alex QW]

Hi PSI,

This project I am doing for my uni project.

Yes I cannot use PWM to control my Peltier plate on-off the voltage source will generate noise


From,
Alex

[Psi]

One option might be a high current opamp. You can get ones that will do 10A. but they cost like $30 each i think.
Some have a current sense system build in and output a voltage proportion to the load current.
So you could build a circuit using the current sense as feedback to your MCU to control the power into the Peltier.

But it would of course be cheaper to use a normal cheap opamp and just drive an external transistor and add your own current sense system.  Just more time/effort. but maybe you'd get better marks for it.

You can also get opamps with a digital input or digital gain, basically the same as a normal opamp but you can control them digitally from a MCU, instead of having to make the MCU output a voltage. So that might be worth a look.
But probably easier to just use a MCU with a DAC and control a normal opamp.

[Alex QW]

Hi PSI,

Do you mind providing a clear image by using a block diagram?



From
Alex

[Psi]

You could probably drop the opamp and just drive a high current transistor directly from a DAC of the MCU.
Just have to figure out the right resistor values to vary output current over the DAC range.

Then you might want to add a current sense system (series resistor and current sense amp IC). and feed the output into the ADC of the MCU so you can measure the current.
You could even simplify that further and make it a low-side current sense system so you can use the MCU ADC to measure the voltage across the low side resistor. Probably need a MCU that has an ADC with some internal gain settings. 20x etc. since you want to keep the sense resistor value low to avoid it getting hot.

You probably don't need to measure the current, but its useful if you want to measure the exact watts going into the peltier. And you would also measure the voltage across the peltier on another ADC channel etc..

Then you just need your PID code to interface with the system above to control the energy into the Peltier, and the temp from your temp sensor.

[Alex QW]

HI PSI,


But my MCU internal ADC only 10bits the resolution is not high enough



From
Alex.

[Psi]

But my MCU internal ADC only 10bits the resolution is not high enough

You can do a lot with 10 bits, why do you think its not high enough?  and you could add an external ADC / DAC if you wanted more bits.

[jbb]

How about the Analog Devices LT8722?
- Edit: 15 V Absolute Max Vin
- So more like 12V or 13 V supply
- 4A max output
- Built in high resolution DAC

https://www.analog.com/en/lt8722

[mikerj]

A microcontroller as a PID controller the problem is I need to control the Peltier module by its current instead of voltage due to controlling by its voltage will be low efficiency and generate a lot of noise will affect my other systems.

Whether a TEC is controlled by voltage or current has zero effect on efficiency and noise.  The way in which the voltage or current is generated will contribute to noise and efficiency.  If low noise is a priority then a linear controller (either voltage or current) will offer the best performance, with the trade-off being efficiency.  Likewise a switching regulator (again either current or voltage feedback) can give high efficiency but the potential of more noise, though this can be minimised by suitable filtering and good layout.

If if need more than the 10 bits that the internal DAC can provide, then use an external SPI DAC of whatever resolution you need.  The LT8722 mentioned above is a good suggestion, this uses an interesting Linear Technology (now Analog Devices) architecture which combines a linear push/pull output and a switching regulator to provide bipolar drive.  Unlike some other manufacturers offering the switching output is always on the same leg of the TEC so you can add however much filtering is needed to achieve your noise requirements, whilst maintaining high efficiency.  Both maximum current and maximum voltage can be limited to your requirements.

The typical way to control something like this is voltage PWM, and that isn't low efficiency, it's pretty high.

Raw PWM is one of the worst and lowest efficiency methods of controlling a TEC.  Filtered PWM giving low ripple (i.e. effectively a buck regulator) is fine.

[Siwastaja]

I need to control the Peltier module by its current instead of voltage due to controlling by its voltage will be low efficiency and generate a lot of noise will affect my other systems.

This makes zero sense. You are mixing some things up. Turning Peltier on/off between full power / no power is bad for efficiency (and possibly EMC, as you say), and I would add also a risk for mechanical lifetime. But controlling the voltage is fine because Peltiers are highly resistive (unlike batteries, LEDs or motors which require current control). Controlling the current is fine, too. Do what is easier. Usually adjustable voltage supply is simplest. This can be unregulated buck (aka filtered PWM).

[Psi]

Raw PWM is one of the worst and lowest efficiency methods of controlling a TEC.  Filtered PWM giving low ripple (i.e. effectively a buck regulator) is fine.

Really? ok, i probably should look into TECs a bit more.

[Chalcogenide]

Yeah, using a buck (bi-directional if you want a faster PID loop) is typical for TEC: see the MAX1968 / MAX1969 which are excellent all-in-one TEC controller chips, although a bit tricky to setup as the entire PID loop is set by analog components.

[Psi]

I assume, being a uni project, the pic32mx needs to control the TEC, and be doing the PID loop.

[Siwastaja]

Raw PWM is one of the worst and lowest efficiency methods of controlling a TEC.  Filtered PWM giving low ripple (i.e. effectively a buck regulator) is fine.

Really? ok, i probably should look into TECs a bit more.

The point is simply this: resistive losses are I^2*R. Running the TEC at 100% rated voltage and current you get the rated efficiency. Now if you run it at 50% duty cycle but still 100% rated voltage and 100% rated current (during ON state), you are still running at rated efficiency; not any better, not any worse. But if you instead run it continuously with lower voltage, you get better than rated efficiency. Same thing with LEDs. Efficiency goes down as current goes up.


On the other hand, if you have designed your Peltier gadget (specifically hot side cooling, which is non-trivial to get right) to work at 100% duty in some conditions, and can handle the heat it generates at that point, then surely PWMing to lower power will not be a problem: you have less heat to work with. It's only that by PWMing, you miss the opportunity to do even better, but if the design handles full power, then one could question, why bother optimize the partial-power use case, especially if it's a scientific instrument (or a toy, for that matter) where energy cost is not that important.

(And if energy cost is important, you would not use a Peltier at all.)

[Kleinstein]

Using PWM still need a suitable maximum power. The nominal current is normally the one for maxium cooling with perfect cooling of the hot side. Depending on the cooling of the hot side the actual maximum useful current is lower to some degree. So one would still need some adjustment for the maximum current to use with PWM.

Another problem with simple PWM is the EMI and thermal stress from an oscillating temperature.

PWM with filtering to make it a kind of buck converter is often good enough, even if there is some resudual ripple.


For good regulation the controler should compensate for the nonlinear reponse of the peltier cooler and possibly also for variations of the the temperature of the hot side. The hot side is often one of the major sources of disturbance. Checking the hot side temperature is anyway a good idea to detect overheating - e.g. with a non working fan.  So a good controller would have a little more than just plain PID.

[Psi]

Raw PWM is one of the worst and lowest efficiency methods of controlling a TEC.  Filtered PWM giving low ripple (i.e. effectively a buck regulator) is fine.

Really? ok, i probably should look into TECs a bit more.

The point is simply this: resistive losses are I^2*R. Running the TEC at 100% rated voltage and current you get the rated efficiency. Now if you run it at 50% duty cycle but still 100% rated voltage and 100% rated current (during ON state), you are still running at rated efficiency; not any better, not any worse. But if you instead run it continuously with lower voltage, you get better than rated efficiency. Same thing with LEDs. Efficiency goes down as current goes up.

Ah, go it