Config check, hunting idle and injector offset confirmation

[jrussell]

You typically want to set your ref dc idle value to give you an idle 150-200 rpm HIGHER than target. And let the PID correct down instead of correcting up. That gives better stall protection.

Other than that, good technique to solve the problem, and thanks for sharing your process. Very nice.

[secretmachine]

ohhh? that process does make sense. i had come across some info stating to make the car idle just slightly below target. this is without any interaction with the closed loop idle, meaning the car would idle like this with the power disconnected from the PWM idle motor. its "limp home" mode so to speak. i'll give a slight higher ref dc a try and see what the car does. thanks for the advice!

[fphil]

this is without any interaction with the closed loop idle, meaning the car would idle like this with the power disconnected from the PWM idle motor

Ths is not the same since additional air is given by the idle air valve refernce position even when P=I=D=0.
D action should may not be necessary. Another general quick way for PI tuning for a first huess is to set I=0, increase P up to the value where system starts to oscillate,  divide this value by 2 to set P. and finally adjust I to get zero error.  If the closed loop system is not stable enough then use D. Be aware that in our case there is a sliding target (IAC target RPM) which slows the rally down to the target in order to prevent too large RPM undershoot

[secretmachine]

We're talking about different things then. I set the car mechanically to idle just below target without any vems interaction. I used an air bypass built into the bosch three pin IAC while the valve had a closed window with the connector removed. I figured that way, the car cannot stall as mechanically it has enough air to keep it going without closed or open loop idle control. Then I set the ref curve, then pid=0, then tuned the pid. The tactic you are mentioning did not work for me in particular, the tuning of P until oscillation then dividing it by 2. My P number until oscillation was very small, so I instead went for the flattest waveform, then trimmed it with I and D. This three pin is a dual solenoid and reacts very fast, so I had to slow it waaaayy down with a lot of derivative gain. Here's a screenshot of the final outcome for stable idle in my application. I also noticed my idle motor, for some strange reason, liked settings that where divisible by 3.