Hey guys,
I was curious on the advantages if using PWM versus Linear voltage for Peltier modules. I read online that people recommend using PWM at high frequencies (15Khz+) but with a LC filter to get a nice DC voltage. Why not just control a MOSFET via linear mode then instead?
I am powering a 12v 25amp module and I am trying to figure out the best method to control its efficiency.
Why PWM when I could control the MOSFET with a voyage divider setup and a digital pot?
The reason why Is efficiency and the control of the amount of cooling power I need. If I don't control the cooling power, it will be below ambient and create condensation.
This is actually used as a liquid chilling process for liquid cooling a high-end desktop for over clocking etc...
Also the TEC module becomes more efficient at lower voltages than the max at 12v. You'll get more cooling power with less heat output with higher efficiency.
The goal is to use Arduino to monitor the temps on the hot and cold side and constantly adjust the signal to find the most efficient power load based on current thermal stress.
I hope this explains what I am trying to do so you guys could provide your expertise on this.
Is PWM for controlling voltage and driving a MOSFET via linear mode only for controlling current?
PWM transients are nothing to worry about if you smooth them out with a cap(s). You could still use the FET for more current, and just PWM the gate. Use caps to negate the transients and you're all good. Experimentation is needed to find your values.
the Peltier assembly which is an assemblage of dozens of diode chips.
No they are thermocouples, in effect just junctions between dissimilar metals.
I agree with Richard on this:-
The reason why Is efficiency
It is way way more efficient to turn a device on and off then put a varying DC voltage to it. The reason is that the amount of power not being used by the device is burned off in the controlling device. Where as on / off wastes very little power. When off there is no current so the power in the device is nil, when on there is lots of current but not much voltage across the device so the power is low.
Thermal mass is large compared with the heat pumping capacity of these devices so on / off at greater than a second is enough for very fine control of the amount of heat.
I written about this in the past so I'm repeating here, sorry.
I have tried PWM using an SSR (110V Heater Coil & 15 AMPS) to power a 2000 degree F kiln.
Fast PWM WORKED... from the heater coil perspective... I achieved my goal of temperature regulation... with 2 side effects. 1) My heater coils "sang" to me. very noisy. 2) My SSR generated a TON of heat.
So, I adjusted PWM to be a 2 second cycle... IE; ON time was a percentage of 2 Seconds. Much less heat was generated on the SSR. Noisyness was gone. Results = same.
Thank you everyone for your comments on this project. After further research, PWM is recommended at high frequencies > 15KHz. However, they recommend creating a LC filter along with it.
My question is, do I add the LC filter to the output of the FET (Drain) like FET Drain -> LC Filter -> Peltier or in line from: Arduino PWM -> LC Filter -> Gate? I assume after the Drain and before the peliter module?
Also, I tried researching on what I need exactly in values for the inductor and capacitor? The Arduino has it's 5V PWM signal and I would be running it at 31,250Hz by adjusting the PWM timer on pins 9 & 10. The peltier demand is 12V and 25AMPs.
My question is, do I add the LC filter to the output of the FET
Not recommended as you will blow the cap with the ripple current, and heat things up with the coil with 25 amps flowing. You seem to be hell bent about feeding a dc signal into a FET and using it in liner mode despite everyone telling you it is not a good idea. Still there's no helping some people. 
I'm sorry if I came across the idea of not using PWM. It was a complete opposite of that.
I got so caught up with using Google and PWM for Peltiers that random individuals were saing use it, others said don't, some said use a LC filter, etc... It makes sense to use a straight PWM frequency, and actually, Richard you are right. 1.5Hz would be better overall.
I'm sticking with PWM as your recommendations, and use 1.5Hz. Right now I'm driving the module straight with a 12V meanwell PSU. I'll try the 1.5Hz compared to the meanwell and see how it works out. 
Efficiency and Peltier cooling don't belong in the same sentence in my book. They are hideously inefficient when the temperature differential is low and it goes down as the temperature differential increases. If you're serious about water cooling your hyper games rig, pump it through a small domestic heating radiator - absolutely no condensation problems.
Throwing 300 watts of electrical power at moving maybe 20 - 50 watts of heat is a real waste. A domestic fridge consuming maybe 50 watts of electric power will move in the region of 150 - 250 watts of heat.
... suggested that the repeated on/off cycles could be physically damaging to the Peltier assembly
Thermal cycling is what causes damage. The ideal solution, from a longevity point of view, might be a variable linear voltage supply but PWM at 1 KHz or higher is a very good alternative. From a performance point of view, manufacturers recommend keeping ripple current below 10% which is likely where the LC filtering suggestions are coming from. For what it's worth, I have been driving a pair of 4A Peltier units for over 2 years using 1 KHz PWM (no filtering).
Random notes: Peltier units are sometimes referred to as TECs - thermal electric coolers. The Peltier units I'm familiar with use an array of bulk P and N semiconductors, not dissimilar metal junctions (but they still aren't diodes). And finally, "hideously inefficient" seems a bit extreme. I haven't looked at the data in a while but I didn't think they were quite that bad.
-Carl