Low Pass Filter for Peltier PWM application

Discussion in 'The Projects Forum' started by basline, Aug 16, 2012.

  1. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    I'm pretty much a total noob to electronics, and seem to be in way over my head trying to complete a project. I'm planning on using two peltier devices (Vmax 15.6V, Imax =26A) as my heat pump for a cooling/sometimes heating application. I've done way more research than i was anctipating I was going to need to, and still feel like I really have no idea whats going on.

    I picked up two meanwell 15V DC power supplies rated for 20A, and Im hoping to wire each peliter separately with a power supply and use a microcontroller to control the voltage they see with PWM. I tried to figure out wiring the supplies in parallel as well we as the peltiers, but it seems like that would be the much more complicated route.

    One of these peltier circuits will be designed to heat and cool with an H-bridge (still muddling through picking parts), while the other will be designed to cool only. An FAQ from a TEC manufacturer suggests that if you used PWM to keep voltage ripple to less than 10% for best performance and longevity. So my goal is to design filters to give me 10% voltage ripple or less at a 50% duty cycle if I'm using a 20kHz signal from an Arduino.

    From my what I think I've pieced together I think I figured out how to design a low pass RC low pass filter (cutoff freq around 2kHz). But given the current I'd be working with adding a resistor will waste alot of power.

    So it seems like I need an LC low pass filter, but I’m having a hard time finding good web resources on this as most of them are for AC power and radio design. With my spotty understanding I took a stab at sizing an inductor and it looks for the size I’d need (on the order of 10's of micro-henries) for the current it needs to handle they will be expensive. If I need 3 of them (one of them to filter cooling in the h-bridge, one of them for heating in the h-bridge and one of them for cooling-only circuit) then filtering may start to get components may start to get prohibitively expensive for this approach.

    I could seriously use help, I'm pretty lost here. I know there is alot I need to learn, but this is getting really overwhelming for what I was hopping would be a pretty simple project. Any help with sizing an LC filter would be really appreciated. I could also potentially change out peltiers for higher voltage (Vmax 31.5V, Imax 17A), can I make this work with my power supplies and would it simplify the filter situation?

    thanks in advance
     
    Last edited: Aug 16, 2012
  2. crutschow

    Expert

    Mar 14, 2008
    13,014
    3,234
    What do you mean by expensive? Here's a 10μH, 38A inductor for $6.63, for example.

    Also, I would think you would only need two inductors. You should be able to use one for both heating or cooling at the input to the H-bridge.
     
  3. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    Have you considered using straight DC on/off control? I mean, I assume you're interested in controlling the temperature of something. You don't need PWM to accomplish that. You might like reading about my related calorimeter project (in Completed Projects). I only used cooling, but the ideas will transfer.
     
  4. crutschow

    Expert

    Mar 14, 2008
    13,014
    3,234
    The reason to use filtered PWM to control the temperature is because the Peltier cooling is proportional to the average current but the Peltier loss due to its internal resistance is proportional to the RMS current (current squared). Thus filtered PWM control is more efficient than On/Off or unfiltered PWM control between maximum and zero current.
     
  5. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    I get that, but I doubt the OP cares all that much about efficiency. I could easily be wrong.

    FWIW, I operate my on/off TEC with the "on" current far below the max current. It is far more efficient and reliable at less-than-maximum current. That lowers its capacity, but adding more TEC capacity is cheap.
     
  6. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    Thanks for explaining the choice of PWM crutschow.


    With regard to inductors:
    My back of the envelope calculation and quick filtering through digikey yielded the cheapest result of something like $30 for a like 35μH inductor that would take 20A. Since I'm still totally new to this I was worried this was the price range they were going to be in. If I only need 2 that also makes this seem reasonable.

    I am a little fuzzy on how to use a single filter with the H-bridge, but I'll hold my questions until I have a circuit diagram to ask with.

    In the mean time, can someone confirm the equations I should be using to determine the sizes of L and C:

    Cutoff Frequency = 1/[ 2 x pi x sqrt(L x C)]

    V = L x dI/dt
    where dt is my PWM freq in seconds, and V is the voltage across inductor

    My questions:
    dI should be my desired current ripple, correct?

    Also, to determine V across the inductor, do I need to know what the voltage drop across the peliter is going to be? Shouldn't that basically be duty cycle x the voltage of my power supply?

    Finally, are there rules of thumb I should be aware of for capacitor sizing? Seems like 22 ηF, give or take 3 orders of magnitude, shows up alot.

    Sorry to need the hand holding, I'm trying to tie together several weeks of having sunk all my spare time into trying to learn this online and I still have some basic questions.
     
    Last edited: Aug 16, 2012
  7. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    Actually I was hoping to make this rig as efficient as is feasible. I know people don't equate TEC with efficient, but I'm hoping my control strategy (assuming I can figure out the power electronics portion) will at least provide COP above 1.

    By operate below max current, does that mean you control the voltage it sees? This has been my working assumption, but again as a noob this is one of those stupid questions I haven't be able to find a cut-and-dry answer to.

    I've also sized my TEC capacity a minimum of 3-4 times my cooling load, in the hopes of being down in the efficient operation ranges.
     
  8. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    Not gonna happen. You'll get about 1 heat unit moved from cold side to warm side for every 9 units of supplied power. So 10 heat units show up on the hot side for every 1 removed from the cold side. More-or-less.
     
  9. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    wayneh are you controlling voltage or current for your TECs? I just assumed most folks are controlling them based on voltage, I'm not even sure how you would control then with current.

    So I don't want to redirect this thread from my questions in post above, but I've read this alot and I'm not sure if that's entirely accurate. I'm still in the designing phase, but my understanding is that the ratio of heat moved to energy consumed (COP) is a function of:
    -heat load the TEC is under
    -the ΔT between the TEC sides
    -voltage available to TEC assuming current is not a bottleneck
    And that these parameters don't follow linear relationships. So at low voltage, high heat load and low ΔT peltiers can cool with COP greater than 1, which for my project will largely be determined by my control strategy. All of this is based off of TEC spec sheets, and an FAQ with calculator a PC overclocker who goes by the handle ultrasonic2 put together.

    Again, I hope this doesn't distract us from my basic power electronic questions, as I'm sill struggling mightily to figure this out.
     
  10. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    I "control" the TEC only by turning it on or off with a 5V (computer PSU) supply.
    The TEC itself takes the current it wants, a bit less than 5A since the resistance is roughly 1.2Ω.

    You are correct that the COP can exceed 1 and even approach 2. It's rarely seen in the typical DIY device encountered here, and that's where my comment was directed. Everybody thinks they want to run them at the rated current.

    As you have already learned, using the TEC far below its rated maximum will get you much higher efficiency than running balls out. Efficiency comes at the expense of TEC capacity. But again, that's cheap to fix if you have the space to place them.
     
  11. crutschow

    Expert

    Mar 14, 2008
    13,014
    3,234
    If you operated it from filtered PWM rather than on/off it would be even more efficient. ;)
     
  12. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    True, but for my calorimeter it's important to have an accurate log of on and off times, to calculate the duty cycle, which relates to an estimated heat removal rate. This would be tough to implement with PWM.

    I'm sure you could keep track of the PWM duty setting instead of trying to actually measure it and this might even be a superior approach, but such a thing is beyond my pay grade. Certainly was when I started building it. Now that I have a LabJack with computer-controlled PWM at my disposal, I would likely take a different approach.
     
  13. crutschow

    Expert

    Mar 14, 2008
    13,014
    3,234
    Well, you could measure the average voltage into the Peltier from the filtered PWM signal to estimate the cooler heat removal rate.
     
  14. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    I could now, yes. I built the device before I had the chops or equipment to do that. Things would be different now. :D
     
  15. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    If you are using PWM into a large inductor then I would just add a diode and cap and make it a DC-DC buck converter. You'll get a nice smooth DC for the peltier and good energy efficiency.
     
  16. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    So after finding some local help, and a whole lot more research, I'm finally getting close. Since I don't need to change polarity quickly I've moved to using two SPDT relays rather than an H-bridge. I've also changed to higher voltage peltiers, which draw less current (Vmax 34.4V, Imax = 12 Amps). Both of these things have revised the circuit.

    My tentative circuit is attached. I still have no idea what I'm doing. I'll be using an arduino as my micro-controller, with PWM of 20kHz or 32kHz. My optoisolator will be 6N137, mosfet driver will be FAN7392 (and only use the high-side). I'll wire my two 15V power supplies in series, use a buck converter to get average voltage, and use a modified sparkfun relay kit that will be SPDT relays.

    I alot of help answering the following:

    -Missing pull-up and pull-down resistors?
    -Are things grounded/floating correctly?
    -What type of diodes should I be looking for?
    -Capacitor sizes?
    -Missing capacitors?
    -Really basic missing components?

    Any help will be greatly appreciated. I'm hoping to order parts this week if I can figure this out.
     
  17. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    I'm not getting why there are multiple caps in parallel (C7, 8, 9).

    I think one big risk with this strategy is that one relay trips while the other hasn't. Isn't there a dual relay that could absolutely prevent that from happening? Otherwise I'd say fuses would be a good idea.

    Of course fusing is in general a good idea.

    I'd add a pull-down resistor on the gate of the MOSFET, to protect against the driver failing to "open". the pull-down would ensure the fet stays off in that case.
     
  18. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    Thanks for the feedback wayneh.

    I added fuses, although I'm not entirely sure if they are in the correct location, as well as a pull down resistor.

    I also changed the mosfet gate driver to something I could find more application notes for, IR2110.

    I'm still unsure about what should be connected to earth, vs the negative of my power supplies. I've tired to be diligent about that in my diagram. As I understand it, I want everything except the microcontroller floating. Is that correct?
     
  19. wayneh

    Expert

    Sep 9, 2010
    12,123
    3,048
    It'd be best if someone else chimes in here - I'm not able to comment on much of what you have.

    Couple questions though: You have your supply caps C4 and C5 in parallel with your batteries. I'm not sure they'll really do much good in that scenario, although I guess the idea is they can smooth spikes that would draw down the battery voltage due to internal resistance of the batteries.

    I guess D1, L1 and C3 are your LPF? Without understanding all this better, it bothers me that the MOSFET source pin isn't directly connected to battery minus.
     
  20. basline

    Thread Starter New Member

    Aug 16, 2012
    20
    0
    Thanks again wayneh for all the input.

    L1, D1 and C3 are my buck converter, which in theory should give me relatively smooth voltage and current.

    I think my lack of knowledge may be part of your confusion. The 12V and two 15V power sources are power supplies, not batteries. I couldn't figure out how to include floating ground rather than ground to earth with my layout software, so I used the symbols I have in there. Since the 30V going into my buck converter will be cycled on and off around 32kHz, I think I need capacitors C4 and C5 to keep from straining the power supplies.

    I'm not sure if I should start a new thread or not. Ill take a look and the forum rules and see what makes sense.
     
Loading...