Logging Subaru STI fuel trims around the "GR stumble" zone

In "Step 4: MAF/Injector Scaling" for the ECUFlash worked example on the Subaru STI, at timestamp 58:55 Andre switches over to a log file which shows an engine speed range of 1800-3800 RPM. This is done after doing a ramp run in 4th gear and then looking at the log data and seeing that there is an AFR error around the 2800-3700 RPM range. Andre talks about using the engine load compensation table to account for both the open loop AFR error and the closed loop fuel trim error around the 2800-3700 RPM range, rather than trying to making changes in the MAF sensor scaling table, however he never mentions what gear the 1800-3800 RPM log file was done in for the closed loop trim data. Does the gear matter while logging closed loop fuel trim corrections around the "GR stumble" RPM zone?

Quick recap for anyone who hasn't watched the ECUFlash STI videos, Andre talks about using 2nd gear and holding steady at 2K RPM for scaling the MAF sensor (avoiding the 2.5K-3.5K RPM range because of the fuel line pulsations and how that can effect the closed loop trims). After that, he uses 4th gear and holds steady at 4K RPM. The MAF sensor voltage changes as load and throttle are applied during the tests. Using the information gathered during logging, Andre then edits the MAF sensor table and changes the values for 3.1V and below by 2.5% and then changes the values for 4.3V and above by 4%. The cells between 3.1V and 4.3V are then changed by 3% even though there was no logged data for this area. This is done in order to keep the MAF scaling curve smooth. After that, Andre repeats the 2K and 4K RPM tests to check the fuel trims and it turns out the MAF scaling is now pretty close to where it should be. During a 4th gear ramp run however, the measured AFR is a bit off from what the ECU was actually commanding. This is most pronounced around the 2800-3700 RPM area and is primarily due to the fuel pulsations at that engine speed. Andre recommends using the engine load compensation table to account for this error rather than using the MAF scaling table and to do that, he uses data gathered from 1800-3800 RPM, presented in a histogram with RPM on the vertical axis and manifold pressure on the horizontal axis. The total trim is presented on the diagonal axis as a percentage. The histogram view matches the breakpoints in the engine load compensation table and the total trim percentages are then added to the corresponding cells in the load compensation table in order to correct the closed loop trims in and around the 2800-3700 RPM region.

In case anyone comes across this thread looking for guidance on calibrating their MAF sensor via street tuning, I ended up using 3rd gear for the engine speed range of 1800-3800 RPM that I mentioned. I got some pretty major fuel trim corrections in that range, even after I calibrated the MAF sensor. Just to summarize what I did, I drove the car in 2nd gear and held the engine speed at 2K RPM while smoothly applying brake and throttle in order to vary the MAF sensor voltage from around 2.0V to around 2.5V, right up until it switched into open loop. I then used the same brake and throttle method at 4.2K RPM while driving in 4th gear to vary the MAF voltage from around 2.4V to about 3.1V, until it switched into open loop. I logged the trims for these tests over the course of multiple days, making small changes to the MAF sensor scaling during each iteration before repeating the tests and making further changes. I also logged the trim during idle after the car was fully warmed up and applied the correction value for that part of the MAF calibration, using linear interpolation for voltages not seen during logging in order to keep the MAF curve smooth. After all that, I was getting fuel trims of under +/-5% up until the 3000RPM mark. As soon as I hit 3000 RPM however, the trims quickly rose to +25% and my wideband started to read lean. Using 3rd gear and logging around 1800-3800 RPM, I then applied the total trim correction values seen within the engine load MAP compensation table. For me, the biggest trims were happening around the 3200 RPM area. I still have some work to do as I log different load situations, since I'm getting quick spikes of about +/-18% and I'm not sure why, but most of the time I'm seeing trims around +/-3% when cruising in low RPMs and +/-6% in high RPMs, and maybe +/-8% while driving smoothly and shifting between gears. I'm pretty happy with that since I was regularly hitting +/- 25% almost immediately every time I drove the car. When I first got the car it already had an aftermarket intake, high-flow downpipe, and a Cobb off the shelf tune that is no longer available and I couldn't find much information on. I think its safe to say the car is much better now, even though it still needs a lot of work.

Question for anyone who happens to come across this, is there a limit to how large of a corrective value you can/should apply within the engine load MAP compensation table? For example, at the 3200 RPM row at both 6.96 PSI and 7.93 PSI columns in Cobb Accesstuner (-7.73 PSI and -6.77 PSI in ECUFlash) I have compensations of about 36%. Despite this, I'm still seeing trims for around +6% in that area. Is a MAP correction of +40% to be expected for a 15 year old fuel pressure regulator?

Another update for anyone interested, I decided to datalog while driving normally, stopping at street lights, shifting through gears, cruising on the highway, ect. I continued to apply the trim corrections to the MAP compensation table over the course of two or three tune iterations. I'm still seeing the occasional spike of about +/-15% total trim but the values shown in my datalogged histogram are currently within about +/-1% in almost every cell of the table while also having large hit counts. There are a couple outliers of +/-2% or so but that's the largest histogram values I'm seeing. The large spikes happen around gear shifts or during low load conditions such as getting the car rolling from stand still, and like I said they're really only about -15% or +10% max and then drop close to 0% after a couple seconds. All in all, I think I've got my MAF scaling and MAP comp table dialed in. I'm now going to move on to getting the transient enrichment where I want it and then I'll take a look at optimizing the cold start and idle. Also FYI, after the multiple tune iterations and making sure I was as smooth as possible with my throttle input, the largest values currently in my MAP comp table are actually +30% to +35% in the 5.99PSI to 8.89PSI region at 3200RPM in Cobb Accesstuner. Hopefully this is useful to someone in the future. Cheers!

Thanks for sharing your detailed process and results. I'm sure that will be useful for other STI tuners.

Michael. Thanks for the great type up. I am currently going through the same module.

For the first set of adjustments to the MAP comp table he uses both a closed loop histogram (using total trim) and open loop histogram (using AFR error Subaru to adjust the 3.87 psi column). After his revision his next histogram (looking at the results) is exclusively using AFR error. I have some questions related to this:

1. I assume that this MAP comp table will adjust fueling regardless of open or closed loop is that correct?

2. To adjust the MAP comp table do I need to take both closed loop logs (using total trim) and wide open throttle open loop ramp runs (using AFR error) separately and make adjustments off both through this rpm range to set it up properly?

3. How do I go about collecting the wide variation of psi/rpm data that he collected in the first and third (revision) histogram? It almost seems as if at each rpm range he increased and decreased the resistance (leading to increase and decrease in psi); then moved to the next higher rpm and doing the same thing?

Overall I'm having trouble understanding what data to collect (open loop, closed loop), and how best to collect it??

Appreciate the help,

Paul

Hi Paul. Concerning your first question, in EcuFlash there is an Engine Load Compensation cruise and non-cruise MAP comp table. In Cobb's Accesstuner software the same tables are referred to as Load Compensation (Manifold Pressure) TGVs closed and TGVs open. In other words, there is a table for open loop and closed loop. I believe Andre recommends making these two tables essentially identical at the 1000 RPM row and above (the cruise table also includes two rows below 1000 RPM). If you do this then it just makes things much simpler and the numbers you enter will apply to both closed and open loop.

For making adjustments to the table while focusing on closed loop trims, you'll want to keep things simple in the beginning and try to not enter open loop at all when datalogging. I did a bunch of datalogs in 3rd gear, varying the engine speed around 1800-3800 RPM and applying the total trims to the MAP comp table. My total trims were constantly over +/-25% so it took some time to adjust. The wide open throttle ramp runs should be datalogged separately and then the AFR error applied to the MAP comp table. Clear the learned trims beforehand, as long term trims may still effect open loop. I did these in 4th gear, starting around 2K RPM and going to redline. You'll find that the data collected here will appear almost exclusively in the far right columns/rows of the MAP comp table. The data from closed loop will appear almost exclusively in the other columns/rows of the table. So in the end you'll get pretty good data for the entire table. This might take multiple datalogs and for me it actually took a few days as I only recorded my driving during my work commute.

So, if you do what I did and focus on the area where there is a lot of trim being applied in closed loop (3200 RPM for me) and then do your wide open throttle AFR error corrections, your MAP comp table will still have a good portion of cells that still need adjustment. Your third question is the same one that I had, as Andre doesn't specifically state how he went about getting that wide variation of data. He's obviously got a dyno so I believe he basically just varied load and RPM, filling the cells with a large number of hits as he moved through the table. I was able to basically do the same thing by just driving the car normally. I started my datalog at idle, drove through my neighborhood in low gear and low speeds, stopped at street lights and stop signs, and then eventually got on the highway and cruised in 6th gear, shifting into 5th a few times while climbing hills, intentionally letting the RPMs enter into the trouble area of 3200 RPM before eventually shifting back into 6th/2500 RPM. The whole time I watched the closed/open loop switch, making sure I stayed in closed loop almost the entire time but otherwise I essentially just drove as I would normally. If I saw it enter open loop, I backed off the throttle and tried to get back into closed loop as quickly as I could. Then again, when looking at the data, you'll use the same total trim histogram as before so jumps to open loop shouldn't really effect the values seen here. Still, I'd try to keep it in closed loop as much as possible since that's the whole point. I did this two or three times, making small adjustments to the MAP table each iteration and now my total trim histogram shows values no larger than about +/-1%. Datalog a wide open throttle ramp run and then switch the histogram back to the AFR error setup and you should see similar values in the last two columns/rows just like before. Looking at these two histograms together, if all of the values reported are relatively low, then you can consider you're MAP comp table tuned correctly.

One other thing I should note, if you're using a Cobb Accessport or going through your logs and looking at the maximum/minimum fuel trim values (correction or learning), you might notice that large numbers are still being seen. As I mentioned in my previous post above, after fine tuning my MAP comp tables and getting everything within +/-1%, I still saw min/max trims of +/-15% and sometimes even greater than that. Looking at the logs, these spikes would happen during gear changes or quick throttle inputs, even though I thought I was being very smooth with my throttle application. I've come to the conclusion that these spikes are a result of an improperly tuned enrichment table (tip-in or otherwise) and so I need to do some more work there. If you also see large trims on the min/max values, just know that they happen fast and relatively infrequently, so you can absolutely still consider your MAP comp table tuned as long as your histogram values look good and you've got plenty on hit counts in the cells.