Arduino PWM Driver for 12v Peltier

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
Hi Folks,

I have started making plans to build a small, efficient and quiet campervan fridge. I'm going to use very high quality insulation, thick insulation, so that I don't need to do too much cooling.

I actually started this project last year, but have not thought about it in a few months, so have forgotten a few things and need some advice.

My general approach is to use an Arduino R3 to sense the temperature in the fridge using a thermocouple, then send a PWM signal to the Peltier modules to bring the current down to the required amount to cool/maintain temperature.

As the system powering the modules is a 12v leisure batteyr, I have bought two TEC1-12709 16v, 9a, 90w modules. My plan is to run them underpowered to maintain efficiency. My understanding is that they don't like PWM due to I^2R losses, so I was planning on putting an LC low-pass filter in. This i where I'm getting a bit lost. A few other threads have said to also in a diode, creating a DC-Buck converter - however, are DC-buck converters not normally for stepping down voltage?

I'm planning on using a logic level mosfet (TIP120 suitable?) to switch the +12v supply. I'm a total circuit/control newbie, so any advice, even if it's a pointer on where to go and learn the basics would be great.

My initial power-switch circuit is attached. I'll worry about the temp sensing part later.
 

Attachments

R!f@@

Joined Apr 2, 2009
9,747
Forget PWM.
Just switch on the MOSFET to cool it and OFF when the temp is reached.
TIP120 is not a MOSFET.
IRL3713, IRLR8729PBF and the likes a Logic MOSFET's
 

crutschow

Joined Mar 14, 2008
27,213
so I was planning on putting an LC low-pass filter in. This i where I'm getting a bit lost. A few other threads have said to also in a diode, creating a DC-Buck converter
Exactly.
You are lowering the average voltage to the TEC cooler to improve its efficiency, which is what a buck converter does.
You don't need a capacitor, just an inductor and a diode.
The TEC doesn't much care about ripple current so you don't need to smooth it with a capacitor.

The inductor will need to be more than one nH, more like a few millihenries depending on your switching frequency (do you know what that will be?).

The selected MOSFET should have no more than 10 milliohms of ON resistance to avoid needing a heatsink.
N-MOSFETs are generally cheaper for a given ON resistance, which you can use if you switch the ground side of the circuit instead of the hot side, as shown below.
upload_2017-4-29_12-24-42.png
 
Exactly.
You are lowering the average voltage to the TEC cooler to improve its efficiency, which is what a buck converter does.
You don't need a capacitor, just an inductor and a diode.
The TEC doesn't much care about ripple current so you don't need to smooth it with a capacitor.

The inductor will need to be more than one nH, more like a few millihenries depending on your switching frequency (do you know what that will be?).

The selected MOSFET should have no more than 10 milliohms of ON resistance to avoid needing a heatsink.
N-MOSFETs are generally cheaper for a given ON resistance, which you can use if you switch the ground side of the circuit instead of the hot side, as shown below.
View attachment 125755
Hi @crutschow,

This general project is on a list (a very long list) of things that I want to do. Reminded by this thread, RE: using a capacitor and ripple, I was looking around and found this other thread on the subject which I know you know about because you had some posts there. That fellow ended up with a schematic using a 100 uf capacitor. Also, these folks (who sell an interface) make an issue about ripple here. Seems like a capacitor wouldn't hurt and may be helpful - no?
 
Last edited:

OBW0549

Joined Mar 2, 2015
3,566
...I was looking around and found this other thread on the subject which I know you know about because you had some posts there. That fellow ended up with a schematic using a 100 uf capacitor. Also, these folks (who sell an interface) make an issue about ripple here. Seems like a capacitor wouldn't hurt and may be helpful - no?
You might also want to look at the discussion on TEC performance degradation due to TEC ripple current on pp. 17-18 of the datasheet for Linear Technology's LTC1923 High Efficiency Thermoelectric Cooler Controller chip. The bottom line appears to be: use capacitors, and keep ripple current below 10% of the DC current level.
 

crutschow

Joined Mar 14, 2008
27,213
I have no problem with adding a capacitor to the circuit for reducing the I²R loss in the TEC due to ripple.
Whether it's needed depends upon the inductor value and the switching frequency, which determines the inductor current ripple.
 

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
Exactly.
You are lowering the average voltage to the TEC cooler to improve its efficiency, which is what a buck converter does.
How can I calculate what voltage the TEC will see?

I do want to be able to run them at higher power/lower efficiency some of the time, for more rapid cooling, which is why my original plan was LC lowpass filter rather than a DC buck. Is it possible to have a DC buck that doesn't actually change the voltage?

I.e. My original aim was to use PWM to lower the current, LC filter to create pure (ish) DC and control feedback to alter the PWM to get to the desired temperature. WHat's the advantage of the diode in the DC-Buck, rather than just a LC Low pass filter?

How do I calculate the required values of L, C etc?
 

crutschow

Joined Mar 14, 2008
27,213
How can I calculate what voltage the TEC will see?
The TEC voltage would be approximately the PWM duty-cycle times the supply voltage.
My original aim was to use PWM to lower the current, LC filter to create pure (ish) DC and control feedback to alter the PWM to get to the desired temperature. WHat's the advantage of the diode in the DC-Buck, rather than just a LC Low pass filter?
Those two are basically the same thing.
The diode is just to allow the current to flow in the inductor when the PWM signal goes to 0V, so you need the diode in either case, otherwise you would get a big negative spike at that point.
Use a Schottky diode for best efficiency.
How do I calculate the required values of L, C etc?
This calculator should help.
 

crutschow

Joined Mar 14, 2008
27,213
How does that work ?

Say for 50% duty cycle ?

Like 0.5 * Vcc ?
Here's a demo showing the average output of the PWM signal for a 0 to 100% duty-cycle change.
Note there is a slight delay for the output voltage as compared to the PWM duty-cycle due to the filter delay (1 time-constant of 50ms).

upload_2017-4-30_13-54-57.png
 

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
Thanks for all the great info so far.

I've ordered a pair of N-Channel Mosfets: https://dlnmh9ip6v2uc.cloudfront.net/datasheets/Components/General/FQP30N06L.pdf

I'm still struggling with the other components. I had a look at the calculator linked above, but wasn't sure of some of the inputs. What is confusing me is that I want the circuit to be able to, in effect, operate at any given duty cycle from, say 10% to 100%, with 100% being 12v @ 9A (but 2 Peltiers, so 18A).

To me, this says my buck-converter input and output will be 12v?? And at 100% duty cycle the circuit won't really be doing anything? Do I have to optimise for certain duty cycles, or will it perform efficiently across the range? Sorry for all the basic questions...
 

crutschow

Joined Mar 14, 2008
27,213
............operate at any given duty cycle from, say 10% to 100%, with 100% being 12v @ 9A (but 2 Peltiers, so 18A).

To me, this says my buck-converter input and output will be 12v?? And at 100% duty cycle the circuit won't really be doing anything? Do I have to optimise for certain duty cycles, or will it perform efficiently across the range?
That is correct.
100% duty-cycle would be 12Vdc, with no switching.
Thus as you increase the duty-cycle, the ON pulses become longer and longer, and the OFF pulses shorter and shorter, until it becomes steady ON (DC) with no OFF pulses.

It will power the Peltier efficiently across the full duty-cycle range without any optimizing needed.
 

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
Ok, well I'm glad I'm starting to understand some things!

I had planned to PWM at 25KHz to keep it out of the audible region, but I' now concerned that the PWM driver on the Uno will not have sufficient current to charge the Mosfet gate at those frequencies. Is the noise something to be concerned of? Or should my capacitor be sufficient to filter it out?

I'm finding this complex enough without adding in a MOSFET Gate driver circuit!
 

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
Are you driving a peltier or a loudspeaker ? :eek:
I drive LED at 200Hz and I hear nothing.

Must something missing that I do not get regarding peltiers. o_O
No no, I just assumed it would be good to do that. Can't even think where I got that from.

I'm now concerned those MOSFETs aren't up to the job without heatsinks! Luckily they're only a few ££. In fact, before I buy anythign else I really need to get a few things ironed out.

1. If I use the above DC-Buck converter circuit, my understanding is that at lower duty cycles, it is the voltage, not the current that is reduced. I.e. the peltier will see 1.2v @ 10% duty cycle, but will it still draw 9A? Therefore the power dissipated in the peltier would be ~ 12W; however, the power dissipated in the MOSFET will still be 3.5 watt. This seems to totally defeat the point in this whole design - which is maximum efficiency when the peltier is in "maintenance" mode.

This is why I originally wanted to use PWM to reduce current, not voltage. Is this even possible??
 

crutschow

Joined Mar 14, 2008
27,213
at lower duty cycles, it is the voltage, not the current that is reduced. I.e. the peltier will see 1.2v @ 10% duty cycle, but will it still draw 9A? Therefore the power dissipated in the peltier would be ~ 12W; however, the power dissipated in the MOSFET will still be 3.5 watt. This seems to totally defeat the point in this whole design
No.
You are calculating the power for a linear controller, not a switcher.

With the inductor, the power supply peak voltage is not seen by the peltier at lower duty-cycles.
It sees the average voltage from the LC filter.

The MOSFET dissipation is just the square of the average current times the duty-cycle times the MOSFET ON resistance.
For the MOSFET you ordered with a 45mΩ maximum ON resistance, the power it will dissipate at 10% duty cycle (if the inductor is large enough for low ripple current) would be (10%*9A)²*45m = 36mW.
However, the power at 100% duty-cycle would be 9²*45m = 3.6W, thus you will need to put it on a small heatsink.

If you want to avoid a heatsink you would need a MOSFET with a maximum on resistance of about 12mΩ to keep the dissipation to a watt or less.

You can drive the MOSFET directly from the Arduino 5V output at 100-200Hz.
 

Thread Starter

Alasdair Fulton

Joined Nov 23, 2016
7
No.
You are calculating the power for a linear controller, not a switcher.

With the inductor, the power supply peak voltage is not seen by the peltier at lower duty-cycles.
It sees the average voltage from the LC filter.

The MOSFET dissipation is just the square of the average current times the duty-cycle times the MOSFET ON resistance.
For the MOSFET you ordered with a 45mΩ maximum ON resistance, the power it will dissipate at 10% duty cycle (if the inductor is large enough for low ripple current) would be (10%*9A)²*45m = 36mW.
However, the power at 100% duty-cycle would be 9²*45m = 3.6W, thus you will need to put it on a small heatsink.

If you want to avoid a heatsink you would need a MOSFET with a maximum on resistance of about 12mΩ to keep the dissipation to a watt or less.

You can drive the MOSFET directly from the Arduino 5V output at 100-200Hz.
I've not had any time to look into this since I've been up to my eyes with blown 2.5 kV IBGTs http://search.abb.com/library/Downl...LanguageCode=en&DocumentPartId=&Action=Launch

I have finally got round to doing the sums, and with a 200Hz PWM I seem to end up needing monstrously big caps and inductors to keep the ripple low. So, I'm in a bit of a catch 22. I'm thinking one way out of it would be to keep the native 32kHz PWM and use a little gate driver to make sure the Mosfet is properly driven. (obviously I could go somewhere in between)

A few questions, does the freewheel diode need to be rated for max current (18A) and 12v? Schottky's this size seem to dissipate a fair pit of power.

When using PWM, is it best to do your cap and inductor calcs at 25% load, 50%, 75% load? I'm more concerned about efficiency at lower to mid settings (i.e. minimising ripple) as at 100% there will be no ripple...
 
Top