It's a permanent magnet. (nice chip btw, but you maybe don't need that fancy). The main thing is you have to ramp speed slow, so as not to get ahead of it.
The fancier controllers look at the phase of the current to ensure they don't do stupid stuff (say, the original Pfeiffer controller). You can tell if you've lost lock that way.
If you do pwm, you don't need the power supply step voltage thing John used, but things get trickier. Obviously you have to know how to set/change the pulse width in that case.
If I were doing this from scratch, I'd just use one of the SOC type microchips and do it all in software driving pwm outputs. I could probably find one that would do 3 phase
sort of naturally, or run one pwm at 3x speed and gate the output around to the motor phases.
The thing you want to sweat is what happens in an inrush accident -- some sort of current monitoring should save you there. Continuing full power drive at the wrong
frequency will make a lot of things hot -- and may even wreck the magnets. This can be tricky to measure, as what you are interested in is an indication that the current on this phase is different from cycle to cycle -- beating in frequency. When you see that, it's time to shut down or do something smarter than whatever you were doing.
Pfeiffer uses the free wheeling diodes to power the controller long enough to vent and do smart shutdown when the mains power fails. It takes a long time to run down when
the vacuum is good.
With a microprocessor, one might look at the signal coming back from the pump once in awhile with the drives shut down to see if you're in sync. One or two roundy rounds with no drive won't hurt a thing.
So at the start, the tradeoff is basically do you want the simplest thing that can work in ideal conditions, or something that's a little more robust.
I find that it's pretty rare to need all the pump motor has in it. Even during initial pumpdown, it gets out of that situation quite fast. In my system, the turbo power draw controls the forepump, and it all kind of works out via a current limit on the turbo. Basically, it doesn't spin to the point where it can draw excess power until the vacuum is already fairly good, and then it can't draw excess power because there's no drag on it. Both start together, but until the roughing pump gets it down some, the turbo just doesn't spin real fast due to the current limit.
I don't know if Pfeiffer does the "check the phase while dropping drive" thing, but that's how I would do it -- easier that way than measuring a current phase angle to see if we are leading or
lagging the rotor.