Preemptive feedback knock correction

wrxt4cy · **Posted:** Mon Jun 11, 2012 3:05 pm

AE5I910V is my rom. I'll give it a test.

I'm comfortable minimizing or zeroing it on an E85 tune since knocking is highly unlikely but with gas tunes, I would love to understand the logic better and possibly be able to tune it to fit the needs of the car.

This is kind of sub-theory hunch I have, but I feel like there could be some logic that exists where the ECU will pull timing using this preemptive method and if it doesn't see knock, then it might not pull it again even if the same conditions are met. I can't tell if it is maybe based on learning between run cycles or if it is a different, undiscovered type of learned timing, similar to FLKC but for the purposes of heavy load change. My hunch comes from noticing that this type of FBKC tends to show up frequently in the 1st or 2nd logs recorded but is normally gone in subsequent logs that are the same conditions as the first. Since logs tend to be taken after a new map is flashed, it could be based on a type of learning that is cleared between flashes. Maybe this isn't true but it might be worth considering while you are hunting around.

Thanks!

td-d · **Posted:** Mon Jun 11, 2012 3:15 pm

Ok - well for AE5I910V the address is D123C - so for Ecuflash add the following to your xml:

Code:

If I get some time, I'll dig out the switches involved, maybe logging them will help identify the conditions.

throttlehappy · **Posted:** Mon Jun 11, 2012 3:19 pm

td-d wrote:

wrxt4cy wrote:

Could the value being pulled be defined like the normal FBKC retard value so that it could reduced or eliminated?

Sure - very easy to define. Which roms would you guys want?

wrxt4cy wrote:

Great find td-d! I'm glad as well to see there is a good explanation for this odd behavior.

Now - how much would it take to understand this sub logic to know what exactly triggers this?

Well - that could take a little doing, as it's a mixture of following the logic and logging. First clue would be to log the switch that determines whether the logic continues to the 'normal' retard, or branches to this one.

I would be happy to know how far off the normal FBKC this subroutine hex value typically is as that area of the ROM is quite structured so would show up quite easily. I intend on dropping it to -0.50

As far as switch points and my research, I think it is MAF airflow related. I mainly see it where MAF rises sharply and then falls away, even by say 5g/sec before rising again such as on cracking the throttle then having the MAP RPM load compensation table takes affect to drop load after having a peak in the table etc.

wrxt4cy · **Posted:** Mon Jun 11, 2012 3:27 pm

throttlehappy wrote:

As far as switch points and my research, I think it is MAF airflow related. I mainly see it where MAF rises sharply and then falls away, even by say 5g/sec before rising again such as on cracking the throttle then having the MAP RPM load compensation table takes affect to drop load after having a peak in the table etc.

That might explain why there is often FBKC between shifts. I'll see it right before or right after on occasion but it usually goes away pretty quickly.

Heide264 · **Joined:** Sat May 07, 2011 3:22 pm **Posts:** 144

I can start logging the value if it helps out

It's an AE51910V

I do see it time to time - although after my latest timing bump (*cough, and raising the FBKC above the 1.2 load minimum) I don't see it much at all. I'll bump the FBKC back into the lower range for a while though.

wrxt4cy · **Posted:** Mon Jun 11, 2012 4:31 pm

td-d wrote:

Ok - well for AE5I910V the address is D123C - so for Ecuflash add the following to your xml:

Code:

If I get some time, I'll dig out the switches involved, maybe logging them will help identify the conditions.

Thank you!

I added a table for this to my RR defs with and I've set it to -.35 for some testing.

If you find the parameters for logging, I can test those as well.

td-d · **Posted:** Mon Jun 11, 2012 6:21 pm

throttlehappy wrote:

I would be happy to know how far off the normal FBKC this subroutine hex value typically is as that area of the ROM is quite structured so would show up quite easily. I intend on dropping it to -0.50

Very close by - e.g. AZ1G101N:

Code:

ROM:000CDC30 Feedback_Correction_Retard_Limit:.float 11.799999 ; DATA XREF: sub_3CE18+13Ao
ROM:000CDC30                                         ; sub_3CE18+178o ...
ROM:000CDC34 Feedback_Correction_Retard_Value:.float 1.4 ; DATA XREF: sub_3CE18+384o
ROM:000CDC34                                         ; ROM:off_3D2ACo
ROM:000CDC38 Feedback_Correction_Negative_Advance_Value:.float 0.34999999
ROM:000CDC38                                         ; DATA XREF: sub_3CE18+3C8o
ROM:000CDC38                                         ; ROM:off_3D2BCo
ROM:000CDC3C flt_CDC3C:      .float -5.0             ; DATA XREF: sub_3CE18+E6o
ROM:000CDC3C                                         ; sub_3CE18:off_3D028o
ROM:000CDC40 Extended_Feedback_Correction_High_RPM_Compensation:.float 1.0
ROM:000CDC40                                         ; DATA XREF: sub_3CE18+100o
ROM:000CDC40                                         ; sub_3CE18:off_3D030o
ROM:000CDC44 Feedback_Correction_Retard_Limit_Alternate:.float 1.4
ROM:000CDC44                                         ; DATA XREF: sub_3CE18+182o
ROM:000CDC44                                         ; sub_3CE18:off_3D03Co

wrxt4cy · **Posted:** Mon Jun 11, 2012 7:16 pm

Here's a log of some different FBKC I captured today. It's pretty hot here today so I was running the A/C on the freeway in 6th gear. I could feel some timing being pulled so I grabbed some logs and managed to capture one of the events which I attached.

This particular example is odd because when it happened, I was keeping a pretty steady load and only had a slight bit of acceleration. My load was pretty high due being in 6th and making some boost at low rpm. It looks like I just barely managed to exceed some limit for load at a given rpm maybe? I know Subarus like to knock under boost at low rpms and higher loads on pump gas so maybe there is some logic built in for this as well?

Also - I have not flashed a new map yet, so its pulling the original amount of timing (1.4).

td-d · **Posted:** Mon Jun 11, 2012 8:02 pm

Ok, there are a number of switches that are checked for the 'on' position (=1), and if they off , this specific part of the routine is called.

The logic follows as such (and this is only a section a very long routine):

If FFFF7D9B is equal to 1

AND

If FFFF7E14 is not 1

AND

If FFFF7E2B is not 1

AND

If FFFF7E2C is not 1

AND

If FFFF7D9A is not 1

AND

If FFFF781E is not 1

Then

Use the alternate FBKC value. So a very specific set of circumstances that calls this logic.

All the switches are 8 bit values. I haven't yet explored other areas of the rom where these are referenced, which might give a clue as to their function.

td-d · **Posted:** Mon Jun 11, 2012 8:08 pm

Interestingly, there's also a switch pretty close to the beginning of the routine, that if not equal to 1 (i.e. condition false), fast tracks the routine right the way down to put -5 straight into FBKC.

wrxt4cy · **Posted:** Mon Jun 11, 2012 8:28 pm

td-d wrote:

Interestingly, there's also a switch pretty close to the beginning of the routine, that if not equal to 1 (i.e. condition false), fast tracks the routine right the way down to put -5 straight into FBKC.

That's very interesting. I've never seen that in a log thankfully. That seems like way too much timing to pull all at once. Maybe its tied to a very low IAM threshold or something.

I sure wish I had the knowledge, patience and time to wade through this logic myself.

I'm very thankful for guys like you have have all three.

td-d · **Posted:** Mon Jun 11, 2012 8:56 pm

RAM:FFFF7E2C relates to the Feedback_Correction_Range_RPM_high_range.

mrf582 · **Posted:** Thu Jun 14, 2012 5:18 pm

Wow. Very interesting. I say we sticky this thread for a while.

Heide264 · **Joined:** Sat May 07, 2011 3:22 pm **Posts:** 144

Got it today on a full to partial throttle pull. Got a double whammy as well on it.

Normally FBKC pulls 2.2 degrees... pulled 2.8 this time around for some reason. Actual knock? Well, didn't hear anything at least? I definitely felt the timing being pulled though.

See row 8155 or so if you can't open the excel file.

Also worth noting, it sometimes increments back to zero partially over 70 samples or so, sometimes it disappears 10 data samples later. No clear pattern.

wrxt4cy · **Posted:** Thu Jun 14, 2012 8:33 pm

Heide264 wrote:

Got it today on a full to partial throttle pull. Got a double whammy as well on it.

Normally FBKC pulls 2.2 degrees... pulled 2.8 this time around for some reason. Actual knock? Well, didn't hear anything at least? I definitely felt the timing being pulled though.

See row 8155 or so if you can't open the excel file.

Also worth noting, it sometimes increments back to zero partially over 70 samples or so, sometimes it disappears 10 data samples later. No clear pattern.

Yeah - it started right when you load peaked. That's where I tend to see it.

Sometimes I also see it step back to zero without incrementing down like in your log.

I even saw a strange variation while tuning a 2012 STi a couple days ago. The tune was set to pull 1* and advance .5* on real knock. I was seeing it pull 1.4 but it would advance .5.

RomRaider

Preemptive feedback knock correction

Who is online