I can't seem to figure this one out. I've been running with the same tune for a while now so I think this is the first time I've given the ECU a chance to start using learned fuel trims. As a consequence I started getting DTC P1274(Bank 1 rich) and P1283 (Bank 2 lean). The problem is that according to the logs bank 1 is actually running lean and bank 2 is actually running rich so the ecu ends up just making the situation worse as it tries to lean out a bank that is already lean and enrich a bank that is already rich. This behavior is supported by my logs which I've included here. I have cylinder trim set to 100% for all cylinders. My correction table is flat at 100%.

Here is a graph of fuel corrections with learned fuel trims. The corrections represents what I must adjust within the MAF table to get the actual AFR to equal the target AFR. Note that I would have to lean bank 2 or enrich bank 1 by almost 10% to get them to be equal.



Here is the same graph now with the learned fuel trims reset. Note how the corrections are almost equal between the banks.



Here is a fuel table built from the data logs with learned fuel trims. Note how the discrepancy is only present in closed loop as seen in the low load cells but when in open loop in the higher load cells my AFRs start coming together.



And here is a fuel table built from the data logs when the learned fuel trims are reset.



My setup is UpRev, HPX PMAS MAF, stock manifold with OEM wideband -> BP turbo -> random tech cat -> OEM narrowbands -> OEM exhaust.

 

Are all your OEM sensors directly in the exhaust stream? Assuming you've got a ST setup you might consider moving the downstream sensors upstream so they can see each bank independently as it's a proven fact the ECU can and sometimes will make corrections based on their readings.

 

Yup. Widebands were untouched. Narrowbands are installed in the same exhaust stream using standard bungs post single cat with some extension wiring. I thought they could be causing the problem so I looked at the voltage from each sensor. Prior to ECU reset the average voltage difference (b1-b2) is 0.0082V. After reset it's 0.0085V. This seems like an insignificant difference, but even if it was significant it should tell the ECU that bank 1 is leaner. It's all very puzzling. I can't find any difference between the b1 and b2 sensors after reset to explain why the ECU should try to enrich one over the other. The voltage difference after reset for the widebands is 0.016V.

The only thing I can think of is that the ECU is looking at the oxygen consumed by the cats to monitor their efficiency. Because I mix the exhaust between both banks prior to the NB sensors a slightly lean bank might appear to the ECU as a cat that needs more oxygen because it appears to have extracted too much because the oxygen content is diluted by the richer bank. The solution for emission is to lean out an already lean bank. If this is how it works then even if I could adjust the cylinder trims to make each bank exactly the same it would be an unstable equilibrium. The only other solution would be to run dual cats if I wanted to continue with UpRev for engine management. I'd rather go with an aftermarket ECU if that is the case.

Is there anybody else who has run a single cat and kept all emissions functions intact with UpRev?

 

Just realized after looking at the logs that the voltage of the WB increases with increasing oxygen content and voltage of the NB decreases with increasing oxygen content. So I guess an average difference between the NB sensors (b1-b2) of 0.0085V technically means that b1 is richer than b2. Since I mix the exhaust prior to the NB sensors, ECU changes between the two banks never make b1-b2=0. I'll play with the cylinder trims to see if I can make b1=b2 but I have a feeling even if I can make them equal on one run they won't stay equal. What I probably need is a splitter so I can use 1 NB sensor to feed both ECU inputs. The trick will be to get the heater circuits etc to work properly.

 

You might be able to Y-off the voltage output wire (assuming it's just one) to both inputs on the ECU, and let the heaters stay independent. However I'd still recommend moving them upstream and keeping them separate if possible. You might try disconnecting them altogether and disabling all related codes in the ECU.

And FYI, B1 is leaner than B2 on a typical VQ (opposite of your findings).

 

This is what I just did. The NB sensors have 4 wires: ground, heater power, heater signal and O2 signal wire. O2 signal wire is black. Just picked one and fed both ECU inputs. So far no DTC's and logging shows both NB are now equal. Will take some time to relearn the long term trim values since I'll be travelling and won't be driving the car much next month.

I'd like to avoid moving the sensors upstream from the cat. I'd like to retain the ECU ability to monitor the cat. This is a street, maybe an occasional track, car and I'd like to maintain the feel that the car came from the factory like this.

 

I'd far prefer to retain the ECU's ability to trim each bank independently over a catalyst indicator. You turbo'd the car and use an HFC, there's no green card left to pull.

 

I think the differences in fuel delivery between the banks are due to immutable differences in architecture and a single adjustment based on logged data to the cylinder trims should be able to address these differences. The ECU trim adjustments only appear to work in closed loop and based on my observations the slightest hint of spirited driving puts the ECU in open loop. I suspect that when I tune the cylinder trim it should apply in closed and open loop. In all likelihood manual adjustments to cylinder trim will make ECU adjustments unnecessary; however, I can imagine some scenarios where if I left the ECU to adjust closed loop and I adjusted trims for open loop the switch from closed to open loop could introduce strange and unfavorable behavior.

Second, I am a little worried that running rich for extended periods of time on the track might destroy the cat and retaining the ECU's ability to monitor its function has some benefit in this setting.

Third--and this is totally anecdotal--a buddy of mine came into town and we swapped cars for test drives: stock 2010 335 with some M sport options, stock 2011 evo mr and my g35. The consensus was that the G rode better than the evo and as nice as the 335 (except for the brakes on my G which sound like a dump truck) but pulled much harder than the other 2 and I did it for a fraction of the cost and frustration that I see other people spending on these cars. And this just reinforced my approach to modifying this car.

Fourth, sometimes doing things the hard way is more fun.

 

I started getting cat codes after about 2 years on the G with dual 3" HFC. It's pretty much a given with HFC (I was lucky to not have them in the beginning), but it doesn't mean they aren't working. Compared to my catless Z the G smells like tulips and roses. I deleted the P04x0 codes for the sake of saving the CEL indication for more serious issues. The funny part about it all is the Z doesn't throw any codes at all with the catless downpipes and all sensors in place. I think the 2003's with all narrowband sensors aren't nearly as picky about catalyst efficiency.

 

Do you remember which specific codes you were getting?

 

P0420 and P0430 -- cat efficiency codes. Anything sensor related other than those two cat codes are most likely affecting how your car runs.

 

Did you ever try analyzing the voltage data to see how it changed over time? It might be that you've lost some oxygen storage capacity. I had been looking at the ratio of WB/NB voltage of the OEM cat and HFC to determine efficiency, but I wasn't getting very meaningful results. I think I need to actually determine the amplitude and frequency of each signal and compare. I think I'll need to use FFT to do this which I've never done before. Looks like I have some reading to do.