Although the vacuum pump doesn't run all the time anymore, I've noticed that it does run more often when it's colder outside. So I installed a gauge and valve near the pump so that I can isolate the pump from the rest of the system. Like this:
The pump stayed on until the gauge read ~23 in Hg. As soon as the pump kicked off the gauge steadily dropped to ~20 in Hg (took about 30 seconds), where the pump again kicked on to bring it up to 23. Once it reached 23 and the pumped kicked off I shut the valve to isolate the pump from the rest of the system. The gauge stayed rock steady at 23. Opening the valve made the gauge dropped immediately. So, there's something going on in the pump causing the vacuum to leak. I think the answer is to put a check valve at the pump intake.
Also, some progress on getting the tachometer to work with the DMOC "speedometer" signal. A search of the EVAlbum showed that there was one other person who used my motor/controller in a Saturn. Oddly enough, it's another Greg, though he uses the extra g at the end of his name (Gregg Witmer). Although he hasn't gotten his tach to work to his satisfaction, he sent a schematic of a circuit that he thought would. Using that schematic as a starting point, a nice EE friend of mine and I succeeded in getting the tach needle on my "spare" instrument cluster to move. Here it is at 1000 rpm equivalent:
The tachometer is designed to accept 6 high-to-low transitions in the square wave from the PCM for every engine revolution, so 1000 rpm on the tach means 6000 transitions per minute from the PCM, or 100 Hz. Gregg's circuit actually uses a flip-flop chip to create a square wave from the pulse that the DMOC puts out, so the DMOC will need to put out twice as many pulses. With our function generator at 1.58 kHz the needle moved up to nearly 8000 rpm, as expected. Yes!
The tachometer draws more power than we expected, however. The 1/2 watt resistor we used to pull the signal down to ground was getting mighty warm! A quick calculation showed that we were actually pulling 250mA through it, which translates to 3 watts. Oops. A 5 watt MOSFET worked fine, though. Next step is to try the circuit with the actual DMOC output to see if it'll drive the tach. Then convert the breadboard circuit to something that'll be able to live in the car.