PI : DI Ratio Discussion

[thambu19]

Originally Posted by steve99

Yes you would think it should. but others have found it not always so.

their was a heap of discussion on it here

http://www.ft86club.com/forums/showthread.php?t=53064

http://www.ft86club.com/forums/showp...5&postcount=12

http://www.ft86club.com/forums/showthread.php?t=3172


jamesm who is a tuner, no longer on this forums suggested this proceedure

http://www.ft86club.com/forums/showp...&postcount=195


ztan found some pi/di fueling error
http://www.ft86club.com/forums/showp...1&postcount=37

So the way we get around is by scaling the MAF to reduce fuel trims? Does not look like the most straightforward approach to me. Maybe this is the only way to get fuel trims right until someone discovers a table or something that should be changed that will make the ECU adjust fuel pulsewidths appropriately when PFI/DI ratio is changed.

[thambu19]

@arghx7 would you know anything about this?

[arghx7]

There's a lot of discussion here among different people and scenarios that I'm trying to unravel.

First question: is this a matter of the fuel trims moving around more than would be ideal or is there an actual driveability, knock, etc problem?

Second: Are we talking about stock port injectors, obviously stock DI injectors, and stock MAF scaling or scaling that has been judged ok before the PI and DI balance was changed?

I suspect there's a bunch of wall wetting compensation, gas flow model, or injector flow compensation that is poorly understood because we don't have actual block diagrams of how the software works, just whatever has been reverse engineered.

My position is, generally speaking and subject to case by case revision--if the fuel trims are just moving around more who cares? That's their job. They're compensating for some inscrutable fuel calculation that hasn't been unraveled, or more likely, production tolerances in injectors, given that the stock values were for some mean spec part most likely.

Is there an actual problem with the car here or are we just trying to make the fuel trim lines on an excel graph look pretty

[thambu19]

Originally Posted by arghx7

There's a lot of discussion here among different people and scenarios that I'm trying to unravel.

First question: is this a matter of the fuel trims moving around more than would be ideal or is there an actual driveability, knock, etc problem?

Second: Are we talking about stock port injectors, obviously stock DI injectors, and stock MAF scaling or scaling that has been judged ok before the PI and DI balance was changed?

I suspect there's a bunch of wall wetting compensation, gas flow model, or injector flow compensation that is poorly understood because we don't have actual block diagrams of how the software works, just whatever has been reverse engineered.

My position is, generally speaking and subject to case by case revision--if the fuel trims are just moving around more who cares? That's their job. They're compensating for some inscrutable fuel calculation that hasn't been unraveled, or more likely, production tolerances in injectors, given that the stock values were for some mean spec part most likely.

Is there an actual problem with the car here or are we just trying to make the fuel trim lines on an excel graph look pretty

@arghx7 @solidONE @steve99 @Kodename47 @ztan

My concern is if the fuel trims go up or down too much say more than 10% it will take the estimated load along with it. Estimated load = Originally assumed load through MAF signal +- corrections from fuel trim. This means the cam and spark lookup will move around unnecessarily?

I dont know how exactly it works on the FA20 so this is just my assumption. This is assuming that most OEMs trust their fuel modeling more than air flow modeling. So when fuel trim moves the load estimation moves as well.

My other concern is how is MAF scaling going to fix this? For example say if by going 100%PFI at low loads causes the LFTF there to go 10% +ve causing a load estimation to go up 10%. The MAF there would be moved up 10% at the same Voltage to bring the LTFT to 0% but we still have the same problem because now the MAF readings are higher (artificially) causing a higher load estimation and hence spark/cam lookup = bad combustion or knock depending on which direction we go in load lookup

[arghx7]

Originally Posted by thambu19

@arghx7 @solidONE @steve99 @Kodename47 @ztan

My concern is if the fuel trims go up or down too much say more than 10% it will take the estimated load along with it. Estimated load = Originally assumed load through MAF signal +- corrections from fuel trim. This means the cam and spark lookup will move around unnecessarily?

I dont know how exactly it works on the FA20 so this is just my assumption. This is assuming that most OEMs trust their fuel modeling more than air flow modeling. So when fuel trim moves the load estimation moves as well.

You're giving Subaru too much credit. It's not that smart, or at least, not that anyone has proven. Unless someone here wants to link to somewhere showing the load estimation being related to fuel trims. It certainly never did that on the EJ engines.

The code is really not that sophisticated. There's no combustion model based spark control either--or if there is, it doesn't matter enough that anyone has bothered trying to understand it. Have you seen the cooled EGR spark controls on the FA20DIT? It's just a dumb look up compensation table, or at least the one that has been found so far seems that way. No residual gas model aspect to it.

My other concern is how is MAF scaling going to fix this? For example say if by going 100%PFI at low loads causes the LFTF there to go 10% +ve causing a load estimation to go up 10%. The MAF there would be moved up 10% at the same Voltage to bring the LTFT to 0% but we still have the same problem because now the MAF readings are higher (artificially) causing a higher load estimation and hence spark/cam lookup = bad combustion or knock depending on which direction we go in load lookup

MAF scaling has always been strategic tail-chasing in this context.

[KoolBRZ]

I have a totally different theory. I think the rich condition in CL is caused by sudden "Tip-in" enrichment, which affects STFT, becoming LTFT. This is then passed on to OL. I think Tip-in should have a lower threshold, and smoother transitions, so it wouldn't adversely affect fuel trims in CL, and then corrupting OL. It doesn't require any mysterious ECU reverse engineering, and it is a fuel source outside of usual channels unaffected by the MAF or MAP sensors. What do you think @thambu19?

[KoolBRZ]

Here are my current Tip-in thresholds above, compared to the OFT OTS map thresholds below. As well as my current Tip-in Enrichment A, and B tables, above, compared to the OFT OTS A and B tables below. I have tried smoothing out the transition between throttle angles 5.86%, and 7.81%, because there is such a jump there, but it didn't work as well when I drove it. So, I just stuck with smoothing the transitions under 5.86% and lowering the thresholds. Do you think the stock configuration of the Tip-in Enrichment and Thresholds could be the source of the problem?

[solidONE]

Originally Posted by arghx7

There's a lot of discussion here among different people and scenarios that I'm trying to unravel.

First question: is this a matter of the fuel trims moving around more than would be ideal or is there an actual driveability, knock, etc problem?

Second: Are we talking about stock port injectors, obviously stock DI injectors, and stock MAF scaling or scaling that has been judged ok before the PI and DI balance was changed?

I suspect there's a bunch of wall wetting compensation, gas flow model, or injector flow compensation that is poorly understood because we don't have actual block diagrams of how the software works, just whatever has been reverse engineered.

My position is, generally speaking and subject to case by case revision--if the fuel trims are just moving around more who cares? That's their job. They're compensating for some inscrutable fuel calculation that hasn't been unraveled, or more likely, production tolerances in injectors, given that the stock values were for some mean spec part most likely.

Is there an actual problem with the car here or are we just trying to make the fuel trim lines on an excel graph look pretty

1. Aside from increased LTFT causing the car to run richer than desired OL operation under certain conditions (average to cooler temps) there is not much perceptible drivability issues or increase in knock that I've noticed.

2. I don't think the stock MAF scale has been very good for anyone that I can remember. For my case, it's a maf scale that worked well while running less PI (more like the table below), but experienced a good amount of increase in fuel trims in areas where Port injector ratio was increased. This somehow led to an increase in LTFT in OL operation causing it to run even more rich than before the changes to the port injector ratio tables.



One would suspect one would spend less time "chasing one's tail" if I simply modified the port ratio table to run 100% DI at .20 load and lower, but where is the fun in that right? lol

It's seems that many of the guys that have already done some "tail chasing" had pretty good success balancing the port injectors to the direct injectors after dialing their MAF scale while running 100% DI. This is what I've decided to do now, but I will take it one step further by taking the IAT compensation into account. I've dialed my MAF while running 100% DI at IAT of 66~70*f by making sure my IAT don't go below 66*f throughout the entire range and not using any of the data with intake temps above 70*f. I think this will give me a good base before making changes to the IAT compensation and port injector scaling.

[solidONE]

Originally Posted by thambu19

@arghx7 @solidONE @steve99 @Kodename47 @ztan

My concern is if the fuel trims go up or down too much say more than 10% it will take the estimated load along with it. Estimated load = Originally assumed load through MAF signal +- corrections from fuel trim. This means the cam and spark lookup will move around unnecessarily?

I dont know how exactly it works on the FA20 so this is just my assumption. This is assuming that most OEMs trust their fuel modeling more than air flow modeling. So when fuel trim moves the load estimation moves as well.

My other concern is how is MAF scaling going to fix this? For example say if by going 100%PFI at low loads causes the LFTF there to go 10% +ve causing a load estimation to go up 10%. The MAF there would be moved up 10% at the same Voltage to bring the LTFT to 0% but we still have the same problem because now the MAF readings are higher (artificially) causing a higher load estimation and hence spark/cam lookup = bad combustion or knock depending on which direction we go in load lookup

Fuel trims, on my car, has not increased beyond 9% (idle LTFT) after changing the port injection ratios. I'd say at most +4%~6% LTFT in the "problem" areas aside from idle speeds. My main concern it the fluctuations in fuel trims seems to be bumping up the LTFT in OL operation causing it to run a bit richer than before. I wouldn't call it a 'huge' concern, but it is something I think I'd address. If anything, I just want to mess with it and see how close I can get it to requested AFRs as I can.

[thambu19]

Originally Posted by solidONE

Fuel trims, on my car, has not increased beyond 9% (idle LTFT) after changing the port injection ratios. I'd say at most +4%~6% LTFT in the "problem" areas aside from idle speeds. My main concern it the fluctuations in fuel trims seems to be bumping up the LTFT in OL operation causing it to run a bit richer than before. I wouldn't call it a 'huge' concern, but it is something I think I'd address. If anything, I just want to mess with it and see how close I can get it to requested AFRs as I can.

Id think 4-6% LFTF isnt a concern at CL but at OL yes maybe. Some ppl tried going 100% DI at loads above 0.7 for knock relief they also said their fuel trims went up. I am wondering if it is because at 100% DI vs earlier 50% DI the charge cooling is increasing the VE and hence the LTFT. If that is the case the engine should run the same AFR at OL as before. Since you said the LFTF increased and made the engine run Rich in OL id suspect that the LTFT was not supposed to go positive but did due to PW not scaling appropriately with PFI/DI ratio change.

Those who have had this issue can you chime in?
Did you see higher LTFT in areas where you increased PFI or in areas where you increased DI?
The LTFT learnt at low loads should not affect OL. Usually it is the LTFT just below OL that gets carried into OL. So if the idle LTFT is off nothing to worry there, atleast not so much

[Kodename47]

At this point in time we aren't sure quite how the LTFT ranges work. There are definitely bands but the standard defined ones don't seem to quite tie in. I think @solidONE has been testing something but nothing conclusive.

[thambu19]

Originally Posted by Kodename47

At this point in time we aren't sure quite how the LTFT ranges work. There are definitely bands but the standard defined ones don't seem to quite tie in. I think @solidONE has been testing something but nothing conclusive.

Okay. The engines Ive looked at not (FA20) the last row of load in CL is what gets used in OL. Are our engines using Wide band O2 sensors/AFR sensors or jus regular switching O2 sensors?

[ztan]

Originally Posted by arghx7

There's a lot of discussion here among different people and scenarios that I'm trying to unravel.

First question: is this a matter of the fuel trims moving around more than would be ideal or is there an actual driveability, knock, etc problem?

Second: Are we talking about stock port injectors, obviously stock DI injectors, and stock MAF scaling or scaling that has been judged ok before the PI and DI balance was changed?

I suspect there's a bunch of wall wetting compensation, gas flow model, or injector flow compensation that is poorly understood because we don't have actual block diagrams of how the software works, just whatever has been reverse engineered.

My position is, generally speaking and subject to case by case revision--if the fuel trims are just moving around more who cares? That's their job. They're compensating for some inscrutable fuel calculation that hasn't been unraveled, or more likely, production tolerances in injectors, given that the stock values were for some mean spec part most likely.

Is there an actual problem with the car here or are we just trying to make the fuel trim lines on an excel graph look pretty

First: Getting DI and PI systems balanced made a driveability difference to my car - just went smoother.

Second: You would think that the Subaru engineers would be quite close with their stock MAF scales and fuelling algorithms. Some cars seem to be close to the mark, others not so. I was happy leaving the DI system stock on the hope that the engineers did their job, scale the MAF to that, then dial in PI system. I had planned to upsize my port injectors so it made sense for me to do it in that order to have a stable base to calibrate my port injectors to.

Third: I agree, there is a huge amount in the code that is not defined. I did disassemble the Tau transient fuelling tables, but have not been able to make sense of how to change them logically, so I've left them alone.

Finally: I'm happy to let the trims to their work and I have given up trying to control them around idle speeds, but I also like a pretty graph, even if the information it gives me is erroneous.

[ztan]

Originally Posted by thambu19

@arghx7 @solidONE @steve99 @Kodename47 @ztan

My concern is if the fuel trims go up or down too much say more than 10% it will take the estimated load along with it. Estimated load = Originally assumed load through MAF signal +- corrections from fuel trim. This means the cam and spark lookup will move around unnecessarily?

I dont know how exactly it works on the FA20 so this is just my assumption. This is assuming that most OEMs trust their fuel modeling more than air flow modeling. So when fuel trim moves the load estimation moves as well.

My other concern is how is MAF scaling going to fix this? For example say if by going 100%PFI at low loads causes the LFTF there to go 10% +ve causing a load estimation to go up 10%. The MAF there would be moved up 10% at the same Voltage to bring the LTFT to 0% but we still have the same problem because now the MAF readings are higher (artificially) causing a higher load estimation and hence spark/cam lookup = bad combustion or knock depending on which direction we go in load lookup

I haven't been able to find exactly where fuel trims get applied, but engine load calculation comes quite close to the beginning and all the trim calculations that I have seen seem to be calculated referencing a figure that has already been finalized. I suspect the trims do get applied to fuelling only and do not affect load.

Changing MAF scale does shift load cell - this should be finalized (or stock values accepted as a reference) first before anything else is touched, as most of the other values reference load. Changing MAF scale later in the process screws around with everything under it in the code.