The microcontroller circuit shown in the figure initially behaved very strangely—rebooting the microcontroller at odd moments, and often latching one microcontroller output high—until resistors R1 and R2 were installed. Investigation showed that a brief negative pulse, caused by inductive kickback, was being capacitively coupled from drain to gate in the FETs and creating CMOS latchup in the microcontroller.
The circuit shown is part of a twophase variable-frequency drive for an astronomical telescope. Two-phase drive, with its staircase-like approximation to a sine wave, is more efficient than square-wave drive. If the pulse width is held constant, the rms voltage is proportional to the frequency, exactly as a synchronous motor requires (Electronic Design, March 3, 1997, p. 179, and M. Covington, “Astrophotography for the Amateur,” 2nd ed., Appendix C).
Typically, FETs driving inductive loads would be protected by diodes from drain to source. That’s not possible in this case, because when one of the FETs switches on, its drain goes to 0 V; the other drain goes to +24 V by normal transformer action, not an inductive spike. A protective diode would conduct during that part of the cycle, which wastes power.
Instead, inductive spikes are dissipated by the avalanche diodes inside the FETs. However, these diodes don’t act instantly. When a FET switches off, its drain is driven negative by an inductive spike, but it takes the diode at least 0.2 ms (according to the data sheet) to recover from the reverse bias and begin conducting. In the mean time, a strong negative pulse (measured at -2 V) is coupled to the gate through the FET’s internal capacitance of several hundred picofarads.
This pulse is apparently capable of throwing the PIC microcontroller into CMOS latchup. The problem disappeared when resistors R1 and R2 were installed to limit the current.
Note that with 5-V gate drive, the FET must be an IRL530 (“logiclevel”), not an IRF530 or the like. The published threshold voltage of the IRF series is 4 V, but that refers only to the onset of conduction; saturation requires at least 8 V on the gate. The IRL series saturates at about 4 V. The symptom of incomplete turn-on is that the transistors will run hot. In the correctly functioning circuit, the heat dissipated by the IRL530s is barely noticeable by touch and thus doesn’t require a heat sink.