Individual Cylinder Trims Discussion

[Terminal_Crazy]

I've read a bunch about your project searching for ideas for mine - it may be worth the attempt. I was pretty frustrated with the parking lot manners in SD/CL but the difference is night and day with the MAF. I have no idea why, but I'm not going to look that gift horse in the mouth. I was on the fence as to whether it was worthwhile to finish installing the A/C piping and charge it, but I'm thinking it'll handle it now.

The only issue I can't seem to resolve is surge / bucking in gear at closed throttle. I know the fix, but don't want to take idle timing out because at 29 the lope turns into a "dead cylinder" feel that shakes the car while at 30 you can barely feel it.

I might switch the MAF back of again then next time i go out.
Mine just felt uptight like itr was being held back.
Timing was around 39. Idle was really smooth but it didn't drive nice. SDOL fattened it up but it cruised nicely evreywhere else.
Increasing the timing smoothed idle, reduced Ibpw, reduced MAP etc etc everything we are looking for.... just drove a bit jerky.

I've since raised my O2 swing thresholds up. It passed the emmissions test like that and drives much better with the lowered timing.

What does your int delay table vs airflow, and corrcl vs airflow look like?

By my estimation the O2s are only around 6-8 inches further from the exhaust ports with mine (stainless works C4). I'm having trouble finding pics of your headers to compare.

[/QUOTE]
Fast 02 Rich/Lean
0 451
16 438
32 424
48 424
64 424

o2 Int Delay - Airflow
0 0.80
16 0.66
32 0.46
48 0.43
64 0.38

CorrCL - Airflow
0 08
16 26
32 26
48 26
64 32

I'll stick a pic of the headers on my website.
http://www.sand-hill.uk/TerminalCraz...Length_004.jpg

The port for the O2 sensor is basicaly at the end of the header pipe rather than in the collector.


Mitch

[spfautsch]

DISCLAIMER: I don't have the slightest clue as to what I'm doing, and have excessive amounts of what I hope is largely false knock. Nor do I have a wideband. So if you follow my advice you might need to have your head examined.

I did pull my plugs a few days ago and none show signs of detonation.

Your int delays seem excessive for such a short header.

I also found things stabilized somewhat when I took most of the "teeth" away from CORRCL.

This is what I'm currently running with the header below.

INT DELAY vs AIRFLOW
725
212
150
125
112

CORRCL vs AIRFLOW
3.14
4.31
9.02
12.55
12.55

0304171713.jpg

As to your comment about feeling "held back", I have also noticed the throttle response isn't quite as crisp as it was when running SD. But if my splits never come back I'll deal with it.

[steveo]

having an integrator delay a bit on the long side doesn't really seem to hurt anything.

As to your comment about feeling "held back", I have also noticed the throttle response isn't quite as crisp as it was when running SD. But if my splits never come back I'll deal with it.

that's typical. the map sensor, tps, and everything else other than the maf responds damn near instantly. ve can tend to provide a natural accelerator pump effect, since vacuum often drops then recovers from quick throttle movements. the maf on the other hand, has a bit of a delay. i think it uses tps to try to account for it (i think kur4o was working on trying to tune that).

[spfautsch]

I want to argue with you on the integrator delays, but I'm obviously doing something wrong. The more logging I do in warm temps the more fuel it's taking out (BLMs dropping when ambient temps rise). I'm starting to think my false knock isn't all false. I also filled up tonight and calculated 21.9 mpg over the last two days. That seems highly improbable @ stoich with a cam with 50+ degrees of overlap.

I've already bumped the O2 swing voltages somewhat.

00	525	525
16	525	525
32	459	459
48	459	459
64	459	459

My guess is the slow nature of the MAF has helped reduce the splits by stabilizing the fueling calcs whereas in SD the MAP would fluctuate due to reversion at low airflows.

[steveo]

I want to argue with you on the integrator delays, but I'm obviously doing something wrong.

fuel is sprayed. o2 reads oxygen content. ECM responds by adding/subtracting fuel.

in this particular control system loop, you definitely need a decently sized delay, otherwise the change happens so quickly that you end up modulating.

the only side effect of being too long is that changes take longer to occur.

so what?

a few percent of fueling error never hurt anything.

The more logging I do in warm temps the more fuel it's taking out (BLMs dropping when ambient temps rise)

warmer air is less dense. humidity also plays a factor. warmer air requires less fuel. the IAT temperature compensation is imperfect. it's normal for trims to change during things like weather fluctuations.

don't get too OCD about keeping your trims perfect. let closed loop do SOME of the work. that's why it's there. otherwise just turn it off like i usually do

IMO........in reality, engines don't actually need or want a stoich mixture most of the time. i suggest you try driving while adding/subtracting a bunch 'o fuel so you get a feel for how much AFR actually affects an engine. staring at these blm numbers too long makes you lose sense of that. just because o2 sensors only read stoich doesn't mean your engine really cares how close you are to it, and a truly awesome tune definitely doesn't have to hover around stoich for most of its life.

I've already bumped the O2 swing voltages somewhat.

don't go too far. the o2 swing voltages should ideally be the actual switching @ stoich point of the sensors as seen by the ECM. the further you stray from that value the less stable things get.

[spfautsch]

in this particular control system loop, you definitely need a decently sized delay, otherwise the change happens so quickly that you end up modulating.

the only side effect of being too long is that changes take longer to occur.

so what?

I suppose I was looking at the integrator delay purely from the standpoint of the engine as an air pump, and that we adjust it to compensate for the increased length of tube headers. Correct me if I'm wrong but isn't that only one (relatively insignificant) aspect of what we're accomplishing by increasing int delay?

a few percent of fueling error never hurt anything.

I would have to revert back to SD to get a more exact #, but it looks like close to 15% over the past few days. I'm definitely running into what I would describe as debilitating lean spark knock in the afternoons in a load range that might only generate ten to twelve knock counts per hour in the morning. It was so bad yesterday afternoon I thought my one wheel bearing that I'd found to be sloppy had started growling. Then I looked down at datalog and noted KR pegged at 12 degrees and KC rolling like the wheels in a slot machine. IATs have been 82f with 70%rh in the AM and 108f with 50% in the PM.

I was thinking this was false knock from (roller rockers, headers, connecting rod clap, ???) because I use my cruise control a lot, and that causes the knock events to graph in a line on the RPM I'm cruising at. <planting face in palm>

Whatever the case, the int delay discussion caused me to think a bit about the closed loop parameters and how they're all using a derived airflow scalar from the reference table. I'm wondering what the program uses when the MAF is disabled / absent. Does it fall back to 00, and was it using my int delay and reference voltage from the 00 cells? That would certainly explain why switching my MAF back on has caused caused trims to shift. Was it Rob that disassembled EE?

the IAT temperature compensation is imperfect.

How would IAT compensation figure into closed loop other than for base calculation? Don't the O2 sensors have the final say?

Edit: Further in that direction, isn't IAT pretty much irrelevant when using an air meter (MAF)?

[spfautsch]

One other thing I've been wondering, does anyone know how long the program samples the O2 sensors after the int delay timer has expired?

My understanding of how the closed loop metering works is that the sensor outputs a waveform somewhat similar to a sine, and the feedback routine read the peaks of the wave and divides by 2 to find the average?

Edit1: The idea I'm arriving at logically is that the engineers that wrote EE didn't intend the integrator delay to have anything to to with relative placement of the O2 sensors. My theory is it's used both as a delay to let the fueling changes settle in, and possibly also the oversampling window for reading the AD converter(s). But without disassembly notes or source code that's simply wild speculation. All I know is if I was building an opposed or crossplane cylinder engine controller with an arduino, that's how I'd do it - alternate between reading bank 1 while bank 2 is stabilizing and vice versa.

Edit2: Further thinking has reminded me that CORRCL is working in between integrator "cycles" so the program must be reading both O2s all the time.

[spfautsch]

in this particular control system loop, you definitely need a decently sized delay, otherwise the change happens so quickly that you end up modulating.

Would this be observable in any way? Fluctuating INT counts?

Tried raising O2 reference voltage in the cells I hadn't already and removing more from CORRCL, and the only place it seemed to add fuel was in cell 16.

Another thing I noticed is that somehow EGR was still coming on at low duty cycles even though I had the enable RPM well above the range it was coming in at, adding a degree of unwanted timing and further diluting the charge. I just zeroed the egr adder timing and set the enable RPM to 7000. Will try to log more before it cools off this evening.

[lionelhutz]

The O2 doesn't output a sine wave and the algorithm doesn't just read the peaks and divide by 2. The output of the O2 is directly read and where it's voltage goes is a result of the control loop operating. Each time the O2 reverses direction it's because the control loop reverses the direction it's going with the fuel. In other words, the control loop adds fuel until the O2 swings to a high voltage and then it starts pulling fuel until the control loop swings to a lower voltage. A control loop is never perfectly stable. It always oscillates slightly around the desired setpoint.

If the integer acts too fast then it can pull fuel past the point where it should start adding fuel again. This happens because the O2's don't see the leaner fuel quick enough so they don't give feedback quick enough to tell the integer to "turn around" in time. So, a longer time makes the integer go slower and it is all more stable.

On a '7747 system, I observed a light throttle acceleration surging caused by the lean swing. I adjusted the O2 switch voltages slightly to stabilize that one.

Going back to your original intent. I can't see you having any sure-fire way to get the individual trims right unless you put 8 wideband O2 sensors on it. Possibly thermocouples could work too. IR sensing is way too sketchy.

[spfautsch]

warmer air is less dense. humidity also plays a factor. warmer air requires less fuel. the IAT temperature compensation is imperfect. it's normal for trims to change during things like weather fluctuations.

I knew / know this happens and I've observed it, but I was a bit in freak out mode because it felt like it was running extremely lean and this is a new / replacement MAF sensor. I killed the 22 year old AC Delco cleaning it, and I had no reason to be overly confident the replacement would be identically calibrated.

I wish I'd noticed the EGR trying to enable sooner, it would have explained a lot. I assumed setting the enable RPM as high as possible and disable at 0 would kill it completely, but it still attempts to enable and adds spark before switching off.

Edit: I wonder if it would be worthwhile patching the EGR out completely, so as to save a few CPU cycles?

don't get too OCD about keeping your trims perfect. let closed loop do SOME of the work. that's why it's there. otherwise just turn it off like i usually do

I do understand we're trying to re-shape the VE tables, not necessarily have BLMs at 128 everywhere. I guess the goal is to either log only in ideal conditions (still trying to determine what exactly that is) or have equal amounts of logging in extreme conditions and let the law of averages find the middle.

[steveo]

I wish I'd noticed the EGR trying to enable sooner, it would have explained a lot. I assumed setting the enable RPM as high as possible and disable at 0 would kill it completely, but it still attempts to enable and adds spark before switching off.

could be a testing routine. 'vette calibrations have lots of extra tests to feed those pre-obdii pids. try disabling the egr error code and see if it goes away. with that cam an egr is kinda pointless, might as well just block it off, if for no other reason that it's ugly.

edit: often disabling the error code disables the test routine too. egr test involves popping the egr open for a moment during conditions when it's not usually operated, but it's unlikely the driver will notice and when vacuum is steady, and monitoring MAP for a slight drop in vacuum.