Forum » EFI Tuning Fundamentals » Fuel temperature compensation and ECT/IAT Bias correction values?

Fuel temperature compensation and ECT/IAT Bias correction values?

EFI Tuning Fundamentals

Discussion and questions related to the course EFI Tuning Fundamentals


Page 1
Author
118 Views

Temperature compensations.. The bane of my existence lol.

I've got a later model Subaru EZ30 with an elite 2500 in it, and in the 07+ ones they run a return-less fuel system. This tends to heat up the fuel quite significantly because the rails sit on top of the motor (Flat 6)

So, to help combat this I installed a fuel temp sensor in the fuel line right before the rail itself.. Seems to work a treat, the rail fuel temp is substantially hotter than what the flex sensor reads; particularly at low vehicle speeds (Low airflow and low fuel consumption)

I'd like to try and work out the required density values to combat AFR drifting lean as fuel temps heat up.

Using a spreadsheet, I transposed the values in the Link ECU's https://i.imgur.com/z5CiJIk.jpg but they didn't seem to be anywhere near aggressive enough to combat the rise in fuel temp (At idle at least). The Motec M1 GT86 values suggested a change of 9% per 10c which is a little better, still doesn't seem to be quite enough.

Here's what I did.

I set the ECT Bias so at idle there's almost no change in calculated air temperature as the MAT sensor heats up. So in theory, the only thing leaning out the mixture would be the fuel temperature. If anything, the actual charge air temp would be decreasing in density and going richer..

That aside..It seemed at idle about a 15% change in density per 10c was required to remain on target as the engine bay heatsoaked. My question is

would this value likely need be less aggressive under larger fuel flows (Maybe closer to the 9% Motec suggested) What's the theory here?

For all I know Haltech, Link, and Motec all calculate fuel density differently into their fuel model (Hence the different numbers)

You'd think testing this would be simple, go for a drive on a hot/cold day and log the particular cells steady state.. But because I don't know if my ECT/IAT bias is accurate I can't know who's to blame.. Is it the hotter intake air, or is it the hotter fuel? I can't set my ECT/IAT bias because I need to maintain a consistent fuel temperature to work that out.

My only idea was to have it on the dyno steady state and run the fuel line through a oil cooler submerged in ice water.. Take the cooler out of the ice and monitor AFR drift as the fuel temps return to a more typical 20c+ above ambient.

That or a very dangerous Aquarium chiller in the passenger seat to maintain a set fluid temp :P (I'm not being serious btw)

I know it's a lot to take in, but would appreciate any input.

Any suggestions, and known good values?

Cheers - Jaydn

So are you using fuel feed back (closed loop fuel) ? Typically the goal here is to minimize your feedback within an acceptable range (keep the fuel trims down). You can drive yourself nuts trying to get what are basically simple models and dumb compensation factors "correct." It's not really intended to be a PhD level thermodynamic model. That wouldn't be practical anyway because it would be even more complicated to tune.

I am. The Haltech has 4D fuel table learning (Which i'm currently only using 3D) I'm just trying to help the long term trim table stop from chasing it's tail every time there's a mild environmental change. If I could get at least one dialed in, I'd just set the 4th axis on the LTT to either IAT or Fuel temp and let it figure out the correction required over time.

Yeah I know that in my heart, I'm just being OCD about AFR. I know realistically variations have minimal performance impact, it'd be nice if the ECU could eventually go "I'm targeting 14.3 here, temp is x, fuel temp is w, rpm is y, load is z.. Yup, we're within 0.5AFR

If you're closed loop it will converge on the target (just not instantly) if your lambda control gains are set right. If it won't achieve the target (or get very close) the gains could be too fast (overshoot) or you need more integral gain or whatever to converge within a tighter band of the target.

Not saying the closed loop system is overshooting or reacting too slow :)

I'm more trying to reduce the LTFT error's from day to day.. So it can refine itself instead of constantly fixing itself due to conditions.

Example.. Let's say that 2500RPM at MAP '-9psi' with a fuel rail temp of 55 and an MAT of 35 degrees.. The correction may be 5.8 learned over a period of steady state driving.

Next day it might be a colder day and fuel temp is 35 and MAT is 25, that learned value might change to lets say 2.5 instead of 5.8.. Which correction is the cause? Is it the fuel temp correction that's over compensating, or is the the CAT/IAT bias?

I know I could probably leave it at "Near enough is good enough" but I'd like to try :)