next topic »

[on:]

Hi guys! I'm hoping to get some help with a project I am working on... I am using a peltier (thermoelectric cooler) as a dehumidifier, and I would like to set the cold side of the peltier to the dew point temperature. So I'd imagine the best way would be to some how use the arduino to throttle the current to the peltier, rather than using some sort of mechanical device like a relay to switch the current on/off real quick. I'd appreciate some help on this, or at least some hints in the right direction

Btw, the peltiers are 12VDC and run at around 11amps... Although I am looking at a much larger 400w version that at 12vdc uses 26amps.

Thanks

[Reply #1 on:]

That will be tricky, there was someone who posted some time ago who said that Peltier devices don't take too kindly to PWM signals. This only leaves the option of linear control of the current. This means the power you don't give to the Peltier devices is dissipated as heat in the controller which would normally be a FET.

[Reply #2 on:]

If you were using a PC power supply for the peltier you could switch it between 0V, 5V, and 12V to get a kinda variable control.

The simplest solution, IMHO, would be to just use a large thermal mass on the peltier so that it doesn't require frequent on/off switching.

[Reply #3 on:]

That will be tricky, there was someone who posted some time ago who said that Peltier devices don't take too kindly to PWM signals.

I've seen that comment as well but it doesn't match my experience.  I have been running a couple of TECs using PWM (at 1KHz) for several years with no signs of degradation.  Information I've found from various vendors says that PWM is acceptable - it's thermal stress from thermostatic control (and condensation if the TEC isn't sealed) that kills a TEC.  TEC vendors do recommend keeping voltage ripple at less than 10% but that is for maximum performance and so should only matter when the TEC is running at 100% PWM.

[Reply #4 on:]

In that case it would seem the best way is to use a good logic level FET and use that to drive the Peltier device.

[Reply #5 on:]

That is what I am doing.  The FET gets pretty warm (12A & Rds(on) of 0.020 ohms) with a small heatsink.  Switching 26A will be more challenging.  I also get a bit of audible noise (whine) from my power supply in a certain low PWM range.  Since my application rarely gets below 50% PWM, I haven't pursued a solution (yet).

For a dehumidifier, be sure to use sealed TECs as the condensation will kill a TEC due to corrosion.  Assuming you want closed loop control of the cold side temperature, you'll also need a moisture resistant temperature probe.  My first attempt using a purchased silicone sealed thermistor was a failure; the silicone didn't stick to the teflon wire insulation so moisture eventually got to the thermistor and changed it's resistance.  Several websites sell water resistant probes built around the DS18B20.  I had already made my own before finding those.

Good luck!

[Reply #6 on:]

The FET gets pretty warm (12A & Rds(on) of 0.020 ohms) with a small heatsink.

It will do it is 2.88 Watts of heat toy are trying to get rid of. You can get better FETs these days than 0.020R, you might have to use a non logic level one though and you will therefore need a driver.

I also get a bit of audible noise (whine) from my power supply in a certain low PWM range.

You need more bulk decoupling on the power supply. How much have you got now?

[Reply #7 on:]

You can get better FETs these days than 0.020R, you might have to use a non logic level one though and you will therefore need a driver

Agreed.  I've already found several logic level FETs with Rds(on) in the 5 mOhm range but haven't tried them out yet.  I currently have a fan blowing on the heatsink that is keeping it tolerable.  The FET I'm using has built-in overcurrent, overtemp, etc protection but that may be overkill for my application.

You need more bulk decoupling on the power supply. How much have you got now?

Good question - thanks for the tip.  I'm probably applying bigger transients than the designers planned on.  Lifting the cover on the supply, it looks like a pair of 1000 uF caps.

Hope the OP is finding all this helpful

[Reply #8 on:]

Hope the OP is finding all this helpful

I certainly am. I'm thinking of adding a TEC to my PC for the drive bays. Probably overkill, but irrespective of that, I'm filing away bits of knowledge.

[Reply #9 on:]

You can get better FETs these days than 0.020R, you might have to use a non logic level one though and you will therefore need a driver.

You can also connect 2 or more mosfets of the same type in parallel, although this will of course increase the total gate charge. Rds(on) has a positive temperature coefficient, so mosfets in parallel share the steady-state current well.