Oddbean new post about login | logout @f3ad697a I see two issues: One is that when the PWM MOSFET is off, the channel MOSFETs potentially can see a high negative V_GS. The other is that to reliably switch the channel MOSFETs, you need a fairly high voltage on their gates since you want them to be hard on even while the PWM MOSFET is still transitioning through its linear region.
@f3ad697a The negative V_GS issue can be solved by turning off the channel MOSFETs with a resistor between gate and source. Since they presumably don't need to switch very often, that being slow shouldn't be a problem. Also it would be possible to switch them during a PWM "off" period.
@7f73a7d1 the upper MOSFETs would be driven by a CMOS gate so the resistor probably wouldn't do anything (it's low impedance 0V or 5V, never high-Z). -5V Vgs on the upper MOSFETs is the bit I missed. I'm now trying to work out if that's actually a problem in practice. they switch on and off in the order of 120Hz at most, and are only handling tens of mA each. the Vgs range is spec'd for ±12V so seems ok?
@f3ad697a I can see a possible issue with this solution though: If the PWM MOSFET has a low-impedance driver and is doing very hard edges, when it switches from off to on, the channel MOSFET's gates will momentarily stay at their old voltage, since they are driven low through high-impedance resistors to the common node, while their sources dive low along with the common node. This leads to a momentary positive V_GS that could possibly turn them on for a bit.
@f3ad697a One last thing I can imagine would be that when the three loads have different voltage drop, like RGB LEDs, then while the PWM is turned off, I can see some of the body diodes potentially be forward-biased momentarily. At PWM speeds below RF that shouldn't matter though.
@7f73a7d1 in this case they're all the same LED, so should be fine, but that is a good point.