Keep in mind that Q0 is gainscheduled out almost immediately when lde gets small (at least in ME7.1), so likely most changes will be in Q2 or Q1

Generally, I leave Q1 alone; if DRL is calibrated right, it shouldn't need much adjustment.

Q2 is usually the term that needs the most attention.

I'll have a better look at those Basano writings.
Ability to log the variables would make a massive difference.

The trouble is that I cannot really find the thing that trips the oscillations.
There are massive variations in the output duty between the different rpms eventhou the commanded P/I/D and duty are identical.
At some points the actual duty follows KFLDIMX (through KFLDRL) perfectly, sometimes it is off by 20% in a "stable state".

I didn't even notice that the F23T has an enlarged wastegate flap compared to K04 
That alone will make a massive difference to the PID programming requirements.

Start with LDRL tuning, then LDIMX/LDDI etc (like phila says), and finally Q1/Q2.

New project ahead, could make a new thread if needed, but it seems like most of the MED9.1 people are in here.

I've got a 2008 VW GTI Mk 5. Have read out the ECU from my car, and learned that it has 1K0907115S VW software ID (ECU ID 8P0907115B).

Started mapping from other reference versions, since there is no map for the 115S, mostly from what majorahole posted last month of his 115H, and this one with matching image / complete map.

Hoping that someone could take a look at what I've got so far and give some confirmation I'm not missing something with this activity. Especially LAMFA, why does it get richer then leaner at 80% then get richer again (pic attached)?

And I'm also curious if anyone around has addresses for this ECU version, or a suggestion on how I can find the smaller maps that I can't search for by cross-referencing other maps... namely:
NMAX
VMAXNB
KLMIMAX
NMAXGA
KFFRTMKI

Will log from stock ECU soon, definitely before flashing any updates.

I've had better results with LAMFA and keeping it out of bts. so that bts will still be a safety measure if needed, not the target

After copying all of the differing PID related maps (not just the KFLDRL, KFLDIMX & KFLDRQ1-3) from S3 binary the small oscillations are mostly gone.
The KFLDIMX/KFLDRL behavior changed completely after that and the actual duty is even further away from the one KFLDIMX/KFLDRL commands.

I still need to improve the response time the ECU takes to stabilize the duty after disturbance (initial boost spike).
Also there seems to be some more undershoot (droop) at the higher revs.

Increasing the "D" (KFLDRQ2) should cover the response time and increasing the "P" (KFLDRQ1) should fix the droop, correct?

I don't think this has come up yet on this thread. If it has, forgive me.

One of the easiest ways currently to request more boost on MED9.1 over the hard 2550mbar absolute pressure limit is to install a higher reading MAP sensor but scale it incorrectly, so that 2.55 bar is calibrated at the (actual) maximum absolute pressure you'll want to see from the sensor when under boost. My question is how will this *not* affect specified load calculation? It seems like you would need to request less load in the ECU to achieve the same actual engine output when scaling boost by this method. Is this true or do I not understand correctly?