ECUFlash - Getting close!

[Td-d]

Originally Posted by steve99

the last one "D" looks a bit suspect. Not exactly sure what these are though :-)

Ok, I'm on to something. The long term trims are stored in 'bins' (think of the table af learning ranges) according to the MAF g/s - that's where the LTFT value comes from. Note that the LTFT value is the same

I'll need to look more closely on a pc, but looks like at least one of the value is correct.

[steve99]

Originally Posted by Td-d

Ok, I'm on to something. The long term trims are stored in 'bins' (think of the table af learning ranges) according to the MAF g/s - that's where the LTFT value comes from. Note that the LTFT value is the same

I'll need to look more closely on a pc, but looks like at least one of the value is correct.

hmmm I have been doing maf scaling so my ltft,s are very low or 0 in that log, have not disabled anything though

[vgi]

@Td-d

and here's mine on a01c

http://www.datazap.me/u/vgi/aflearni...12-13-23-24-26

[Td-d]

Ok, did some very extensive trawling through the logic today, mind numbing. There are actually 2 sets of AF learning range tables, with slightly differing values that seem to be blended in arriving at the AF learning applied. I suspect that making both tables equal would make MAF scaling (and any manipulation AF correction / learning applied) much easier.

So - I've made both table available in Romraider now (updated definition available here), and Vgi, can you log the following for A01C:

E_AF_Learning_1_A_StoredExt_E44_Range_A FFF81684 
E_AF_Learning_1_A_StoredExt_E44_Range_B FFF815C4 
E_AF_Learning_1_B_StoredExt_E45_Range_A FFF8168C 
E_AF_Learning_1_B_StoredExt_E45_Range_B FFF815CC 
E_AF_Learning_1_C_StoredExt_E46_Range_A FFF81694 
E_AF_Learning_1_C_StoredExt_E46_Range_B FFF815D4 
E_AF_Learning_1_D_StoredExt_E47_Range_A FFF8169C 
E_AF_Learning_1_D_StoredExt_E47_Range_B FFF815DC

The two learning range tables, side by side look like this:



I had seen the second table before, but had assumed that like in the older roms, the table A was the main ingredient, with the outer values (1.6 and 60) as the min and max values - but looks like table two is used as well, which would explain why in the first values I gave you, in some areas, long term trims match one of the cells under the stored learning, and in some areas not. There is also a separate minimum and maximum MAF g/s limit (1.6 and 200).

[vgi]

Originally Posted by Td-d

and Vgi, can you log the following for A01C

Will do, if not tonight then tomorrow morning.

[vgi]

@Td-d
here's the log:
http://www.datazap.me/u/vgi/test3?lo...-9-10-28-29-31

doesn't look right but i left the scaling and data size the same as before. should it be different?

; E_AF_Learning_1_A_StoredExt_E44_Range_A
paramname = AF_Learning_1_A_Stored_Range_A
paramid = 0xFFF81684
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_B_StoredExt_E45_Range_A
paramname = AF_Learning_1_B_Stored_Range_A
paramid = 0xFFF8168C
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_C_StoredExt_E46_Range_A
paramname = AF_Learning_1_C_Stored_Range_A
paramid = 0xFFF81694
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_D_StoredExt_E47_Range_A
paramname = AF_Learning_1_D_Stored_Range_A
paramid = 0xFFF8169C
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_A_StoredExt_E44_Range_B
paramname = AF_Learning_1_A_Stored_Range_B
paramid = 0xFFF815C4
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_B_StoredExt_E45_Range_B
paramname = AF_Learning_1_B_Stored_Range_B
paramid = 0xFFF815CC
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_C_StoredExt_E46_Range_B
paramname = AF_Learning_1_C_Stored_Range_B
paramid = 0xFFF815D4
isfloat = 1
scalingrpn = x,100,*

; E_AF_Learning_1_D_StoredExt_E47_Range_B
paramname = AF_Learning_1_D_Stored_Range_B
paramid = 0xFFF815DC
isfloat = 1
scalingrpn = x,100,*

[Td-d]

Well - the 'B' range looks to be correct - you can see the LTFT correlates with the A to D ranges regularly, and also interpolates between the cells at times. Obviously, the A range is not correct - will carry on looking. If you are feeling adventurous, makes the learning ranges tables the same (I would use the A table), and see what that does to trims.

[vgi]

Originally Posted by Td-d

Well - the 'B' range looks to be correct - you can see the LTFT correlates with the A to D ranges regularly, and also interpolates between the cells at times. Obviously, the A range is not correct - will carry on looking. If you are feeling adventurous, makes the learning ranges tables the same (I would use the A table), and see what that does to trims.

Not sure if it's anything but I noticed that the B range has contiguous addresses than there was af_learn_3 then went A range. If the address space has any meaning I would guess the addr range for A range should be right before B range.

[Td-d]

^ tried that, still didn't find them Have to do it the hardway, through tracing the logic...

[Td-d]

Alright, Vgi - try these ones:

E_AF_Learning_1_A_StoredExt_E44_Range_A??? FFF81618
E_AF_Learning_1_B_StoredExt_E45_Range_A??? FFF81620
E_AF_Learning_1_C_StoredExt_E46_Range_A??? FFF81628
E_AF_Learning_1_D_StoredExt_E47_Range_A??? FFF81630

[vgi]

ok, will post tomorrow morning

[D-VO]

@Td-d: Here's my logcfg.txt and logcfgout.txt for the B01C ROM. I also added a log of what shows up under the "RPM" and "Speed" columns. I disabled the defogger switch start to get it logging.

[Td-d]

@D-VO - your defogger switch is not working because you've got it defined as an 8 bit value - needs to be float.

Engine speed parameter has the wrong address, should be FFF8DB98
For speed, try using FFF8DB9C rather.

I suggest going through the logcfg and comparing it to the address list posted, looks like some might not have been copied across.

[Td-d]

Originally Posted by vgi

almost forgot to post

Bingo...

I'm seeing a lot more correlation now between the LTFT and these stored values. In fact, there's only one value that I'm not seeing (-5.49806, which kicks in above 20 g/s - i.e. D range of Table A) which may be in another ram address - but on the most part you can see how it's using values either from the ranges in B or in A, depending on the MAF g/s.

I suspect that if I manage to dig out the fueling mode, it will demonstrate how it's switching across the two tables.

So that being said, it's not as useful as in the older roms - I was hoping to provide a way to see a 'learning view' in the absence of having that tool for the BRZs - but if the tables are different, it switches between the tables constantly.