PWM on shut-down pin? (gate driver switching strategy)

[webgiorgio]

I plan to make a mosfet H bridge with two IRS2004 gate drivers and 4 N-channel mosfets, to drive a 24V 10A DC motor.
The gate driver has two inputs, IN and SD (shut down).
https://www.infineon.com/dgdl/irs2004pbf.pdf?fileId=5546d462533600a401535675b86b2782

Let's say that I PWM only on the first leg, and keep the S4 closed and S3 open.
Schematic attached or https://upload.wikimedia.org/wikipedia/commons/thumb/d/d4/H_bridge.svg/1200px-H_bridge.svg.png

Case 1) If I PWM on the pin IN I always have one switch closed and the other open. (the gate driver takes care of the dead time)

Case 2) No, I put pin IN high, and I PWM on the pin SD. I have the advantage that the bottom mosfet S2 is always open, and the upper mosfet S1 does the open/close. As only S1 is switching in this case, I have lower switching losses in S2 compared to the case 1). However, the return path for the current is the diode in parallel with S2. So, the "current*0.7V*time" is probably more energy than switching the S2 mosfet.

So, on which gate driver pin is it best to attach the PWM signal?

[Siwastaja]

Switch it fully with minimum safe amount of deadtime.

At 24V,10A, your freewheeling losses in the diode are likely going to be so massive they are enough to blow up the MOSFET without proper cooling. If you really need to do this, use an external schottky.

Think about running at high torque, low speed for any reason, so that the motor voltage and current are, say, at 2V, 15A (input could be, for example, 30V at 1A). In this example, the duty cycle would be 7%; the diode conducts at 15A for 93% of the time, producing losses around 10-15W (depending on the exact Vf of the diode at said current). Compare this to the synchronously rectified Rds(on) losses if you switch it on.

Disabling both FETs is used to make the motor freewheel.

Note that you can make the motor regenerate by accident or on purpose, basically boosting to your DC link with almost no voltage limit. Use enough bulk capacitance to limit the voltage rise time on your input DC bus, then use voltage measurement to shut down the gate driver to let the motor freewheel in case the voltage rises near your max ratings. Letting it freewheel prevents boost operation, and you'd need to externally turn the motor very fast (faster than what you get when you apply 100% your input voltage to the motor) to still increase the bus voltage (through the body diodes).

Using a microcontroller ADC / Analog watchdog / analog comparator is fine for this protection.

Also consider a current control loop. A 24V/10A motor sounds big enough it's not doable without current sense, or would either require massively oversized components, and/or careful software tuning and still be prone to blowing up. The current limit doesn't need to be super fast, something like a few tens of kHz bandwidth is going to be enough.

Despite common misbelief, a motor is not a voltage-controlled device. It's like an LED (with the twist that the equivalent "voltage drop" depends on the RPM) - you need to limit the current. Note that your 24V,10A motor likely takes at least 50A, possibly 100A on stall if you just apply a voltage to it, only limited by resistance in the system. Still it doesn't produce much more torque than what it would at more sane currents, because the iron saturates. Hence, you need active current limiting circuit, which gives you:
1) Torque control
2) Protection
3) Good efficiency when running at high torques

[webgiorgio]

Hi, thanks, I will do more reasoning following your advices.

I have a i2c current and voltage monitor IC, which has a programmable output "ALERT", open collector, which I plan to use to drive low the "Shut down" pin of both gate drivers if either voltage or current goes above programmed limit.
For temperature protection the idea is to use a 8A 30V Polyfuse bended over the D-PAK mosfets.

[Siwastaja]

The I2C current monitor is unlikely to be fast enough for current control, especially if it's located on the DC link before the bulk capacitance. Although, it could be workable... Depends.

The polyfuse bended over the D-PAK mosfets cannot protect them from anything - it won't couple thermally, and even then the trip temperature is likely too high. If you are worried about the cooling (i.e., long term heating, where the time constant is in range of at least tens of seconds), use an actual temperature sensor (an NTC will do fine) and try to thermally couple it with what actually heats up - the D-PAK drain tabs. This allows you to set proper temperature limits in your software.