2-way temperature control with Peltier Element

Hi, I am working at a project to hold constant the temp of about 15 liters of watery liquid, which tends to stray down below its ideal 22.6 C by about 3 degrees at night and tends to go up about 5 degrees during one working day (this is to give you an idea of what sort of power comsumption I'll need). Accuracy of 1 K hysterese is fine, even 1.5 K is ok. I'll use two DS1820 to check the temp of the liquid flowing to and again from the peltier

I have a 12 V 8 A peltier and also a suitable power supply (12 V 15 A). I have a heatdrain on one side of the peltier element so that is also not a problem.

The problem is the 2-way regulation of the element. I would have simply used an H-Bridge but PWM to run a peltier is supposed to be a nogo, at least without LC filtering as wavy supply the peltier reduces efficiency significantly.

Mechanical relay may be fine but I was thinking of programming the current to the peltier to drop step by step as delta-K gets smaller. A mechanical relay sending full blast power to the peltier is perhaps not very elegantsetup.

Hmmm, any views?

...Ahh, did I mention, I'll be using an Arduino Duemilanove.

Just because you are using an H-bridge doesn't mean you have to PWM your control power.

analogWrite(myPin, 255);


digitalWrite(myPin, HIGH);

will put the power on full, without any pulsing. So, if you are happy with simple on/off control and hysteresis rather than proportional control, you are fine.

The H-bridge is all about reversing the polarity. If you feed it with pulses it will turn on and off with pulses, but its just as happy with a simple on/off.

Oh, ok. So I could, as an example, simply use the Freeduino Motor control shield, run it "full open" in either direction and it would output clean, flat DC?

I was thinking, if, for example, I run it at 1 kHz, 100%, it'll still have 1 KHz impluses in the output.

Thanks for your input.

I re-read your reply. You meant, not to run it at any frequency at all, so no question of spikes or pulses?

OK, got it now I guess.

Nope, in fact, just use digitalWrite rather than analogWrite.

Check the current handling specs before you order. I think Rugged Circuits do a higher power version.

In any case, the nominal current capability of the Sparkfun motor shield I bought, was very optimistic. It soon got hot and turned itself off running a bit more than 1 A per channel.

15 litre of a watery liquid isn't going to move by a degree very quickly with a 96W peltier turned full on. A peltier on cooling is very inefficient (single figure percentages) so a few watts on 15 litres is next door to immaterial. My 3kW litre kettle takes around 3 minutes 20 seconds to boil 1.5 litres from 20 degrees C, 15 litres would take about 33 minutes or around 2.5 degrees per minute. Extrapolate that down to say 10 watts of cooling and it comes out at around 0.5 degrees per hour, hardly PWM territory. Heating it would be much better, you'd get the heat its transferring + the electricity its guzzling, say 105 watts. So it would "only" take a few minutes to move it 0.5 degrees, thats hardly PWM territory either.

Just switch it on and off under control of your sensors.

Yup, loads of thermal inertia, but their problem was that they thought that H-bridges came with pwm - rather than that they are 2 separate things.

They want h-bridge because they want to reverse the polarity of the P-element (I can't spell) so that it will cool as well as heat. I assume that sometime the ambient is higher than the desired temp.

I need the system to be quick (I know it is a relativ term) in the morning to bring up the liquid to working temp. The liquid gains about 350,000 Kj during a working day (15 liters x 5 Kelvin x 4.2 Joules per degree per gramm water), so I have the entire day to drain off that extra energy.

In fact, if the system ain#t quick enough in the morning, I probably would add a resistive heater as a worst case scenario.

My problem really was that I kept thinking H-Bridge and PWM are married! That doesen't seem to be the case so I'd happily go hand in hand with the H-Bridge and leave PWM standing :-)

At the risk of stating the obvious.

Why don't you insulate your tank really well, and save yourself a whole load of energy?

Si, it is a developer tank in the plate making dept in the printing industry, it is 3% NaOH solution with some antioxydants. We let the temp drop during the night to reduce oxydation. The correct developing temp is 21.6 (Sorry, I had mistakenly written 22.6 in my first post). During the day, we run about 5.5 sqm aluminium sheet of 0.3 mm through the bath per hour. I guess it isn't only the ambient temp that's ruining my day but chemical reaction etc as well. The original cooling and heating system conked off... so did the manufacturer :-(. Now you know the rest of the story.

Fortunately we are, due to our relatively constant ambient temp, already quite close to working temps so we need a small, simple system to really just "nudge" the temps right. Our delta-t is really small so I seriously think our grappling with the problem may bring us success. Thanks for all the information and suppost :-).

Thanks too to Pluggy for the useful input, I seriously thought the H-Bridge NEEDS to be modulated to function.

Good. Sounds like you have a plan :slight_smile:

If you are interested, there is a good description of H-bridges here:

If you are feeling brave, its not that hard to make your own, but a motor shield is even easier :wink:

Actually the shields protect themselves fairly well, so if they get too hot they drop the power. You should include a bit of ventilation (holes) in the box and possibly one of those self-adhesive heat sinks for the chip in the middle.

I have a feeling that the shield you mentioned is waaaaaay underspec for the current requirements. You're going to need something far more beefy. As has been suggested, a basic H bridge isn't very complex to make. The biggest challenges come when PWM is needed (and the fast switching required with that) so you shouldn't need much more than a handful of mosfets(P and N channel), some resistors and some diodes.

I have 10 MOSFETS F14N05L lying around. That's 14 Amp, 50 Volt N-Channel Type. I would use those, should work, no?

Hunter62: I have 10 MOSFETS F14N05L lying around. That's 14 Amp, 50 Volt N-Channel Type. I would use those, should work, no?

Looks perfect to me.


Just a thought but a H-brigde does not have to be simply the on / off thing we normally use for controlling motors. You can run each leg in the linear region and have a continuously varying see saw. Mind you it pushes up the dissipation in the H bridge active devices.

N-Channel logic level Mosfets are fine for the low side switching. For the high side, you're going to need to provide a higher gate voltage (minimum of 5v + source voltage). I know a lot of folks around here are big fans of logic level mosfets but I'm not so sure.

For the high side, you're going to need to provide a higher gate voltage

Unless you use P-channel FETs on the high side, with the gate pulled up to +ve, you then pull it down with a transistor or other n-channel FET.

That's what I originally suggested Mike, however the OP responded to say that they happen to already have 10 logic level N channel mosfets. :)

Sorry missed that. :blush: