O2 sensor tuning with Headers

[GHuggins]

JBZ, with all due respect, I'm 99.9% sure that this (ie. step-wise changing airflow mode) is not how it works...



Reasoning:

1) When you have a wide dynamic range and you only have a limited set of parameters to cover that range you tune these parameters in each selected operating point. Once you've done this you will cover the areas between each tuned point by interpolating. Why is that, you might ask? Because it's the best bet not having the closed loop circuit go outside stability limits. A step-wise change could very easily do this in a in-between point you haven't tuned.

2) 2D and 3D interpolation are the bread and butter of every engine controller for the same reason, one simply can't tune everything at the required resolution, this is no different than anything else. To my understanding, in most cases when it comes the ECM's the calculation involves multipliers after another. The end result would easily be extremely discontinuous leading to very rough running and likely very dangerous output results. In the early 90's when I was in the process of developing our own engine management with my friend, I studied major aftermarket ECU's very carefully, Haltech, Motec, a few UK made, etc. and some OEM stuff as well. When I was studying VW Corrado G60 Motorola 68HC11 based code, clearly written with assembler, one could only admire how elegantly the 2D and 3D interpolations were done. Yeah, then I had to graduate and get a real job and that project was never finished but it taught me quite a bit about the field that still interests me a lot.

3) The reason you've seen O2 oscillation frequency change so "rapidly" is that the control loops run at millisecond (or in some cases some tens/hundreds microseconds) levels compared to 1 - 1.4 s cycle time seen in O2 feedback, ie. plenty of time & cycles to interpolate between new airflow modes.

Barum, I'm also curious as to how you might dial all of this in? The OE has big jumps between some cells and smooth transitions between others, so just curious?
Thanks

[Michael_D]

Michael, for all intents and purposes I wonder if you can blow up each section of your good looking (anyways) O2 swings, try and find your proportional fueling for that zone in the main 02 tables and more or less copy that into the tables that are not so good. in this screenshot I blew up your injector PW (in the range that it operates at during just this portion of the log) and you can time stamp between when it goes rich/lean and when you get a reaction out of the O2's. that should give you a good baseline for integrator delay, now if you already know your proportional fueling maybe try and copy that value into the tables that are not so good, or increase/decrease (based on direction) by a marginal amount (take a stab at it) and see what it does for you.

edit BTW in this particular screenshot injector B1 (red trace) I set limits to .9 and 1.5...I think and o2 bank 1 (purple trace) I left alone

I have a pretty big monitor in my office. Big enough to open several windows with several logs open. I have my layout set up so I can then compare the logs, zooming the same amount, targeting similar rpm/load areas where I 'feel' the engine coughing. Of all my different tests, the O2 swings are the smoothest with the Proportional multiplied by 1.8. But, the car did not like this. Coughing was pretty bad, and I believe that was due to the over/under swings from commanded AFR and actual as seen by my wide band. I think the over/under correction is key to smoothing out the closed loop problem areas.

Interestingly, I am also trending AFR. I just plot commanded and actual with the same range. When I moved the O2 switching voltage from 450/450 to 550/450, the mean average trend did shift lower (richer). The engine was a tad smoother as well. That could also be another key element for dialing in O2s.

I'm beginning to wonder just how accurate my new AEM wide band is though. I ran through an open loop/VE cal, and had to drastically change everything under about 2000 rpm after all the EOIT and ETC changes I've made. Looked really good at the end (+/- 2% error). Put the thing back into CL, VE and everything I changed is now about 6% rich. I'm tired of chasing my tail with this crap.

[JBZ]

Put the thing back into CL, VE and everything I changed is now about 6% rich. I'm tired of chasing my tail with this crap.

so I have always had my wideband read a few % more rich in CL opposed to OL. wish I new why or how to fix that...maybe we can get that figured out. I have asked HPT again for more tables and parameters and specifically those ones that Ghuggins talked about in an earlier post.

so on a side note...today during my lunch break I decided I was going to try to start measuring my injector PW vs O2 swings and start getting an idea of different airflows and my integrator delays. well that didn't work out so well. I modified my chart vs time to show only 1 group. crazy thing is...I brought my MAF reading in so I could write down the airflow to correspond with the table. well it looks like the injector PW tracks airflow so tight that its hard to tell where the PW is added/subtracted for rich/lean fueling in closed loop. GEEZ

[gmorris]

I've been experimenting with O2 settings for a couple days now trying to improve idle, reduce AFR swings and push everything a tad leaner if possible. What I've discovered is that leaving the switching voltage at 450/450 and reducing bank 1/2 down to 50mv across the board did push the average AFR .1-.2 leaner. I then reduced the o2 limit from 2 all the way down to .01 and that made a dramatic difference. At cruise my AFR swings at most .1 point compared to over 1 point previously. It still swings about 1 point at idle but with 13 degrees of overlap and idling at 720RPM I expect that but the swings don't seem to affect the idle quality nearly as much as before. Overall the motor feels happier and more willing to rev than it ever has. I haven't found any negative consequences to dropping the limit so much. I also reduced both the int and prop airflow in various cells until I got the switching to smooth out pretty much everywhere.

Now I just need to figure out why my laptop decided to tank a dump and lose corrupt my last log file.
07_28 final v2.hpt

[GHuggins]

I've been experimenting with O2 settings for a couple days now trying to improve idle, reduce AFR swings and push everything a tad leaner if possible. What I've discovered is that leaving the switching voltage at 450/450 and reducing bank 1/2 down to 50mv across the board did push the average AFR .1-.2 leaner. I then reduced the o2 limit from 2 all the way down to .01 and that made a dramatic difference. At cruise my AFR swings at most .1 point compared to over 1 point previously. It still swings about 1 point at idle but with 13 degrees of overlap and idling at 720RPM I expect that but the swings don't seem to affect the idle quality nearly as much as before. Overall the motor feels happier and more willing to rev than it ever has. I haven't found any negative consequences to dropping the limit so much. I also reduced both the int and prop airflow in various cells until I got the switching to smooth out pretty much everywhere.

Now I just need to figure out why my laptop decided to tank a dump and lose corrupt my last log file.
07_28 final v2.hpt

It's amazing how the O2's can change how "peppy" something is isn't it

I think the "limit" setting is key for the swings... But I haven't tried dialing them back that much - going to have to do that for now on

The rich/lean settings - both bank and max/min - contrarily to most beliefs does control the "richness" and "leanness" averages or in other words drives the O2's to a more "prioritized" rich or lean area... You'll only ever see that in the averages...

On a side note, it sounds like most on here are having the best luck dialing things in individually or rather focusing on a particular airflow mode and dialing in just that mode - guess I've got more learning to do

[Michael_D]

Do you have cats? I'm thinking for those of us who do, we would need to be careful to not limit rich/lean switching too much. What too much is, I have no idea.

[gmorris]

Do you have cats? I'm thinking for those of us who do, we would need to be careful to not limit rich/lean switching too much. What too much is, I have no idea.

Good point, I do not have cats. Its a swap into a 69 chevelle with long tubes and full dual 3".

[topher455]

I came across this thread aging today while lurking and reading and thought even though it is gen 3 it is still pertinent. This is actually the thread I was searching for when I found this one. He basically forces the car into ol and then changes the target afr lean because of bad reporting from the o2's and then monitors bsfc.. Im understanding the theory but not how to apply it to an e38. Only one part about it bothers me and that is the contention that the car is actually richer than the narrow and wide bands report because of the fuel in the exhaust caused by overlap. Thoughts?

http://www.hptuners.com/forum/showth...hat-stock-feel

[JBZ]

I came across this thread aging today while lurking and reading and thought even though it is gen 3 it is still pertinent. This is actually the thread I was searching for when I found this one. He basically forces the car into ol and then changes the target afr lean because of bad reporting from the o2's and then monitors bsfc.. Im understanding the theory but not how to apply it to an e38. Only one part about it bothers me and that is the contention that the car is actually richer than the narrow and wide bands report because of the fuel in the exhaust caused by overlap. Thoughts?

http://www.hptuners.com/forum/showth...hat-stock-feel

I didnt read all of that thread but injector timing may play into that. I have a stock cam so other than emissions the injector timing doesnt pertain to my situation, but my thoughts are...that needs to get disl in first. Then onto o2s. When i finally get a cam that will be one of the first things i do, next to just getting it to idle decently.

[GHuggins]

Do you have cats? I'm thinking for those of us who do, we would need to be careful to not limit rich/lean switching too much. What too much is, I have no idea.

Typically OE cars will keep actual air fuel swings to .1 or .4 at the very most... I'm not so certain that having the swings dialed back in how much they go +/- would hurt the cats function? I think all that really matters is having the correct and quick swings to promote catalyst functionality... BUT I could be wrong?

[JBZ]

Typically OE cars will keep actual air fuel swings to .1 or .4 at the very most... I'm not so certain that having the swings dialed back in how much they go +/- would hurt the cats function? I think all that really matters is having the correct and quick swings to promote catalyst functionality... BUT I could be wrong?

Plus to add to that...at least in my stock tune every time i would cone to a stop (before i disabled this) the pcm intentionally goes really rich 12.1 then lean 16.0 or something as i guess a cat test.

I dont think that dialing the swings back a little would hurt, adding more boost on stock cats or a bigger cam is more likely to take out the cats than lessening the swings by a few points.

[GHuggins]

I've been experimenting with O2 settings for a couple days now trying to improve idle, reduce AFR swings and push everything a tad leaner if possible. What I've discovered is that leaving the switching voltage at 450/450 and reducing bank 1/2 down to 50mv across the board did push the average AFR .1-.2 leaner. I then reduced the o2 limit from 2 all the way down to .01 and that made a dramatic difference. At cruise my AFR swings at most .1 point compared to over 1 point previously. It still swings about 1 point at idle but with 13 degrees of overlap and idling at 720RPM I expect that but the swings don't seem to affect the idle quality nearly as much as before. Overall the motor feels happier and more willing to rev than it ever has. I haven't found any negative consequences to dropping the limit so much. I also reduced both the int and prop airflow in various cells until I got the switching to smooth out pretty much everywhere.

Now I just need to figure out why my laptop decided to tank a dump and lose corrupt my last log file.
07_28 final v2.hpt

That's the ticket Thank You..... Multiplied Limit by .01 ---

There is one "sort of" side effect that I did very randomly notice... If air fuel went "way" out - such as into the 13's or 15's - it does take just a few seconds for fueling to come back in line... SOoooo just make sure your fueling is dialed in withen 3 to 2 percent and your golden Air fuel swings at idle were 13.9 at absolute richest to 14.7 at absolute leanness and then at cruise it stayed right around where I have stoic dialed into - 14.3 to 14.4 average with .1 to .2 swings at most... I did have to go back to stock proportional and a 1.2 multiplied integrator - it was just "happiest" here and air fuel was more in line with where I was commanding stoic Beginning to wonder if blowers need to be considered as much as the length at which the O2's have been moved downstream - in other words wondering if you install headers and a blower - if you aren't better to "bias" more towards the header correction and just don't dial the integrator up "as much"...

Still obvious things to work out, but really really happy with the limit settings........

[Michael_D]

Typically OE cars will keep actual air fuel swings to .1 or .4 at the very most... I'm not so certain that having the swings dialed back in how much they go +/- would hurt the cats function? I think all that really matters is having the correct and quick swings to promote catalyst functionality... BUT I could be wrong?

That's good to know. I haven't spent much time observing "stock" engine logs....lol

But I suppose that as long as fuel goes rich/lean, the cats will be OK. I do know that a constant stoic AFR they won't last long.

Played with the limit table. First cal, I just cut the values in half, then for the second cal, I cut them in half once again. AFR swings are tighter. Could not detect any oddities. Possibly a slight improvement. I think my next test will be going back to the proportional table.

The logs files are too large to upload. Stinking road construction delays. Was caught for 25 minutes the first log, 45 the second.... Grrrrr....

[schpenxel]

I'm running stock Z06 manifolds on mine still, it usually swings +/- 0.02 lambda when in closed loop, sometimes less. It's a very tight swing (most of the time)

[GHuggins]

Cut her on down This is the only table in the gen 4's that mimicked the error table in the gen 3's - wondering if they aren't mislabeled in one gen or the other? The gen 3 had two rows of 2's in the "error table" where the gen 4 platform has a single row of 2's in the "limit table"... The gen 3 I had cut down to .055 through multiplying the difference in correspondence to the proportional and integral tables - per gmorris I cut the gen 4 down to .01 - it was very nice... Owner was extremely happy with how it turned out and drove - he's especially happy with the power now

[schpenxel]

What'd you end up at power wise?

[GHuggins]

That was the one that made 892 to the tire He came by and drove it today... Don't think I want to ride with him again, but he was pretty happy...

[JBZ]

so im going to have to try cutting the proportional gains. ive landed on that table so many times but didn't know what it did so I haven't messed with it. multiplying it by .01 will reduce it greatly. mine is still stock value of 2.0 so were talking about .020 after that cut. also going to try to reduce the bank 1 and 2 switching as well.

so I did try out some new settings and its not bad at all. last night I back calculated my O2 swings with lean/rich with Injector PW. kind of a bitch to do, it took me at least an hour 1/2. this was mainly to try to calculate my integrator table and based on what I calculated (at least 5 lean/rich swing per cell, per bank then averaged them out) the integrator delay was way off. I tried to get the airflows right in the middle of the cells that I posted earlier to most match the middle of each cell, the higher cells from 56 up were damn near impossible to get long enough data in to make a calculation so I guessed. I didn't really do anything else closed loop wise other then move my proportional all 1 cell to the right as I idle in the 8 cell...hadn't really thought about doing that before.

I try all the above mentioned settings one at a time and see how they pan out.

these are my new settings and are not bad at all so far, if you take a look back at what I had in there in post 159 the integrator delay is a lot different here.

[GHuggins]

Looks like you increased integrator and decreased proportional?

[JBZ]

Looks like you increased integrator and decreased proportional?

Yeah i increased the integrator alot and porportional i didnt decrease nessassarly but yes kind of other than cell 8 i increased by shifting all my data i had one cell to the right. Of course from 16 to 40 it decreased l, everything beyond that is unchanged.

Edit:no sorry my mistake, it all decreased to cell 40. I was not looking at the first cell correctly taking out a zero. Yes decreasing by moving to the right.