[Alan Anderson]

Don's explanation is a good overview of the concept, but please don't pay too much attention to the details of what he said. He's essentially correct for some H-switch circuits, but the Victors don't work the way he described. Eugene is correct in that there are four states, corresponding to various combinations of plus voltage, minus voltage, or open circuit on each of the two output terminals.

The Victor circuitry is very fancy. It doesn't use the PWM signal to drive its output directly. Instead, it actually measures the PWM input, turns the pulse width into a number, and uses that number to control the outputs with an internally generated pulse width and direction control signal.

Here's more than you needed to know:

The IFI control system assigns a value from 0-254 (not 255) to the range of pulse widths from 1.0 to 2.0 milliseconds, with 127 corresponding to 1.5 ms. The Victor nominally turns a measured 1.5-2.0 ms into a 0-100 forward value, but the output actually stays at zero for at least the theoretical 0-6 range, and it reaches 100 with an incoming pulse width that is somewhat less than 2.0 ms. It turns 1.0-1.5 milliseconds into a 100-0 reverse value, with the same holding at zero until it reaches 7 or so, and getting to 100 before the pulse width gets down to 1.0 ms. Experiments have shown that the Victor's center value is not precisely 127; it's actually more like 134. So the Victor output isn't symmetrical around the theoretical neutral value of the IFI system.