RomRaider

Another update - added more specific details on how the delay value is chosen with the USDM 02-05 WRX. Also added a question that explains the reason behind the problematic delay with the USDM 04/05 WRX from the factory as compared to the USDM 02/03 WRX since there's a myth that is still floating around that it is due to higher delay values, which is not the case.

Thanks for this explanation guys

I have noted that my STi 06 UK has been holding onto the closed loop for far too long. It will sometimes hit 2 engine load before switching. I guess that is because the injector sizing has gone from std 505 to 850 now and the Closed Loop Calculated Load values are far too high.

I am not wanting it to run in OL all the time so I have therefore scaled the load figures down by a factor of 0.6 from 5900 to 3540 to get it to switch to OL when it hits 1.1 load (as in the OE map)

This was the cause of a lot of knock retard at about the 1.5-1.75 load range at 2800-3600 RPM because it was still stuck in CL mode and now the penny has dropped

It won't be OL all the time if you set the closed loop delays to zero (read the first question in the FAQ of the OP).

I got you.

Trialled this out today and it has cured ALL the transitional knock I was getting. Managed to add around 1.5 degrees of advance back into the 1.5-1.75 2800 rpm -ish area (that I had previously taken out)

I wish I had read this before

merchgod maybe, what i'm writing about is too obvious or mentioned earlier but...
I zero'd the CL/OL delay on my '06EDM FXT and was looking for the changes in the logs (and in my driving impressions). Much better throttle response and smoothness were nice improvements. Logs before/after looked quite similar except leanest AFR seen was 15.3 (0 delay) vs 15.7(stock '625' delay setting) Then I noticed that A/F Learned #1 A Jumped from -4.30% to -6.60%. My only suggestion on what can cause this is first pair of values in 'Throttle Tip-In Enrichment' Table:

Angle Change 0.0 1.0 1.9 3.9 5.9 ...
Enrichment 288 332 432 680 880 ...

I assume that with 0 delay Tip-In Enrichment Table isn't ignored anymore in CL. Is this true? I'll

zeroing the delays will not directly impact tip-in enrichment, nor will it cause you to be in open loop all the time (assuming a similar to factory fuel map in low load areas). As far as A/F Learning, it can impact range D (typically) since you are limiting the amount of time spent in this airflow range while in closed loop and therefore affecting learning for both closed loop and open loop in that range. A/F Learning range A should not be impacted by zeroing the delays unless you've changed the fuel map (from effective 14.7 AFR to richer than the minimum OL threshold) in the idle load area of the map.

No, my only change was CL/OL delay



know that, but i mean something else: as I understand tip-in enrichment table is totally ignored in CL. But there is some amount of tip-in enrichment at zero angle change. Can this amount ('raw ecu value 288') come in play if there is actually no CL/OL delay?

It shouldn't be ignored in closed loop as far as I know.

The value at zero change is only for interpolation above the minimum tip-in throttle threshold (also defined in the latest ECU defs). There's no tip-in enrichment unless tip-in throttle % is greater than this threshold.

I have also messed with this extensively. I have found that with the varying grades of gasoline around the country, our cars must be able to learn fuel trims to hit the correct real AFR that the car was tuned for. This means that you must not 0 the O/L C/L delay, or you must change your O/L fueling table to run 14.7:1 in all of the cells (and surrounding cells as the ECU interpolates) that hit up to 80 G/S airflow. To do this, you have to take your open loop fueling table and figure out which cells are going to hit the 80 G/S MAF and less. Set all of these and the lesser airflow (unless of course they are at 5k rpm or more or very high loads) to 14.7:1. FWIW, I did this for about 3k miles. It worked ok, but the engine seemed to still take a long time to learn fueling trims in the D range.

So I switched up to making a delay of 200 MS. I also made changes on when the switch occurs and modified the tables that just 0 the delay (there is one table that will 0 the delay without checking any others based on throttle in the STIs at least). I have found that these settings learn fueling trims as fast as factory, but give you a nice trade off in smoothness and fuel learning to make a more consistent tune. You can also do a Vishu reset this way and force the D range to learn and check your tune (say it leans out 0.5 AFR points and it knocks, then you know to either modify the MAF scaling or just add more fuel to your O/L fueling tables).

Since I often travel over 400 miles from my home, who knows what gas I will get and my fuel trims will adjust in the D range which is carried over into open loop. If you just simply 0 the delay, you will see that the D range fills up very slowly.

My settings are available in the v2 Stage 1 STI free Rom I have posted.

One more thing. If you are modifying MAF scaling to make it so that you get close to 0 in all learning ranges this is what you want to aim for:

A: -5% (You want the ECU to pull fuel at idle for smooth idle, this also helps with crisp response when you mash the throttle as there is a ramp to the fuel trims applied, up to 2 seconds, so if you are very positive, no matter what your D range is, you could be adding fuel when in open loop at WOT and losing power)

B: -2% (Same as above)

C: 0% (This does not matter much, but around 0 is best)

D: 0% (This is most important)

Gabe

This is interesting, I would assume once a value is learned, it is effectively the same as adjusting your MAF scale to 0%.. idling is much rougher than it needs to be.

It is effectively the same, but starting the car and initial idle will be better if you have a -5% learned A range LTFT. If you watch the ECU adjust fuel (AFC #1) at idle you will see it move around a good bit. Having things on the rich side (and letting the ECU learn to lean out) makes the car start and idle better. Also, if you are positive, when you get on the gas, there will be a ramp where postive LTFT from the A range is applied at higher flows throwing off your AFRs (making them richer than need be). This is why on some AFR dyno plots, you see a dip in AFR when the gas is mashed at 2k (some of this is tip in as well).

Gabe

So it's also 10ms per point on 32-bit ECU's? That makes the cold delay on LGT's 12.5 seconds, or rather never go into OL. Is this really right?

I think it might not be a good idea to just extend the non-enriched areas of the fuel map. Extended periods in these areas should probably go to OL, hills, lots of passengers, etc. This will never happen if the cell is just set to 14.7.

I'm also wondering if the stock O2 is good for down to even 11.5, why not set the minimum open loop enrichment to 11.5 and let the ECU run the proper fuel for a lot more of the map. How about plugin a proper wideband and run 100% of the time in CL?

Someone would have to the log the counter to get an idea for the 32-bit ECU. The delay value selection is not the same for the 32-bit ECU as the 16-bit ECU and it varies by ECU (has not been determined in detail). It should be noted that it is not a delay per se, but the period of "time" over which either primary threshold has to be exceeded continuously in order to move on to the next step.


Changing the minimum threshold to 11.5 would cause you to run in closed loop in any cell leaner than an effective 11.5:1 AFR, not open loop.

That's the intention, closed loop isn't a bad thing. I thought the reason it isn't done more often is because wideband O2 sensors have been too expensive. Since newer cars apparently already have a wideband, why only run closed loop at 14.7?