An Idiot Learns to Tune, Part 2

[spfautsch]

I've been working on the tune for my car for quite some time. A slipped optispark hub, an unfinished ignition controller project, and life in general has slowed the progress immensely, but I've arrived at a point where I feel like it's 95% perfect and I finally have all the pieces to the puzzle together (development halt on the ignition controller and working + accurate wideband). Last fall I decided when I got to this point I wanted to run a tank of ethanol free gas through it to see what happened to trims and PE AFR. Here in the Midwest it's difficult to buy gasoline that doesn't contain "up to 10%" corn liquor. Since I did most of my tuning on a steady diet of 93 octane E10, I wasn't sure what to expect, but the shift was quite distinct so I thought I'd share it with anyone interested. Here's a quick comparison running the same tune for all the log data.

Here are the logs if anyone's interested. [link] Weather conditions were fairly consistent - hot and humid. The ones with "E0" in the filename were captured running on 10 gallons of roughly 90 octane pure dino juice. This is a MAF tune - since I didn't have a really accurate injector constant for the modified LS2 injectors I have, I made the assumption the factory MAF calibration is accurate and tried to determine the flow constant with trims and / or wideband data.

E10

E0

It's difficult to see from the graphs, but the PE lambda is fluctuating between 0.86-0.88 on the E10, and 0.83-0.85 on E0 (BLM locker patch in use).

Essentially this is a very verbose and graphic presentation of the simple idea that E10 is stoich at about a 3% richer mixture than pure gasoline. What I learned from this is that when tuning with E10 it's best to try and have your trims adding around 2%, whereas I was targeting trims that were removing about 2%. This doesn't seem like a lot, but with the E0 my cam surge is far more pronounced, so I think I'm going to scale my injector constant up a bit and see how it reacts.

The cam surge is about the only aspect of the tune I'm not really happy with. Once I have the injector constant so E10 produces BLMs in the neighborhood of 130-132, I'm then going to go back to fine tuning individual cylinder trims to see if there's any improvements to be made. I finally broke down and bought a high quality IR heat gun and after taking measurements on a couple different days I'm seeing #2 seems to be running hotter than the other 7 holes. I'm not sure if this means it's rich or lean, but at least I have an idea which cylinder might be the culprit in causing a slightly lean condition on the right bank.

[spfautsch]

I wanted to add that if anyone looks at the logs you'll need this line in ee_definition.csv to see my wideband in 'user parameters'.

54,D27_LAMBDA,Pin D27 Lambda, λ,EXTENDED,8,2,0.001533,0.75,0x00,0xFF,,,,,

[A95Y]

Changing the *Stoich AFR target* to 14.1 doesn't compensate for ethanol?

[spfautsch]

You can put any AFR you want in those cells. I'm sure it impacts some fueling equation somewhere, but once you go closed loop the trims will correct you to stoich for whatever the oxygen sensors tell the ECU is stoich. With methanol that's around 6.5:1 and with bituminous coal it's more like 18.0:1. The oxygen sensors don't care what fuel you're burning - they just give 0.455 volts at precisely 0.0% oxygen concentration.

[steveo]

I'm sure it impacts some fueling equation somewhere, but once you go closed loop the trims will correct you to stoich for whatever the oxygen sensors tell the ECU is stoich.

based on that logic there's no point tuning fueling at all in closed loop range

changing that stoich AFR target will land you closer to what is stoich so closed loop doesn't have to correct as much, which is a good thing

[spfautsch]

based on that logic there's no point tuning fueling at all in closed loop range

I'm not tuning closed loop fueling, I'm trying to extrapolate my injector flow constant based on trims and PE lambda so I have the most accurate representation of this for whatever type of fuel it will run. From this flow constant I will be scaling my startup prime pulsewidth table and the proportional table of the closed loop PID controller. I don't want to bake a 3% error into these if I can help it.

changing that stoich AFR target will land you closer to what is stoich so closed loop doesn't have to correct as much, which is a good thing

I think you're missing the point of the exercise. There are millions of cars driving around today burning E10 with 14.7 as their stoich afr target, and they're all running just fine. The ECUs running these cars lack the capability to change this dynamically based on actual fuel composition. Does that mean their fueling is wrong? No, the O2s correct the trims to whatever stoich is for the fuel they're burning. Even if someone accidentally fills the tank with E85.

I'm not terribly concerned with the AFR target - it does serve a purpose but only until BLM learning is enabled. Edit: Without the BLM locker patch this will make PE fueling adjust for fuel compensation and bank to bank imbalance as intended, and that's what I plan to do.

Outside of that, what's wrong with trying to build a baseline tune that's as close to what GM would have done with these cars back in the 90s before the country became infatuated with burning corn liquor? I'm not going to live forever and my kids aren't gearheads, so this car will probably end up belonging to someone else down the line. That person may live next to a large body of water where ethanol free fuel is widely available.