jroorda:
There are two different and conflicting issues here. The first is that a traditional PWM scheme is a bad idea because it will be inefficient. As you increase the voltage, or current through a Peltier the cooling output will increase reach a peak and then decrease due to the additional power dissipated in the Peltier. In a PWM controll scheme you are emulating a lower supply voltage / current by pulsing a higher one on and off. But in this case the higher your input voltage is the less efficient the device will be, so you are much better off having a lower steady input voltage than a higher pulsed input voltage from an efficiency standpoint. That said, if your Peltier is sufficiently oversized you might be able to just write off the efficiency loss.The other issue with PWM is thermal cycling. Anything heated and cooled enough times will be mechanically harmed by the effects of differential expansion and contraction of materials. What you want to avoid is a rapid thermal cycling of the Peltier device. If you PWM quickly (faster than 10 Hz) no appreciable temperature change will occur in the device during the cycle and you should be fine.
If you can capacitivly filter your PWM signal so that it is essentially a DC signal you will solve both of these issues at one time, so that is the best solution, but a quick PWM with a maximum voltage that is still in the good efficiency range would probably work.
This is very good info. A PWM signal filtered by a capacitor looks like the best approach for temperature control, but what about humidity control on top of it? It seems to me that switching the PWM signal on and off according to measured humidity levels would lead to the same expansion and contraction issues I am trying to avoid. What, then, would be the best way to accomplish the same effect?