for years there have been giant threads on the pros and cons of adjusting injector latency vs tweaking the maf table for fueling errors.
hopefully this post will illustrate why they should not be considered equivalent in terms of outcome.
i have recently turned back to maf tuning after years of speed density. while some things about maf tuning are a pain in the ass, i can safely say that measuring airmass directly (or extrapolating it as the case may be) is a far simpler endeavor than attempting to compensate for the inability to "know" the exact temperature and pressure of the charge air in the chamber.
thankfully, since i have a stock maf sensor in an oem maf tube/airbox, i started with the oem translation table.
i plugged in some sensical value for injector scale and basically started raising latency until idle was stable. i was getting to values that seemed high for what i am running (dark blue wrx injectors with 65psi base pressure), but it was a good starting point, and got the car idling. the latency was much higher than the oem values i had been tuning sd with, but that was part of the reason for switching to maf--to dial in more accurate values for the injectors, and to get better alignment between SD g/s and maf g/s.
then i turned, as so many of us do, to translation table tuning.
i suggest that everyone (maf or sd) tunes open loop exclusively until things are behaving very predictably, as doing so eliminates a confounding effect of the ecu fuel trims chasing their own tail due to latency.
in all fuel tuning a primary metric is some kind of "fuel error." if you run in closed loop you rely on the oem ecu to learn the trims and pump out some numbers. for open loop fueling you don't have trims being determined for you, so you have to define your own.
in my case i've called that term "ve error" because it was used in tuning the VE table, but in any event it is
measured afr divided by
target afr.. (i also take that number and multiply by 100 to get a percent error, where 100 is perfect, 95 is 5% rich, 105 is 5% lean, and so on. the units are unimportant at this point, as we're interested in trends.)
if we just look at the fuel error vs maf voltage, things look pretty centered around 100%:
Attachment:
File comment: mafv vs fuel error
SNAG-0002.png [ 65.63 KiB | Viewed 16885 times ]
you would think that latency and scaling tuning was done, and you'd move on to maf table to get things finally dialled in. that's what i thought, and that's what i did.
i quickly realized that despite adjustments i still had excessive errors at certain operating points. after some consideration i realized that there are a great many operating points that will net the same mafv: 10psia MAP/5000 RPM is about the same as 20psia MAP/2500 RPM. you can imagine these equivalent mafvs as diagonal lines from bottom left to top right across a MAP vs RPM surface.
however--and this is the key point--the load (and therefore the IPW) is approximately doubled at the 20psi/2500rpm condition.
since latency errors are IPW based (having the most profound effect at low IPW) this means that the two operating points will have different fueling errors, and adjusting the maf translation table to correct for one will throw off the other. sure enough, this is exactly what i've seen happen.
so how can we visualize this trend? we need to look into another dimension of data. let's first consider RPM as that was one of the variables in the example:
Attachment:
File comment: rpm vs mafv
SNAG-0003.png [ 55.92 KiB | Viewed 16885 times ]
now it's a lot more obvious that for a given mafv (ie, steady mass flow rate) the fueling error follows a trend that runs leaner at low rpms (where load and therefore IPW is high) and richer at high rpms (where load and IPW is low). what is nice about this way of looking at the numbers is that your maf translation table errors are irrelevant: you are looking for trends while holding mafv steady.
the same data can be visualized by plotting against IPW directly:
Attachment:
File comment: ipw vs mafv
SNAG-0000.png [ 72.76 KiB | Viewed 16885 times ]
clearly for each row of mafv there's a trend of too rich at low IPW and too lean at high IPW: a classic symptom of latency that is set too high for the injector.
now, why had this latency issue not raised its head in a more obvious way when i was tuning SD? because with SD tuning we DO have a map (volumetric efficiency) that has a tuning axis which represents engine load/ipw, and that is MAP, so we can tune around latency issues. with maf tuning you only have the translation table and that does NOT have a load/ipw reference.
further results to follow...
a savant could just look at the numbers and give you an answer!
