actually, I've completely redone the circuit
the valve is ACTUALLY a solenoid which rotates the little vane inside the idle speed controller, so ideally, it's best to have a constant voltage source going to the valve.
So I made one.
What I'm using now is an 8 bit R-2R ladder out of my processor which goes into op amp A from an LM412, with a 2.4x gain. This takes the signal from 5V max to 12V max at the output. Then I stick that output into op amp B on the chip. The output from op amp B goes into a TIP120, and that sources current to the idle valve. For feedback, I use the drain from the transistor going through a 1k resistor into op amp B. This gives a good, strong, variable output, with no current to speak of required from the microprocessor, which avoids the problem of needing enough current to drive the transistor.
At low PWM time from the microprocessor, there wasn't quite enough juice to crank over the transistor until about 60% duty cycle, but with the new setup, it's pretty much perfectly linear from about 10 to 250 or so on the R-2R ladder outputs.
8 bit resolution on the R-2R ladder should do it I think. That gives up to 256 bits of resolution (while the MS only works at 1% increments anyway).
What I have YET to test, however, is how the op amps will respond to the voltage spiking and stuff on the car's 12V rail. What I'll likely do is build a DC-DC regulator that will always put a constant 15V on the positive side of the op amp. That'll make sure that it's linear, too.
I'll draw up some schematics in a minute. I just got home.
EDIT:
Oh, and one reason why I'm using a co-processor is because I'm designing my own version of the megasquirt board, and I want everything to be absolutely perfect for when I finally get around to it. The MS-II main processor just doesn't have enough extra I/O for everything that I plan to do with it, so I'm practicing communicating with the main chip.
Does anyone know if the MS-II chip supports the Two-Wire Interface (aka I2C)?
My plan for the board I'm designing is to use an Atmel AT90CAN128 as a co-processor, talking through the CAN bus, and then any extra stuff I need to add on top of that if that by itself still isn't enough, I can talk to using the TWI, which is fast enough to handle ~4800 baud or so, plenty for basic peripherals.
DOUBLE EDIT:
On the other hand, so long as you run the PWM output fast enough, and you sufficiently filter it, you don't NEED an R-2R ladder. So it WOULD be possible to do this cleanly on the MS. I'll test both tomorrow and see which setup works better.
)
though, i guess you could put up with no idle control while the tuning software was connected it could work? or am i confused?