0

I'm creating a dead simple motor driver:

schematic

simulate this circuit – Schematic created using CircuitLab

I apologize for any mistakes in the schematic, I tried to portray the motor as it's equivalent components in the editor but I don't really know what the actual values for them should be. It is a 36V ~25A motor.

I'm planning the frequency of the PWM signal to be 22kHz.

I learned that there come a bunch of problems when you start switching FETs fast such as voltage spikes, increased current consumption for energizing the gates, something about capacitance between gate and source which I don't fully understand and possibly more.

To tackle all of those I should use a MOSFET driver IC. I found TC4420 which has a continuous output current of about 1.5A which should be well above what those gates will waste. Is it possible to use a single TC4420 IC to control all 8 FETs?

php_nub_qq
  • 880
  • 8
  • 24
  • 1
    Look at the data sheet for your MOSFETs and find the "gate charge" for your conditions. Then determine the current you need to supply all that charge in the time available, some small fraction of the 22 kHz period. – electrogas Jun 25 '20 at 01:49
  • @electrogas thank you for the pointers. It is apparently more complicated than I anticipated. So I might be wrong but looking at [fig. 6 on page 4](http://www.irf.com/product-info/datasheets/data/irlz44n.pdf) I determine that total gate charge will be 35nC. The time in which the FET must switch is `1 / 22000 = ~50 us` but 50 micro seconds is the entire length of the period, so I'm guessing the switching needs to be faster than that, say 5us? In which case I'm a little stuck. Looking at [this answer](https://electronics.stackexchange.com/a/20511/164399) I'm understanding that ... – php_nub_qq Jun 25 '20 at 17:07
  • ... the switching will happen in about `2.3 * R * C`, C I calculate as `35nC / 5V = 7nF`. I currently have no resistors on the gates but let's just pick an arbitrary 1k value, switching time becomes `2.3 * 1000 * 0.000000007 = 16 us`. However in this calculation I don't see current anywhere and this confuses me further.. – php_nub_qq Jun 25 '20 at 17:11
  • 1
    Just in case you hadn't realised... Your Diode is very likely to suffer more than the MOSFETs. It may be rated at 30A but when it's taking the freewheel current it will be losing the full motor current x the body diode drop, so ballpark 25W. And you've only got one of them. The model also is not a type that likes being switched at 22khz. Just a heads up :) – David Molony Dec 25 '22 at 00:18
  • @DavidMolony can I use two [MBR1645](https://www.onsemi.com/pdf/datasheet/mbr1635-d.pdf)s in parallel? I'm having a very hard time finding a *cheap* appropriate diode for this job.. – php_nub_qq Jan 22 '23 at 23:46
  • 1
    You can use multiple in parallel, though they might not current share especially well. One sensible option is to use a MOSFET as the diode, with gate and source tied together. Try a TOLL MOSFET, they can handle crazy current, and you get maybe 0.8V on 20A... 16W power. You will still need a heatsink... Really what you want is to make a half bridge with an actual driver and run synchronous PWM with dead time. – David Molony Jan 25 '23 at 18:15
  • @DavidMolony isn't a mosfet going to conduct through it's internal diode? I put two 1645 diodes in parallel for now, didn't put heatsinks though, gonna pray. Will be able to test it out this week, light a candle for me. – php_nub_qq Jan 25 '23 at 20:48
  • 1
    That is the point, use a MOSFET conducting through the internal diode as the diode. But you should be OK with two 1645 diodes. – David Molony Jan 27 '23 at 13:58

1 Answers1

3

So current is charge/time, and you need to transfer 35nC in whatever the switching time is. Now picking 1/10 the the period probably isn't good enough, because you want the interval during which the switching takes place to be short because this is the time when the current flowing through your device and voltage across the device are both large, hence power dissipation. From the looks of your data sheet you might want to shoot for a switching time of about 100 ns, so then your gate current will want to be about 35/100 or about .35 amps. You could probably do it with one of your TC4420 drivers since its peak current rating is up there around 7 amps I believe. I think you should put separate series resistors in each gate line, perhaps 2 ohms, to equalize the currents.

electrogas
  • 661
  • 4
  • 8