I have some peltier units that I would like to deliver variable voltage to, and according to app notes I've read, it is best to feed these straight DC rather than PWM. An Arduino or similar PWM output will be controlling the voltage level to them so I would like to have a power supply which can have its voltage output determined by a PWM input. The power supply needs to deliver at least 6A and input will be 12-14V (automotive). I dont know yet what max output voltage will be required but hopefully not more than the input; but to be safe it may be best to hope for a buck/boost type. This will be an embedded affair, so a bench top style supply won't be suitable. I was hoping to find some module on Amazon or such, an inexpensive little PCB that would do the job, rather than building my own supply. If anyone knows something that would fit the bill please let me know.

 

Basically the problem with PWM and TECs is thermal stress/expansion/contraction related, so as I under stand to minimize these you must use a very high PWM frequency. (so the next pulse comes before contraction)

I know that is not the information you asked for, just something I learned when I was doing research on thermal electric units.

So the option to use PWM straight to the TEC is still available.

 

You can also add a capacitor filter at the output of the PWM supply, which will smooth the pulses and deliver a more constant voltage level. And it will still be rather efficient.

 

The other problem with direct PWM to a TEC is that it increases the IR loss (and thus reduces efficiency) of the TEC.
The TEC resistive loss is proportional to I² so for example, at 50% duty-cycle PWM would generate a loss I²R/2 where I is the full voltage current of the TEC and R is its resistance, whereas running at 50% DC voltage would generate a loss of (I/2*I)²R or 1/2 the loss of PWM.

If you want to use a switching regulator, the example below shows a simple way to control the circuit with a voltage (either from a DAC or filtered PWM). It shows a buck converter but it should work as well for a buck-boost.
Discussion here.

 

You can also add a capacitor filter at the output of the PWM supply, which will smooth the pulses and deliver a more constant voltage level. And it will still be rather efficient.

Not so much.
Charging the capacitor through the PWM switch will generate significant resistive loss in the switch unless a series inductor is added (to make it like a switching regulator).

 

Basically the problem with PWM and TECs is thermal stress/expansion/contraction related, so as I under stand to minimize these you must use a very high PWM frequency. (so the next pulse comes before contraction)

I know that is not the information you asked for, just something I learned when I was doing research on thermal electric units.

So the option to use PWM straight to the TEC is still available.

The option is there, of course. In fact these Peltier units are from (in) the heated/cooled seats of Jaguars/Land Rovers and the climatic seat controllers in those vehicles use PWM. But according to app notes (example) it is much more efficient to use a constant voltage*:





(*) They are using a constant current supply in the app note, but my understanding is that these modules have a relatively linear voltage:current relationship, so a constant voltage supply should work just as well.

The hope is that (as @crutschow ) pointed about the I²R losses, I should be able to get more cooling out of these seats than they can manage in their factory control scheme, without any ill effect. Also the Peltier units in these seats are prone to failure, i suspect in part to the PWM control scheme. So hopefully this will also be a more robust design.