Hi dd,

Do you mean KFLDIMX i.e. what I've set the map to?  Or is there another variable/s I could be logging specifically for that (I-limit)?

I can have a go at logging these, if that's what you're refering to:

ldimn_w
ldimx_w         
ldrki_w         
lditv_w         

ldtv is post linearization.

What does ldimxak_w look like?

That notch in DC at the end of the run looks like a changing adaptation range.

Post ldimxr, ldimxak, ldimx, lditv, ldptv, and ldtvr.

Does it always underboost like that?

Aaargh, my mistake. You are absolutely correct, ldtv_w is definitely post lin! That'll teach me to concerntrate.


Yes, it's very consistent like that. Well-behaved underboosting!

I added some more variables from LDRPID (including all you mentioned) into the ram logger and captured a run this morning.



The red, dark blue and light blue lines are the contributions from the P, I and D terms. In the graph, all the waste gate duty cycle (yellow) comes from the I term at the rpm range I’m interested in (5000 rpm upwards)

P – ldptv_w
I – lditv_w
D – ldrdtv_w
W/G – ldtvr_w



The red and blue lines are desired and actual boost, showing the deviation from 5000 rpm upwards. The green and purple lines are the waste gate duty cycle pre and post linearization (KFLDRL).

Desired Boost – plsol_w
Actual Boost – pvdkds_w
W/G duty cycle pre lin – ldtvr_w
W/G duty cycle post lin – ldtv_w



The red and blue lines are the I-limit (from KFLDIMX) and the actual I value being used by the PID.

I-limit – ldimx_w
Actual I value – lditv_w

Looking at the data and speculating out loud, I’d make these observations (but I’m no expert. I haven’t done this before and I make mistakes all the time. If I’m wrong, don’t be shy to tell me!)

•   No issue with KFLDIMX. I ride the I limit during spool and midrange, but from 3500 rpm upwards, the ceiling is well above the actual I value.
•   KFLDRL (pre and post lin) isn’t doing very much, especially at the range that I am interested in.
•   All the duty cycle is coming from the I term


The attached csv log is pretty comprehensive and has got all those other variables you mentioned. But I didn’t graph them so as to keep things clearer. I’m happy to pull them out if you ask.

Did you log ldimxak?

Is KFLDRL 1:1?

Is the bump up in ldimx related to LDDIMXN  values?

I would make Q1ST slightly more aggressive (increase) and see if it tracks better. If you go too aggressive it will become unstable.

I would also raise dynamic Q0 slightly in the spool region and Q2, but it might be best to save that until after you get this sorted.

Just curious where that odd jump in ldimx is coming from. I'm guessing it was done like that to keep spikes in check down low. Do you spike harder when boosting above 4k RPM?

It may actually be better to start by raising KFLDRL as the response is clearly not linear.

IMO for a constant boost req your pre-lin DC should not be moving much... I should not be moving, unless your req boost is tapering significantly, in which case you should be losing I, not gaining.

let KFLDRL bring up DC, let your PID stay mostly inactive

increasing I generally means your DRL isn't calibrated correctly, unless your req boost is increasing!