AF#3 (Rear O2 sensor) correction limits

MFB · **Posted:** Sat Feb 25, 2012 7:30 am

Hi, can you also get the AF#3 address for the ff ecus:
AZ1G900P, AZ1E400U, AE5F301C and A8DK100P

Thanks in advance Td-d!

td-d · **Posted:** Sat Feb 25, 2012 7:49 am

No problem, as soon as I get near my computer.

I must be as disciplined as NSFW and upload instructions!

dschultz · **Posted:** Sat Feb 25, 2012 5:12 pm

I believe that you can query the float (4byte) value of AF3 Learning by logging the value equal to the address of "A/F Learning #1 D (Stored)*" + 8
i.e.: if "A/F Learning #1 D (Stored)*" = 0xFF2630
then "A/F Learning #3 (Stored)*" = 0xFF2638

Then later in the code I see this:

Code:

ROM:0002D7A0 D3 5F       mov.l   @(h'17C,pc), r3 ; [0002D920] = E_AF_Learning_3_StoredExt_Exx
ROM:0002D7A2 F3 38       fmov.s  @r3, fr3
ROM:0002D7A4 D2 5F       mov.l   @(h'17C,pc), r2 ; [0002D924] = E_Closed_Loop_Fueling_Target_4byteExt_E121
ROM:0002D7A6 F4 28       fmov.s  @r2, fr4
ROM:0002D7A8 D1 5F       mov.l   @(h'17C,pc), r1 ; [0002D928] = unk_FFFF56A5
ROM:0002D7AA F4 31       fsub    fr3, fr4

The most important line is the last one which is the CL Fuelling target - the AF3 Learning value.
Just need someone to log to validate this.

I've further tracked this to the DTC code for P2097_POST_CATALYST_TOO_RICH_B1 and P2096_POST_CATALYST_TOO_LEAN_B1 evaluation.

td-d · **Posted:** Sat Feb 25, 2012 5:17 pm

^ Hmm, thanks Dale - let me track it in my rom, and see if I can log it.

2bfi · **Joined:** Fri Dec 25, 2009 8:46 am **Posts:** 93

I logged AF Learning #3 using the instructions above.

It came up with "6600" for all values of AF Learning #3. Correction limits are still stock at -50% to 50%.

Attached is the log file, and related portion of the log.cfg file. This is for an AZ1J500G '11 EDM STI.

Code:

paramname = A/F_Correction_#3(%)
paramid = 0x0000D0
databits = 8
scalingrpn = x,.078125,*,5,-

paramname = A/F_Learning_#3(Stored)*
paramid = 0xFF2D78
databits = 8
scalingrpn = x,100,*

2bfi · **Joined:** Fri Dec 25, 2009 8:46 am **Posts:** 93

td-d wrote:

2bfi wrote:

I logged AF Learning #3 using the instructions above.

It came up with "6600" for all values of AF Learning #3. Correction limits are still stock at -50% to 50%.

Hmm. Interesting - I guess we need to dig out what the scaling is - the raw value would have been 66.

I'll look through your rom more thoroughly, and send you some values to test via PM for the learning parameter, if you're willing? Your SD rom is ported, btw, will just need you to test the RAM holes

Yes I'm willing. However, I'm such a noobie at Subaru's I hope I can do a good job.

Just send me the instruction on what to do and I should be able to figure it out.

My experience with SD is 12 years on Hondas and have tuned countless cars. And, I understand how SD is implemented in the Subaru.

One last thing, I zeroed out the AF Correction #3 min and max limits. Logged AF Correction #3 remained at 0. It works!

dschultz · **Posted:** Mon Feb 27, 2012 11:26 am

2bfi wrote:

I logged AF Learning #3 using the instructions above.

It came up with "6600" for all values of AF Learning #3. Correction limits are still stock at -50% to 50%.

Attached is the log file, and related portion of the log.cfg file. This is for an AZ1J500G '11 EDM STI.

Code:

paramname = A/F_Learning_#3(Stored)*
paramid = 0xFF2D78
databits = 8
scalingrpn = x,100,*

^------ this is a float 4byte value not an int 1byte value.

Try this:

Code:

paramname = A/F_Learning_#3(Stored)*
paramid = 0xFF2D78
isfloat = 1
scalingrpn = x,100,*

td-d · **Posted:** Mon Feb 27, 2012 5:11 pm

Didn't notice that... Defined as an 8 bit value.

2bfi · **Joined:** Fri Dec 25, 2009 8:46 am **Posts:** 93

dschultz wrote:

2bfi wrote:

I logged AF Learning #3 using the instructions above.

It came up with "6600" for all values of AF Learning #3. Correction limits are still stock at -50% to 50%.

Attached is the log file, and related portion of the log.cfg file. This is for an AZ1J500G '11 EDM STI.

Code:

paramname = A/F_Learning_#3(Stored)*
paramid = 0xFF2D78
databits = 8
scalingrpn = x,100,*

^------ this is a float 4byte value not an int 1byte value.

Try this:

Code:

paramname = A/F_Learning_#3(Stored)*
paramid = 0xFF2D78
isfloat = 1
scalingrpn = x,100,*

Whoops. Ok, I've corrected it to "isfloat = 1". Output is a different constant "6379.76" all the way. Learning limits are at -50% min & 50% max stock values.

dschultz · **Posted:** Tue Feb 28, 2012 11:19 am

2bfi wrote:

Whoops. Ok, I've corrected it to "isfloat = 1". Output is a different constant "6379.76" all the way. Learning limits are at -50% min & 50% max stock values.

It's possible in the "newer" ROMs that it's in a different location. I was working off of an '05 ROM. I'll take a look later...

dschultz · **Posted:** Wed Feb 29, 2012 2:23 am

dschultz wrote:

2bfi wrote:

Whoops. Ok, I've corrected it to "isfloat = 1". Output is a different constant "6379.76" all the way. Learning limits are at -50% min & 50% max stock values.

It's possible in the "newer" ROMs that it's in a different location. I was working off of an '05 ROM. I'll take a look later...

I finally got around to logging this on my car and the value is always 0. For the moment I can't see where this value is updated after it is initialize to 0 in code.

For the AZ1J500G ROM, a quick look seems to indicate that there is no stored Learning value, not in the same location at least.

MFB · **Posted:** Wed Feb 29, 2012 4:45 am

Awesome update on the addresses! (which I missed on the first post) Thanks Oren :-)

td-d · **Posted:** Wed Feb 29, 2012 4:51 am

dschultz wrote:

I finally got around to logging this on my car and the value is always 0. For the moment I can't see where this value is updated after it is initialize to 0 in code.

For the AZ1J500G ROM, a quick look seems to indicate that there is no stored Learning value, not in the same location at least.

Interesting - seems to be the case for pretty much all the roms. I guess it's a piece of code that was incorporated, and never 'switched on'. Given that the AF3 corrections are within a pretty narrow range, maybe learning was superfluous.

I'm still curious to see what happens when the rear O2 is completely disconnected - besides the CEL, whether this removes the drop to 13.5~13.7 AFR. Mark - have you had a chance to try it? I'm too busy tuning SD to get to it

So much tuning, so little time...

MFB · **Posted:** Wed Feb 29, 2012 5:51 am

td-d wrote:

dschultz wrote:

I finally got around to logging this on my car and the value is always 0. For the moment I can't see where this value is updated after it is initialize to 0 in code.

For the AZ1J500G ROM, a quick look seems to indicate that there is no stored Learning value, not in the same location at least.

Interesting - seems to be the case for pretty much all the roms. I guess it's a piece of code that was incorporated, and never 'switched on'. Given that the AF3 corrections are within a pretty narrow range, maybe learning was superfluous.

I'm still curious to see what happens when the rear O2 is completely disconnected - besides the CEL, whether this removes the drop to 13.5~13.7 AFR. Mark - have you had a chance to try it? I'm too busy tuning SD to get to it

So much tuning, so little time...

I got the car up the lifter earlier today to swap out sway bar bushings. It did cross my mind to remove it and prove a point. haha If it affects close loop (I go rich to 13.5-.7) then I have to lift it up again right? :mrgreen:

At this point, its confirmed that af3 learning does not take place and af3 correction is disabled with the table you found. Does it affect close loop? It should and its better turned off for catless set ups like ours. :-)

dschultz · **Posted:** Wed Feb 29, 2012 10:58 am

td-d wrote:

Interesting - seems to be the case for pretty much all the roms. I guess it's a piece of code that was incorporated, and never 'switched on'. Given that the AF3 corrections are within a pretty narrow range, maybe learning was superfluous.

I also logged the #3 correction. It is limited to +/-2.5% and holds steady at it's last value during OL or CL deceleration.

RomRaider

AF#3 (Rear O2 sensor) correction limits

Who is online