Buck converter diode

Discussion in 'General Electronics Chat' started by vortmax, May 31, 2014.

  1. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    I'm building an asynchronous buck converter (schematic attached) and am attempting to simulate it in LTSpice just to make sure I have everything working properly. Everything seems to work great, except when I measure the forward current through the diode, it is consistently above 20A (albeit for microseconds at a time).

    Is this just an artifact of LTSpice using interpreting the diode as being ideal, or should I plan for this much current? All the app notes I've read just say to rate the diode for load current + ripple, which in my case would only be ~5A. Do I need a resistor in series with the diode to limit the current? I have yet to see an example buck converter layout with such a resistor.
     
  2. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    I think what you are seeing is the reverse recovery time of the diode. It should be very very short. I think if you use the ideal diode (no name) it should go away. It is real but the time is so short you can usually ignore it in the power calculations unless the frequency is very high. Use a fast diode.
     
  3. crutschow

    Expert

    Mar 14, 2008
    13,042
    3,243
    Is that a Schottky diode?

    Post your .asc file.
     
  4. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    It is a schottkey, but thanks for saying that...I was using a standard diode in spice. I still see the same thing with the shottkey, although the effect is somewhat diminished. My asc is attached
     
  5. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    Ahh, I see. It is running in what is called discontinuous mode because the inductor is to small to store enough energy at such a low frequency. Try changing the inductor to 100uh and the output cap to 4700Ufd.
    Here is a handy calculator for sizing the inductor and capacitor:
    http://www.daycounter.com/Calculators/Switching-Converter-Calculator.phtml
    So you could use .6v for the diode, .1v for the FET, 10khz, 4amps, 50% for inductor ripple, and .5% for voltage ripple.
     
  6. crutschow

    Expert

    Mar 14, 2008
    13,042
    3,243
    Why do you have such small resistors and large capacitors in you simulation circuit as compared to the original circuit?

    To use such a small inductor and avoid the discontinuous mode you can also increase your switching frequency to about 100kHz.
     
  7. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    I would be wary of that gate driver at 100Khz.

    I think it's better <50kHz and increase the inductor again, maybe to 220uH.
     
  8. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    oversight....The intent of this simulation is to pick values for the resistors and caps that make up the gate driver, and I was experimenting with values. When I ran the numbers to pick the inductor, I did use 100kHz as my freq...it looks like I was just off one decimal point when I plugged it into spice.


    I am trying to keep the physical size of the inductor down, and 10 uH seemed to offer the right size package with a switching frequency I could handle (>8 bit resolution with a 30MHz clock). I would like to avoid increasing the size of the inductor as much as possible....any suggestions for how to make the gate driver circuit more robust for this freq? Would I be better off using a dedicated driver (e.g. LTC1693-5)
     
    Last edited: Jun 1, 2014
  9. crutschow

    Expert

    Mar 14, 2008
    13,042
    3,243
    Yes, a driver such as the LTC1693-5 should help you operate at a higher frequency with reasonable efficiency.

    What does 8 bit resolution with a 30MHz clock have to do with the switching frequency of the circuit?
     
  10. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    I'm regulating output voltage using pwm, so a 30 MHz master clock driving a 100kHz pulse train translates to 300 clock ticks per pulse. That means I get 300 discrete pulse widths, or just over 8 bit resolution.

    If I go with a driver like that, what considerations do I need to make in terms of MOSFET gate charge?
     
  11. crutschow

    Expert

    Mar 14, 2008
    13,042
    3,243
    Mainly that the gate driver has adequate current capability to charge and discharge the gate charge sufficiently rapidly to keep switching losses at an acceptable level. Likely anything less than 100ns for the charge and discharge time should be adequate.
     
  12. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    At 100kHz, YES. (Which Crutschow already said).

    Your big problem is that you are using a high side PFET, so the driver has issues with Vin max/variation etc.

    If you decide to go for a FET gate driver IC I would be temped to go for a high-side NFET driver IC. You get to use a cheaper NFET with better Rds-on spec, which means better efficiency and less heat. That also takes care of most of the Vin issues.
     
  13. vortmax

    Thread Starter Member

    Oct 10, 2012
    103
    18
    I thought about a high side N-Fet, but I'm having problems finding a driver with a full time charge pump. Most of them use the high side switching voltage to bootstrap the high side gate voltage... but I want to be able to run this at 100% duty cycle, so the charge pump needs to be independent. I suppose I could just add an external charge pump, but that seems like a lot of extra parts.
     
  14. crutschow

    Expert

    Mar 14, 2008
    13,042
    3,243
    What about one of these?
     
Loading...