Ok, I’ve managed to gather a little more information. The drivers will run on 5V, and ESP32 pins are 5V tolerant, so I can safely probe everything when running on a 5V supply. The mosfet gates are receiving correct PWM signals. High side waveform matches the input pin, low side is inverted.
Probing the mosfet output seems to confirm what the multimeter shows. When high side gate is fully on, output is fully on, otherwise the output is almost fully off but with sporadic blips.
I also tried measuring the gate pins with the multimeter and they seem ok, except the high side gate is somewhat lower than it should be. With 5.09V power supply, 100% duty reads 4.1V, 2/3 duty reads 3.3V, 1/3 duty reads 1.5V. Trying with the 7.7V battery, 100% duty reads 6.7V, 2/3 duty reads 4.7V, 1/3 duty reads 2.3V. Corresponding low side gates are just as they should be, 100% duty 0V, 2/3 duty 2.5V, 1/3 duty 5.1V.
No clue what the problem could be. The mosfets are wired up just like Lepton, and receiving seemingly valid inputs, yet they don’t work.
I think it’s one or the other depending on mode. Inverting mode is INHx pulldown, INLx pullup (both mosfets off if input pins are floating). Which means that I actually could use hi-z/freewheeling if I disable the timer and configure its pins as inputs
But since the pins are externally connected, you get a voltage divider via the external connection?
Ok but this would indicate that the drivers are in fact working? What’s the threshold voltage of your MOSFETs and what happens if you go to 12V gate drive voltage?
Oh! You’re right. I doubt it matters since the signal from the MCU should overwhelm the ~200kOhm pull-up/down, but that does mean they’re in an undefined state at startup until the MCU pins are configured as output, and freewheeling mode is not possible after all. But the cross-conduction protection should prevent any real trouble.
Correct, the drivers seem to be working, but the mosfets are not. The datasheet says RDSon=7.5mOhm at Vgs=4.5V, 5mOhm at Vgs=10V. There’s a graph that I don’t quite understand, but seems to indicate threshold voltage is around 3.5-4V. So the 5V output probing results are somewhat suspect, especially with the high side gate seemingly only receiving 4.1V, but the 7.7V battery should be plenty to get it going. I’ve done a lot of motor testing on Lepton with it.
I tried with a 3S lipo (presently charged to 11.7V) and got similar results. 100% duty reads 11V on the mosfet outputs, lower duty is around 100mV or less. Gate voltages show a similar pattern as before, with high side being a bit lower than expected. High side 100% duty->10.6V, 2/3->7.4V, 1/3->3.7V. Low side 100%->0V, 2/3->3.8V, 1/3->7.7V.
Thanks for all your help so far! Hopefully we’ll get to the bottom of it eventually.
EDIT: Tried setting PWM frequency 1KHz and made no difference. I thought maybe since the mosfets do seem capable of switching on and off, doing so at a slower speed might help.
EDIT2: Next test: Ultra-low frequency. No more timers, just set the pins to output mode and use digitalWrite to switch them on and off, and then probe mosfet output with ESP32. Even at one on-off cycle per second, the output is asymmetrical (more off-time than on-time), and it gets progressively worse until 8Hz there’s almost no on-time left, and 16Hz+ there is none visible at my fairly low resolution of 244us per pixel. Mofset gates are still good, and the output waveform matches the input frequency, it just seems to be very slow at switching on.
Correct me if I’m wrong, but the high side FETs are usually switched via a bootstrap circuit. That often requires external component(s). Maybe they’re too tame to boost the FETs?
The DRV8300D has the bootstrap diodes integrated, only the capacitor is needed externally.
But of course the bootstrap can be a source of problems, esp. at 100% duty cycle.
You can check it by checking the capacitor against GND, it should charge up to Vmot +Vbst every cycle. Using an oscilloscope I mean.
Yeah, 5V sounds like too little but 7.7 is probably ok on 4V threshold.
We’d have to do some math to see how quickly the gate charges up in these conditions.
But if it worked on Lepton for you then probably there’s another issue to discover.
Ok, I have some actual oscilloscope measurements now. PWM frequency is 25KHz and measures correctly now. Apparently I was doing something wrong on my ESP32 when it seemed to be doubled.
Battery 7.7v
100% duty:
Low side gate 0v
High side gate 6v
Bootstrap capacitor 3.5v
Mosfet output 6.4v
2/3 duty:
Low side gate alternates 0v and 8v
High side gate alternates 0v and 7v
Bootstrap capacitor is mostly 7v, with 230ns blips up to 10v
Mosfet output is mostly 0v with 230ns blips up to 3.4v
1/3 duty:
Low side gate alternates 0v and 8v
High side gate alternates 0v and 7v
Bootstrap capacitor is mostly 7v, with 230ns blips up to 13v
Mosfet output is mostly 0v with 230ns blips up to 6.2v
The blips seem to occur during the deadtime window, starting as the high side gate is switched off, and ending just before the low side gate is switched on.
For comparison, I took the same measurements on Lepton:
Low gate alternates 0v and 8v
High gate alternates 0v and 15.5v
Bootstrap capacitor alternates 7.4v and 15.2v
Mosfet output alternates 0 and 7.7V, with small spikes at the transitions to about 8v and -0.3v.
So apparently the bootstrap is not working, but I still don’t know why.
Yeah, this really sounds like an issue with the bootstrap capacity. With the exception of the 100% duty cycle test, which is an edge case, they should charge up to around 15V (2x your VM) each cycle. And they should charge both during the dead-time and the low side on-time, not just during the dead-time.
Are the caps used actually 470nF like in the schematics?
You could try replacing all the 470nF caps with 100nF ones and see if it helps? I wonder if it’s an issue that the total capacitance is 4x that of a single driver? If the drivers are all switching in sync?
But probably you’re testing only with one driver active at the moment?
Results are the same whether one driver is running or all four. I may sacrifice a board to try adding gate resistors (though I’m not terribly confident I can solder to a 0.15mm trace…), and in that case I could try changing to 100nF capacitors while I’m at it.
I’ve pretty much confirmed that the internal diodes are not the problem. I used a jumper wire to connect a diode to battery+, and manually touched it to the bootstrap capacitor terminal. The mosfet output blip remains basically unchanged, aside from a slight drop in amplitude while the diode is in contact.
Transplanted DRV8300 onto a Lepton to see if it would run. Results inconclusive but not encouraging.
I took a Lepton with burnt mosfet and removed that mosfet, the FD6288Q driver, and two of the bootstrap diodes (one left to see if it makes any difference). Removed one driver and mosfets from the Gooser and transplanted them onto the Lepton (after some surgery to the PCB to disconnect the mode and DT pins, which were unused/connected to ground on FD6288Q).
When I first powered on the Lepton, I heard a snap, and the newly transplanted mosfet now has a bubble on it, as they do when burnt. I then discovered I had uploaded a bad .bin, but I don’t think it would have activated any driver inputs so I don’t know what happened.
Probing with the oscilloscope, the driver inputs from the CPU are all good, but outputs to the gates are not. Only the phase with the probably-burnt mosfet is working. The other four gates never change. I can’t find any solder shorts with the multimeter, but wouldn’t be surprised if it’s a soldering problem, or heat damage from reflowing 3 times now (including the initial assembly by JLC). The probably-burnt mosfet acts just like all the ones on Gooser, so it may still be functional. Output mostly 0v, with blips during deadtime, and bootstrap capacitor sitting at battery voltage with matching blips above that.
The last thing I can think to try is adding gate resistors, but I don’t have any on hand so I’ll have to order them. The various gate peak current ratings in the datasheet have quite a wide range (400mA to 2100mA), but I think 500mA should be a safe target. My final battery voltage will be 13.2v (4s LiFePO4), so maybe 25 ohms? With the 7.4v battery I use for testing, that would be 300mA, or 170ns to charge the 51nC gate, which is cutting it a little close on the ~200ns dead time, but probably ok for testing.
