Connect with us

Mosfet switching losses

Discussion in 'Power Electronics' started by Rixen, Oct 21, 2018.

  1. Rixen

    Rixen

    98
    22
    Feb 16, 2016
    Hi all,

    Some of you know i've been working on this induction heater for quite a while and I got it working working some time ago, yay!
    (Video if anyones interested..)

    [​IMG]

    Now, it's been powered from my isolated transformer in the background, that can only supply about 3A, to get around that I was switching the circuit abit above resonance, to bring the current down.. I want to move away from doing that, but before I do, I want to make sure I have the correct heatsinks for my switches and diodes..

    That's my circuit as it is right now.
    [​IMG]

    So, im trying to find the power losses in the switches M1,M2 and the diodes D1,D2.
    I've googled and googled on how to do this, and I understand 'where' my losses are, but I dont understand the math behind it, let me try explain where im comming from..

    Since P = U*I, I have to know the voltage across the FET (Vds) at any given time and multiply with the current going through it, but there is a time component here that I dont understand how to work with, the switching losses are at 10%-90% Vds, so I need to somehow find the losses in the FET when it's switching from 10-90% (and the other way around) and then I think(?) multiply with the switching frequency, to get the power loss in seconds..

    Then there's the conduction losses which im thinking would be Rds_on * current^2 * duty cycle*fsw..

    The inverter is supplied by rectified mains, about 325Vdc where I am, the inverter has to supply about 13.79A, (lets round that up to 14A) to the RLC circuit, there's 165Vdc in the middle of the FET's, so about 2.28kW total.

    Datasheets:
    Mosfet: http://ixapps.ixys.com/Datasheet/DS100310B(IXFH-FT-FQ50N60P3).pdf
    Diodes: http://www.onsemi.com/pub/Collateral/MUR1520-D.PDF

    I hope it's clear what I want to do, if it's not, I apologize, I will be happy to try and clear it up.. I hope you guys can help me out, if any additional information is required I will be happy to try and provide it. ;)
     
    Last edited: Oct 21, 2018
  2. Harald Kapp

    Harald Kapp Moderator Moderator

    8,686
    1,691
    Nov 17, 2011
    To manually calculate the power dissipation requires you to express voltages and currents as functions of time and calculate RMS power from these time-dependent values. Usually some approximations (linearizations) are made to simplify this process.

    Since you have already modeled your converter in LTSPICE, you can let LTSPICE do the work for you. See this explanation. Add a reasonable safety margin to prevent your real world circuit from blowing up.
     
  3. Rixen

    Rixen

    98
    22
    Feb 16, 2016
  4. Rixen

    Rixen

    98
    22
    Feb 16, 2016
    Or I could be using the PULSE source wrong, either way..

    Heres the top gate signal.

    [​IMG]

    Top PULSE source settings
    [​IMG]

    Heres the bottom FET gate

    [​IMG]

    Bottom PULSE source settings

    [​IMG]

    Top and bottom pulse source..

    [​IMG]


    Edit: Damn, the PULSE is wrong now, gotta try fix that ..
    Edit 2: Fixed, delay was 12u not 6u. :)
     
    Last edited: Oct 21, 2018
  5. Harald Kapp

    Harald Kapp Moderator Moderator

    8,686
    1,691
    Nov 17, 2011
    Hint: Use differential voltage measurement to display V4 and V3 to see the 'real' gate-source voltages. Measuring only V(n001) and V(n008) as in your waveform plot gives you voltages referenced to GND. While this is o.k. for V4, as '-' is connected to GND, this will not give you the gate drive voltaeg V3 as '-' of V3 is not connected to GND. Differential measurement will take care of that.
     
  6. Rixen

    Rixen

    98
    22
    Feb 16, 2016
    Hi,

    I know that. This is kind of my first bout with LTspice and I couldnt find a differential 'probe', I think it still does kind of show what I wanted to see. :)
     
Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.