Not sure whether you guys deem it worthwhile but I'm thinking of making some very small adjustments to the derivative function of the PID to try and get the little overshoot (when actual meets requested) in check. the s4 wiki leads me to believe i'll need to slightly raise the values I  KFLDRQ2 to cure it.

One axis I have RPM and the other is hPa which from what I've read is 'ground pressure' which goes from 100 -700hPa. Am I right in saying I should only need to tweak the values for the 700hPa row at the say 2 rpm figures where I am seeing the overshoot. Finally, is there a quick way to calculate how much I should increase them by. I guess i could bump them up a small amount and do lots of logging and revisions but if theres another way that'll mean i only have to do half the amount of revisions that would be good to know 

IMX seemed good for sorting undershoot, I bumped up the values ever so slightly at certain points and got the boost closer to requested. However when i tried dropping the values around the area of overshoot it didn't seem to do much but shift the overshoot along in the rev range.

I'll have another go though, maybe I didn't reduce the values in IMX enough to see a proper change.

I've made some changes to IMX and removed a fair bit of the undershoot and feel fairly happy now - though i need to drop ldrxn a tad maybe since I think even with 95% DC i'll not hit the requested boost higher up in the rev range. It's not severely out though.

I'm wondering if its maybe time for me to adjust the differential term of the PID to tame the overshoot and ripple (i think that's what the subsequent undershoot is called) I've attached a log and some pictures lde included in the log.

So if tweaking the PID may cure it, I'd be needing to increase some values in KFLDRQ2. Using lde to help me determine which row or rows i need to change values in, as well as rpm - would I be right making these changes at the rpm's just before the overshoot occurs rather than at the RPM of overshoot as presumably that's too late for me to see the desired changes occur?

In dummy terms do I need to reduce the speed at which I am approaching my set-point (desired boost) the closer I get to it, to reduce my error? and that could be achieved by increasing the D term? Please excuse my lack of correct terminology, I'm just trying to get a half decent grasp of why I might tweak KFLDRQ2 rather than just doing it because.

log attached, the IMX i'm currently using and a screen shot of boost and lde.

and perhaps increase requested from 2000 rpm to 3000 to make things happier

Thanks Nyet for your previous comment, I pulled a fair amount out of IMX and now have no overshoot. Got a slight dip but maybe that'll be fixed if I increase requested in that area?

If I bump up IMX a tad where i see the dip and increase requested hopefully i'll end up with a fairly smooth looking line.

Can you explain why? For learning purposes

To make sure PID is stable during spool - P should be big enough to provide 95% wgdc, and I should have had had enough time for full windup so it is riding IMX before peak boost.


See above regarding integrator windup. Also don't forget there are other terms (P and D) and the LDRL table as well.

Also, this isn't really the place for me to explain integrator windup; it would take me forever, and I don't know your math background. You'll need to do your own learning regarding PIDs.

As an exercise, though, under what conditions would you expect wgdc to EXACTLY follow IMX?

Not even close

I am not sure how much time i want to spend here

Anyway:

P result + I result + D result goes into ldrl, and comes out as wgdc.

I result is limited by IMX if the integrator is completely wound up.

The rest of the answer to my quiz follows from that reasoning, but I fear it would take me way too long to get you to understand PIDs. You really need to do some independent learning first. It would take me way too much work to teach you everything you need to know.