I just wanted to bring KR based AFR tuning to a conclusion as it proved spectacularly successful in my case.

My observations:

I suspect that factory S4 M-Box tune (other S4 tunes might be the same) contains unintentional "test" version of M-box from factory or Audi engineers went berserk and tuned the car based solely on BTS + ATR basis leaving KR AFR corrections essentially disabled. The bit0 for CWLAMFAW which governs what corrections are used for KFLAMKRL (Lambda corrections during knock), is set to opposite value than all the other tunes for turbo audis effectively rendering this Lambda correction path ineffective due to the way knock input for this table is calculated.

At the same time, BTS is overly eager to enrich starting at 500 deg C, which is implies that BTS kicks in as soon as your car gets any boost. Even Stock configuration S4 will run deep into BTS lambda corrections. Stage 3 S4 without modified BTS table will run 0.63 lambda (9.2AFR!!!!) at above 6000rpms at full load.

I've seen my engine reaching 500C anytime I was just spooling up @3.5k to reach fulll boost ramping from there to 800+ and end of run. Now keep in mind that underscaled MAF configurations will not see temps this high since load is lower on these cars and therefore the model temp for BTS will be much lower.

It is noteworthy to add that RS4 tune has the temp threshold much higher (700C or higher) and its BTS table most enriched point is meager 13.0 odd AFR.

Since RS4 was tuned with much less aggressive BTS and it has full blown KR Lambda Correction enabled, it leads me to believe that this is indeed the way S4 should be tuned as well.





What needs to be done to S4 M-Box:

1)  Change bit0 CWLAMFAW @ 0x18ECB to "0" (not set)
2)  Neuter BTS by raising TABGBTS to some high value. I actually just set it to a value that works for my case which is 920. I reached 940C BTS temp at the end of 4th gear run (meth disabled) running at 12ish AFR where ATR (EGT sensors) sttarted kicking in indicating around 15-20C higher temp than BTS on bank1.
3) Adjust KFLBTS table to be more reasonable. I elected to scale it uniformly so that the higher RPM/Load AFR is 10.9 now.
4) Now the fun part. Adjust KFLAMKRL. This is your new LAMBDA/AFR table for ALL load points. You're going to say "wait a minute, it only drives Lambda values under knock...right?". Yes and no. In stock form yes, with modified KR axis it does not. Simply establish KR Corrections table and shift/modify first row to read "0.0" corrections and all of the sudden this table becomes responsible for all load points. Look up the definition of it under "X Axis" of KFLAMKRL
5) Redefine Load axis the same way. I elected to have 40,60,80l,120,168,191 load points.
6) Tune the data points in KFLAMKRL table. You want to leave lambda at 1.0 up to load of 120 at Kr of 0.0 as you should not see any knock until load of 168 (next load point) at stock AFR (14.7) and if you see any, the current AFR will shift to a row with respective knock corrections on it. You can leave most of the table stock except for 168/191  points and maybe massage 120 a little bit if yo uwant to ramp up nicer than just abrupt drop when you expect knock.

The results are superb.

I did several pulls with Meth injection off and settled on -6.0ish corrections at 11.5 AFR (theoretical and real matching) running FATS of 3.05.

When I enabled meth, I settled at -3.0 corrections while running real AFR of 11.3 and ECU fueling at 11.8 running FATS of 2.81.

It works. Should I run out of meth I will be ok since S4 now corrects AFR based on knock while before it would just die miserable death from detonation if -12 corrections were reached (limit, can be changed). With the new tune I would slip to 0.74 lambda which is pig rich imho.

I will probably massage the values a bit to get better AFR with meth but overall it just works since RS4 is doing it the same way!

Here is the KFLAMKRL table I currently run:

Oh, did you leave LAMFA stock?

This is excellent, great work Julex.

It means bit 5 or 6 is set, 0 is unset since it is an even number. Bit zero means 1 in decimal.

I can tell you it would be bit 4, but the effect of that I don't know off the top of my head.

My mind = blown.  As rob asked above, this theory of KR based AFR tuning would not work well with E85, considering MBT can usually be reached without significant knock throughout the rev range, yeah?

I wonder how long it will be before we start seeing some professional tunes cropping up that utilize this method.

start your windows calculator, punch in 113 and hit "bin" button (XP) or switch to programmer mode (Windows 7), enter 113 and hit "bin" again. It will show you binary representation (bits) 113 represents which is 01110001. This can also be viewed in WINOls by hitting "111" icon on top when in a table of interest. They are reversed order - bits: 76543210. Bit0 is set in "R" to 1.

I checked RS4 and euro S4 bins and bit0 is set to "0" in there. Hence my overstatement, I guess, but then we already determined that US tunes are retarded anyway. I am almost tempted to flash euro bin one of these days and see if I pass emissions.