Alright, I think I know what is going on here. On my E85 rom, I have had the long term fuel trims disabled for a long while now (by setting the closed loop target addative table to zero). I believe this is the reason I have not seen the "limp mode" rich fueling before.

Everything I have read on romraider indicates that the ecu only really cares about the sensor signal to the ecu, and if that is missing or out of range, it will go into limp mode with the rich closed loop targets. 02 can be unplugged, so long as the heater dtc's are turned off and you are feeding a narrow band signal to the ecu in place of the secondary 02 signal.

What I don't understand though, is why zeroing the AF3 tables does not remove the rich target as it normally does on the other subaru ecu's that have this limp mode issue.

I think the only real "fix" for removing the rear 02 on this platform is to feed a narrow band signal to the ecu.

So can we successfully disable the secondary O2 and replace it with an aftermarket wideband for external logging? Or is it better to just have another bung welded in on the factory header or front pipe (not sure where the LC-1 will fit) and avoid having to muck around with disabling codes and zeroing tables (I can't see one the one ztan mentioned either on A01G RR defs).

I just want to be able to have wideband AFR logging on the laptop for street tuning.

If you plan on removing the rear 02 to install a wideband in it's place, you will need to run the widebands narrow band emulation signal to the ecu rear o2 signal input. Otherwise you will get limp mode (rich closed loop afr target of about 13.7 rather than 14.7). I have tried every combination of dtc's, AFR 3 settings etc and none of them make any difference when the rear 02 is out of the pipe, plugged in or not. The only thing that does stop the rich target is dissabling long term trims completely by zeroing the closed loop load compensation map. That will eliminate the rich target and you will target 14.7 all the time in closed loop.


===================================

As my final test, I put the rear 02 back in the pipe with no other changes, my rich closed loop target went away, the previous config was with the sensor plugged in but not in the pipe. This matches all the info on romraider about the newer 32bit ecu's.

What does not match, is that zeroing all 3 (or any combination of) of the AF3 maps has zero effect on the rich target when the rear 02 is either unplugged or in free air and the long term trims are enabled (closed loop target comp table is non zero). At least for my ZA1JA00C ecu. Mine is one of the freaks that likes to reset IAM to the rom's default if it sits for more than a couple of hours.

For What it's worth, mine never obeyed the AF learning limits regardless of how they were set when i tested them a long while ago. That is true for both brzedit and the open source tools. Same is true for AF3. No clue why but I'm starting to think there may be multiple hardware version ecu's out there or i just got bad ecu from the factory.

It's either that, or brzedit changed a piece of code or a register in the ecu that the other flashing tools does not overwrite.

Now that I am on my second tank of pump gas (and fuel trims are very low again) I can test turning off the long term trims and running with the rear 02 unplugged again, but I am positive it will work fine that way, just without any long term trims. I may just bite the bullet and run the narrow band signal to the eco to keep it happy since I will absolutely want my wideband connected if I add a supercharger latter this year.

Sounds too hard, may as well leave factory o2 alone. I'll either get a bung put in on the over or front pipes to connect the LC-1 or make up a clamp that can stick it in the tailpipe like we would on the dyno. I don't need it there 24/7, just while tuning.

For a temporary solution you could just make up a socket or pin (not sure which you need since I'm not looking at the connector) connected to the lc-1 analogue output wire, unplug the factory 02, insert your lc-1 connector on or in ping 3 of the ecu side connector and use the secondary 02 bung for the lc.

I've had a look at the disassembly to see what happens to the rear O2 sensor input.

A01G ROM logging for Rear O2 sensor input:
An example of where the Rear O2 sensor data is pulled:
The Rear O2 sensor data is pulled when the Pull_16_bit_sensor subroutine (12F24 in A01G and 12FCC in A01C) is called with the value 0x2C in r4. If you have a look for references to the subroutine for pulling 16 bit sensor data with 0x2C in r4, there are a lot of them.

It should be possible to put a hook into the sensor pull subroutine and feed it narrowband data (based off the front O2 sensor) if r4 = 0x2C, whilst using the sensor input for something else without giving the ECU too much illogical information and sending it into some kind of limp home mode.

How are people disassembling the roms? I'd love to get my hands on this.

IDA: http://www.romraider.com/forum/viewtopic.php?t=6303

Great info man!