Case Study: Fuel Trims and the degeneration of an otherwise good tune

A little background:

Basic stage 2 hardware (Invidia catless UP, Invidia catted DP, full SPT catback, AVO TMIC, stock intake, stock subaru paper filter, clean MAF, etc.). My tune's been dialed in over a hundred revisions of repeated logging after many conditions including heat soak, rain, cold, heat, highway, traffic, etc. etc. It took a LONG time to get it dialed in to my liking where the LV and logs were completely clean and I nailed target boost within +/- less than 1 PSI.

Here is LV 08-03-2013 after driving about 2 months without resetting:

http://legacygt.com/forums/attachment.php?attachmentid=188623&stc=1&d=1406597310

Clean, fuel trims are settled in, a little more negative than I'd like (-5.3 and -5.8 in C and D) but otherwise a nice looking LV.

Next LV 2013-12-23 - 4 months later

http://legacygt.com/forums/attachment.php?attachmentid=188624&stc=1&d=1406597310

Learned Knock has developed, but without logging it, there's no way to tell for sure what happened. Trim C has gone more negative, but Trim D has settled in better. Strange. Still not a big deal. Maybe the -1.40 will unlearn itself.

Next LV 2014-02-28 - 2 months later

http://legacygt.com/forums/attachment.php?attachmentid=188625&stc=1&d=1406597310

Some more knock here and there, minor bits. But again, without logging, I shouldn't be pulling timing. Fuel trims seem to be in a nice range.

It's important to note that there are some big temperature changes going on here so all sorts of compensation tables come into play that affect MAF scaling as well as Ignition Timing. I haven't had the opportunity to sit down and log everything and to be frank, my car's been slightly neglected as far as attention. Originally when I started seeing my fuel trims go past -5 I wanted to look for hardware issues (intake leaks, sensors, etc.) but since they weren't too far out of wack, I left it alone until I could really devote my attention to it.

Fast forward to 2014-07-26 - 5 months later and much warmer

http://legacygt.com/forums/attachment.php?attachmentid=188626&stc=1&d=1406597978

Trim D has gone decidedly negative. IAM remains at 1 but FLKC all across my high load column. Still not horrible though because it's only -1.40 at most, and some of it is unlearning. Trims A, B and C seem fine, so perhaps it's not a hardware issue. However, a -7.10 in Trim D can easily cause the knock in the 2.50+ column because that indicates 7% fuel being subtracted from OL fueling!

So, this is a typical case of how when a tune is dialed in while Trim D is very close to 0 and then later on Trim D going negative and causing some knock. It took nearly 5 months for Trim D to settle in like this.

I have a few options at this point and I'm not sure how to proceed:

1. Look for hardware issues that may be inducing a negative trim of -7.10 in Trim D. It doesn't appear to be a significant hardware issue and maybe -7.10 is my car's "natural" fuel trim being that ABC are all good now. So maybe there's no hardware problem at all in which case I would have to consider the next option
   
2. Rescaling my MAF to get Trim D closer. I'm not sure I want to do this being that I don't have a WBO2 installed... this may be the kick in the butt I need to get a WBO2... safer, better to have a WBO2, but costs money and requires a hardware install. If I don't want to do this, then I can go to the next option:
   
3. Wait for the fuel trim to settle before logging and tuning. This is what I'd like to do BUT it's difficult because it seems to take forever for Trim D to settle in, especially in NYC driving.... I'd likely have to pull some timing in the cells where I'm knocking, or else add fuel to the cells where the knock is.
   

I am strongly inclined to add a WBO2 sensor for peace of mind so I know that my fueling map is correct and I don't go dangerously lean.

Suggestions?

I'm not sure if this would play a factor since you have almost a year of data, but I'm wondering if ethanol content and/or other additives that vary from winter to summer blends and differing brands of fuel could affect this. Just some food for thought.

Edit: Also +1 on the wideband idea. Not really necessary when cruising in CL but gives you immediate data on WOT pulls that could keep you from running lean on too many passes.

Injectors need servicing? Spark plugs? All maintenance items taken care of?

x2 On the injector servicing. I've been running my DW 740cc for over 2 years and finally got them serviced after developing slight cylinder roughness. One of the symptoms I noticed is some minor leaning out on the top end and various random learned knock counts. Fuel trims were not terrible, but not steady.

Based on my pre and post test, mean static flow was increased about 1.5%, and "high pulse (8ms)" variation between injectors was reduced from 3.2% to 0.9%.

Doesn't seem like a whole lot, but when you consider that the life of your engine depends on the precision of these components, it seems worth it to me.

Yeah, but I feel like before I do any sort of rescaling I want to make sure the AF ratio is actually right.

Anyway, new development today... boost is inconsistent. Seems to be requiring a LOT more WGDC to hit my targets than they did about a year or two ago. Need to bring the car somewhere and tear it apart to look for an exhaust leak, or to see if the WG door is sealing properly. AF trim going negative for Trim D I suppose could also indicate a boost leak that only shows up during high boost levels.

Not a good time to mess with the tune.

Thanks for the suggestion about the fuel injectors also. I think my baby needs a thorough refresher for a bunch of stuff right now.

How many miles on stuff? Is this the stock vf40?

Also Dom from Maxwell Power/TiC just wrote a nice article on NASIOC on how Subaru MAFs should really only be trusted for 100k miles. As they get older they run slightly less voltage especially when it's required the most at max air flow causing lean conditions.

I had a very similar issue as the OP and determined a dirty maf sensor was my problem. The problem initially surfaced seamlessly when I moved to a state with different elevation. That made it hard to diagnose. Maf cleaning spray and a quality air filter fixed my problem.

Sent from my iPhone using Tapatalk

In my case this was a small range of cells toward the lower right side of my Engine Load Comp table.

I would almost never be in those cells long enough for any real learning to take place so my D range would stay fairly happy. Occasionally though holding steady acceleration on a long upward hill on a highway would cause fairly serious negative learning in the D range that would never really unlearn.

For the longest time I drove without my D range enabled since I logged and visually watched all my data for all my driving. I re-enabled it after going to E85 since it's much more forgiving. On gas though, I didn't want to risk running 5-10% lean or tune 5-10% rich just in case.