EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: poorchava on November 19, 2015, 08:05:42 pm
-
Hey, so I have an application, where i need to implement drive for a high power Peltier cell (~50W max to be delivered). The project is kind of cost-aware (not to an extreme level).
Because of power level linear control is kinda out of question. So buck it is.
After examining some options and comparing against BOM cost limit that I have and available real estate on the PCB I thought that maybe I can make it a low-side switched buck. This would look something like that:
http://i1261.photobucket.com/albums/ii588/poorchava/junk/2015-11-19%2020_33_54-LTspice%20IV%20-%20Draft40.asc_zps3rymuecu.png (http://i1261.photobucket.com/albums/ii588/poorchava/junk/2015-11-19%2020_33_54-LTspice%20IV%20-%20Draft40.asc_zps3rymuecu.png)
The advantage is that gate drive will be simple and compact using some integrated low-cost low-side mosfet driver (eg. MCP1415). Disadvantage is obviously difficulty in current measurement, but a differential amp or some cheaper current shunt monitor should handle that.
I know that this topology is sometimes used in LED drivers, so it is practical, but I wonder whether there might be some additional pitfalls when compared to classic, high-side switched buck?
-
Hey, so I have an application, where i need to implement drive for a high power Peltier cell (~50W max to be delivered). The project is kind of cost-aware (not to an extreme level).
Because of power level linear control is kinda out of question. So buck it is.
After examining some options and comparing against BOM cost limit that I have and available real estate on the PCB I thought that maybe I can make it a low-side switched buck. This would look something like that:
http://i1261.photobucket.com/albums/ii588/poorchava/junk/2015-11-19%2020_33_54-LTspice%20IV%20-%20Draft40.asc_zps3rymuecu.png (http://i1261.photobucket.com/albums/ii588/poorchava/junk/2015-11-19%2020_33_54-LTspice%20IV%20-%20Draft40.asc_zps3rymuecu.png)
The advantage is that gate drive will be simple and compact using some integrated low-cost low-side mosfet driver (eg. MCP1415). Disadvantage is obviously difficulty in current measurement, but a differential amp or some cheaper current shunt monitor should handle that.
I know that this topology is sometimes used in LED drivers, so it is practical, but I wonder whether there might be some additional pitfalls when compared to classic, high-side switched buck?
both of the output are "hot", but I'd say that is more of an issue with a LED because a heatsink would need to be isolated
-
TI have some interesting LED drivers that use this topology that avoid high-side current measurement by only measuring current through the low-side switch during its on period, and using the mid-point of the rising slope of the current wave form as the control point. The main advantage for them is that it allows common-anode configurations. It would be tricky to get good current regulation with that method from something as dynamic as a TEC, but for simple monitoring it might be adequate, and presumably it's the temperature that you want to regulate rather than the TEC current.
Ah, here it is: http://www.ti.com/product/lm3414 (http://www.ti.com/product/lm3414)
-
For my application I think I am going to use a low frequency PWM signal to control a low side MOSFET, just switching the Peltier element on/off, there should be enough thermal inertia to smooth things out.
-
This is a bad solution. You don;t PWM a Peltier, because the heat transfer action is proportional to current, but losses are proportional to current squared. Pwm is a high current - no current situation, while a buck is medium current all the time. You'll have the same heat transfer, but much worse resistive heating using PWM. On top of that Peltiers are ceramic, and they don't really like rapid turn on/turn off (I guess some might be piezoelectric too? Dunno).
I will use high side sensing, using either a dedicated current shunt monitor such as INA19x or a classical differential amplifier. A classical diff-amp is problematic, because of common mode range (unless I use gain of 1 and then a non-inv gain stage, which theoretically requires only 12V of opamp power supply in this scenario). I was also thinking about using a low side sense to measure inductor current during on-time only, similar to the way they usually do it in current mode boost converters. Since this will be driven from an uC, I know the PWM duty and if i can get the average value of current during on-time, I could calculate the average current through the load. I guess I could even place a low value shunt between MOSFET source and GND, degeneration should not be a problem.
-
Why not just use a constant current LED driver? Then just have your micro adjust the output current based on your temperature control algorithm.
Linear Tech has some nice chips that can do all the hard work for you and you just add the inductor/schottky/FET on the outside. A part like the LT3755 seems like a good option - its output current can be controlled via a single analog input (poor man's DAC on a micro's output pin - even easier if your micro has a PWM output).
cheers,
george.
-
Why do you need current sense? Peltier is mostly resistive and works rather linearly when controlled with varying voltage. I'm sure that unregulated buck will work just fine. No voltage or current feedback needed, as your input voltage is probably regulated (coming from a switch mode PSU).
Unless you need some exact power measurement, if you are doing some kind of precision thermal analysis instead of just cooling/heating.
-
PWM is a good solution, had to try it, take a look at the attached picture.
Upper trace supply voltage 12V.
Middle trace drain voltage.
Lower trace drain current.
-
PWM is a good solution, had to try it, take a look at the attached picture.
No, it's not. Once more:
- Using PWM instead of smooth DC, the Peltier efficiency is lower. This may not be a problem, because you usually design the hot side cooling for the highest power level, where the losses are highest anyway. But if you adjust your Peltier to 50% electrical power, smooth DC gives more cooling effect than the same electrical power provided by PWM. You need to provide more power to the element and get more hot side heating to get the same result with PWM.
- PWM can slowly (or not so slowly!) damage the Peltier element which has limited number of on/off cycles. It's a widely known problem, even without any PWM control. PWM can use the number of cycles very quickly. Peltiers are sometimes marketed as "robust" "solid-state" devices but they are really not.
-
While you can direct PWM such a load, you do still have to account for dynamics of the filter (ideally, it's critically damped by the load resistance, which fortunately is known in this case!). And it would be nice to have current limiting or some sort of protection, so you don't, y'know, fire pulses into a dead short, or something dumb like that (the electronic equivalent of STOP HITTING YOURSELF STOP HITTING YOURSE--...) under fault conditions.
FYI, the coefficients for an 2 pole Butterworth (good enough) with "zero ohm" source, cutoff frequency f, and nominal load R, are:
L = 1.4142 * R / (2*pi*f)
C = 0.7071 / (2*pi*f*R)
(at least, that's what the book says..)
Choose f << f_clk so the ripple has a chance to filter out. Probably 1/5 to 1/20 is enough.
Tim
-
PWM is a good solution, had to try it, take a look at the attached picture.
No, it's not. Once more:
- Using PWM instead of smooth DC, the Peltier efficiency is lower. This may not be a problem, because you usually design the hot side cooling for the highest power level, where the losses are highest anyway. But if you adjust your Peltier to 50% electrical power, smooth DC gives more cooling effect than the same electrical power provided by PWM. You need to provide more power to the element and get more hot side heating to get the same result with PWM.
- PWM can slowly (or not so slowly!) damage the Peltier element which has limited number of on/off cycles. It's a widely known problem, even without any PWM control. PWM can use the number of cycles very quickly. Peltiers are sometimes marketed as "robust" "solid-state" devices but they are really not.
Good point I didn’t take the power loss in account, found this explanation http://www.meerstetter.ch/compendium/pwm-vs-direct-current (http://www.meerstetter.ch/compendium/pwm-vs-direct-current)