Filtering PWM signals

Discussion in 'The Projects Forum' started by SoulBlade, Oct 11, 2012.

  1. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    Hi,
    I intend to drive a Peltier (Tellurex C2-40-1504) from a microcontroller (a 3.3V powered TI MSP430) using an opamp and a logic-level mosfet; with a low Vgs(th) of either 2.5V or 4.5V.
    I'll be using a 2KHZ PWM signal from the MCU. However the peltier requires a PWM signal with below 10% ripple, so i'm targeting a 5% ripple. ( which should be about 0.05x3.3 = about 165mV p-p ripple, right ?)
    The signal from the mcu is to be passed through an RC low-pass filter to achieve a near DC signal with the above 5% ripple, but.. i'm not sure of what RC values to use. Please any help here would be appreciated ?

    I intend to amplifer the filtered signal to around 10V, so i was wondering what final voltage would be outputed after this RC filtering ?

    thank you for any responses.
     
  2. Sensacell

    Well-Known Member

    Jun 19, 2012
    1,124
    266
    So, you want to create a DC control voltage from a PWM source to drive a linear power stage that drives the peltier? am I reading you correctly?

    This sounds rather inefficient from a power point of view?

    A peltier module obviously has a long thermal time constant, (unless it's super tiny?) 2 Khz is a super high PWM frequency for this application, if you filtered the PWM with an RC filter with a 1 Hz time constant you would have almost no ripple...

    I am not sure why the peltier cares about ripple?

    Most designs I have seen use a PWM switching stage to drive the peltier through an LC filter, like a 'class D' audio amp, very power efficient and compact design.

    The drive can be either unipolar (cooling only) or bipolar. (heat and cool)
     
  3. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
  4. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    the heat load i'm trying to remove is less than 1W. 400mW to be exact.
     
  5. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    the peltier is 40mmx40mm in size.
     
  6. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    i already have the temperature control part sorted out and others. my issue is with the switching of that peltier to provide the cooling, via the mosfet,opamp,filter,mcu path.
    The peltier is rated 15.5V,4.4A, Qmax 43.3W. Giving that i'm moving less than 1W thereabout, i don't intend to run the peltier at it's maximum. I intially intend supplying it 10V,3A from the Buck-regulator, but, from the curves on its datasheet ( http://www.tellurex.com/pdf/C2-40-1504-Specifications.pdf ) supplying it 1A, 4.0-4.2V seems to be sufficient enough for moving 1W of load.
     
  7. Sensacell

    Well-Known Member

    Jun 19, 2012
    1,124
    266
    The Tellurex data sheet says nothing about ripple being a problem, and I cannot see any reason why it would be.

    Why not drive the FET with your PWM signal directly? Use a level shifter to get a good solid gate drive. This would yield good efficiency if you are powering the circuit from a switcher. A tiny FET could do the job- no heatsink required.

    You already said you will drive the device with a lower voltage so regulating peak current is not an issue.
     
  8. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
  9. shortbus

    AAC Fanatic!

    Sep 30, 2009
    4,000
    1,512
    You should be aware that the mosfet you've chosen is NOT a logic level component. To fully turn on it takes 10V Vgs. The threshold voltage is actually the turn-off voltage of a mosfet, the point at where it is just barely conducting. Always use the Vgs voltage when choosing a mosfet.
     
  10. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    which is why there would be an op-amp in my circuit (after the filtering) to amplify the mcu voltage to around 10V.
     
  11. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    uh.... your definition of 'logic level mosfet' being ?
     
  12. shortbus

    AAC Fanatic!

    Sep 30, 2009
    4,000
    1,512
    Logic level mosfets have a Vgs of 3.3 to 5 volts. Why use an opamp for this when a mosfet driver is made for it? A mosfet driver will not only turn the mosfet on fast it will also turn it off fast, both of which are important for efficient mosfet switching.
     
  13. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    okay, so if i decided to stick to the above setup of mcu -> filter -> op-amp -> mosfet -> peltier , and if intend to use the filter; what RC lowpass filter would i need to get the mcu's 3.3Vpp 2khz pwm signal with a 5% ripple ?? and are there formulas for calculating it ?
     
  14. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    @Sensacell,
    so assuming "ripple" wasn't an issue, and i just wanted to pass the 2KHZ pwm signal and remove the higher frequencies leaving the 2khz pwm intact, what filter values would be appropiate ?
     
  15. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    if an RC filter was used that'd pretty much give a linear DC signal... probably not what i intended. the mosfet and peltier WOULD need the PWM. Hmm...
    @Sensacell, regarding that " mcu -> level_shifter -> mosfet" route, could you suggest a suitable schematic ?
     
  16. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    alternatively if i did away with the RC filter and just put a smoothing capacitor instead , would that be better... so like: mcu -> capacitor across to ground -> opamp -> mosfet -> peltier. ? if this was the case, what type&value of capacitor would be appropriate ??
     
  17. SoulBlade

    Thread Starter Member

    Sep 17, 2012
    37
    0
    given the attached image, which would be more doable: using the muc -> filter->opamp -> mosfet -> peltier route Or using the mcu -> level shifter -> mosfet -> peltier route ??? :confused:
     
  18. crutschow

    Expert

    Mar 14, 2008
    12,973
    3,219
    To perhaps clarify, a Logic-level MOSFET requires a Vgs of only 3.3 to 5 volts to full turn-on the MOSFET (look for the ON resistance value test conditions in the data sheet). That's not the same as the Vgs threshold voltage (the point where the MOSFET just starts to turn on) which is much lower.
     
  19. crutschow

    Expert

    Mar 14, 2008
    12,973
    3,219
    If you want a linear voltage to the Peltier (with poor efficiency) than feed the PWM signal to an RC filter (not just C), then to an op amp driver and then to the Peltier.

    If you want best efficiency with low ripple then feed the PWM signal to a driver switch and then filter the output with an inductor and a flyback diode (cathode to ground in front of the inductor).
     
  20. Sensacell

    Well-Known Member

    Jun 19, 2012
    1,124
    266
    The circuit you originally linked to with the opamp will work, but it will require a drive transistor with heatsink to dissipate substantial power. It also has another hidden flaw- the transfer function will be non-linear, because it relies on the FET's gate voltage / current relationship which is not linear. This could bugger your control algorithm a bit.

    Just drive the peltier directly, using the FET as a switch to ground, no 'filtering' is necessary- the thermal time constant of the peltier unit is your filter- and it's probably in the range of hundreds of milliseconds at the fastest.
    The transfer function will be 100% linear.

    You must however make sure, (as we all keep mentioning) to drive the gate with a 0 to at least 10 V drive to ensure that it is fully turned on- otherwise it will dissipate power and the circuit will perform poorly.

    There are FETs with lower gate drive levels, but 3.3V is really low- better to drive it with a higher voltage.
     
Loading...