An Idiot Learns to Tune, Part 2

I've been working on the tune for my car for quite some time. A slipped optispark hub, an unfinished ignition controller project, and life in general has slowed the progress immensely, but I've arrived at a point where I feel like it's 95% perfect and I finally have all the pieces to the puzzle together (development halt on the ignition controller and working + accurate wideband). Last fall I decided when I got to this point I wanted to run a tank of ethanol free gas through it to see what happened to trims and PE AFR. Here in the Midwest it's difficult to buy gasoline that doesn't contain "up to 10%" corn liquor. Since I did most of my tuning on a steady diet of 93 octane E10, I wasn't sure what to expect, but the shift was quite distinct so I thought I'd share it with anyone interested. Here's a quick comparison running the same tune for all the log data.

Here are the logs if anyone's interested. [link] Weather conditions were fairly consistent - hot and humid. The ones with "E0" in the filename were captured running on 10 gallons of roughly 90 octane pure dino juice. This is a MAF tune - since I didn't have a really accurate injector constant for the modified LS2 injectors I have, I made the assumption the factory MAF calibration is accurate and tried to determine the flow constant with trims and / or wideband data.

E10

http://www.pfautsch.com/wp-content/u...ms-150x150.png

http://www.pfautsch.com/wp-content/u...is-150x150.png

http://www.pfautsch.com/wp-content/u...da-150x150.png

E0

http://www.pfautsch.com/wp-content/u...ms-150x150.png

http://www.pfautsch.com/wp-content/u...is-150x150.png

http://www.pfautsch.com/wp-content/u...da-150x150.png

It's difficult to see from the graphs, but the PE lambda is fluctuating between 0.86-0.88 on the E10, and 0.83-0.85 on E0 (BLM locker patch in use).

Essentially this is a very verbose and graphic presentation of the simple idea that E10 is stoich at about a 3% richer mixture than pure gasoline. What I learned from this is that when tuning with E10 it's best to try and have your trims adding around 2%, whereas I was targeting trims that were removing about 2%. This doesn't seem like a lot, but with the E0 my cam surge is far more pronounced, so I think I'm going to scale my injector constant up a bit and see how it reacts.

The cam surge is about the only aspect of the tune I'm not really happy with. Once I have the injector constant so E10 produces BLMs in the neighborhood of 130-132, I'm then going to go back to fine tuning individual cylinder trims to see if there's any improvements to be made. I finally broke down and bought a high quality IR heat gun and after taking measurements on a couple different days I'm seeing #2 seems to be running hotter than the other 7 holes. I'm not sure if this means it's rich or lean, but at least I have an idea which cylinder might be the culprit in causing a slightly lean condition on the right bank.

I wanted to add that if anyone looks at the logs you'll need this line in ee_definition.csv to see my wideband in 'user parameters'.

Code:

---

54,D27_LAMBDA,Pin D27 Lambda, λ,EXTENDED,8,2,0.001533,0.75,0x00,0xFF,,,,,

---

Changing the *Stoich AFR target* to 14.1 doesn't compensate for ethanol?

You can put any AFR you want in those cells. I'm sure it impacts some fueling equation somewhere, but once you go closed loop the trims will correct you to stoich for whatever the oxygen sensors tell the ECU is stoich. With methanol that's around 6.5:1 and with bituminous coal it's more like 18.0:1. The oxygen sensors don't care what fuel you're burning - they just give 0.455 volts at precisely 0.0% oxygen concentration.

Quote:

---

I'm sure it impacts some fueling equation somewhere, but once you go closed loop the trims will correct you to stoich for whatever the oxygen sensors tell the ECU is stoich.

---

based on that logic there's no point tuning fueling at all in closed loop range

changing that stoich AFR target will land you closer to what is stoich so closed loop doesn't have to correct as much, which is a good thing

Quote:

---

Originally Posted by steveo

based on that logic there's no point tuning fueling at all in closed loop range

---

I'm not tuning closed loop fueling, I'm trying to extrapolate my injector flow constant based on trims and PE lambda so I have the most accurate representation of this for whatever type of fuel it will run. From this flow constant I will be scaling my startup prime pulsewidth table and the proportional table of the closed loop PID controller. I don't want to bake a 3% error into these if I can help it.

Quote:

---

Originally Posted by steveo

changing that stoich AFR target will land you closer to what is stoich so closed loop doesn't have to correct as much, which is a good thing

---

I think you're missing the point of the exercise. There are millions of cars driving around today burning E10 with 14.7 as their stoich afr target, and they're all running just fine. The ECUs running these cars lack the capability to change this dynamically based on actual fuel composition. Does that mean their fueling is wrong? No, the O2s correct the trims to whatever stoich is for the fuel they're burning. Even if someone accidentally fills the tank with E85.

I'm not terribly concerned with the AFR target - it does serve a purpose but only until BLM learning is enabled. Edit: Without the BLM locker patch this will make PE fueling adjust for fuel compensation and bank to bank imbalance as intended, and that's what I plan to do.

Outside of that, what's wrong with trying to build a baseline tune that's as close to what GM would have done with these cars back in the 90s before the country became infatuated with burning corn liquor? I'm not going to live forever and my kids aren't gearheads, so this car will probably end up belonging to someone else down the line. That person may live next to a large body of water where ethanol free fuel is widely available.

I just ran the last few drops of E0 out of the car and refilled with fresh 93 octane E10. The engine almost died 5 miles before I got to the gas station when I finished making a left hand turn and maybe got too wild with the go pedal. Trims were going > 128 on E0 so I'm pretty sure the fuel pump was sucking some air for the last 10 miles give or take. It fell on it's face as I Iet off the accelerator pedal pulling into to the gas pump slot.

Edit: Yesterday afternoon I gradually increased the injector flow constant from 41.494 lbs/HR to 42.966 lbs/HR which explains the trims shift mentioned below.

Immediately afterwards trims were largely > 128, but more importantly the bucking from cam surge immediately went from about 8 to 4 on a 1-10 scale. I'm planning to commute in it tomorrow and will report back.

I'm not trying to refute your knowledge steveo. Exactly the opposite is true. But if we get to a point where our egos tell us we have nothing left to learn on a given subject, difficulties like this will arise. I'm just trying to take the immense effort you've put into eehack to the next level. If you don't appreciate my efforts, I have nothing more to offer you than "I'm sorry you feel that way".

Cam surge calmed with shit fuel is interesting. Could be the e0 also not pure e0 but some mix of additives with little pure gasoline.

I found 2 things that calms cam surge. Running pig rich or fine tune individual cyl trims with proper spark advance.

I almost have a running car again. Things that left are oil change and some leaking o-rings and will be doing some tunning also. I found that evap valve can stuck open and cause all kind of weird situation. After fixing that the trims were much more stable and constant.

The next level of tuning guide will be power tunning and spark advance tweaks. More of an advance science experiments.
Interesting to see how that IR measurements will give some good data.

How about adding some termocouples on each tube and measure precisely all 8 of them with some modified controller code. Can that atmega take 8 channels of data.
Getting EGT in the equation of tunning will bring us closer to lean cruise patch and the benefit of it.

I am really interested to see trims with stock ignition vs 8 coils.
How about perfectly stock healthy engine running e10 with stock tune for 94` gas quality. Will the trims goes higher since you will be adding fuel to compensate for lower power output.

Since it is really hard to find anything other than e10 can we benefit of it and take advantage in the tune like adding more spark advance.

Quote:

---

Originally Posted by kur4o

Cam surge calmed with shit fuel is interesting. Could be the e0 also not pure e0 but some mix of additives with little pure gasoline.

---

I'm about toast for today, but will revisit this tomorrow when the ice+bourbon cocktails have worn off. But what I feel compelled to point out is that with X amount of integrator shift, there will come a proportionately larger combustion change with fuel that burns at a higher AFR like E0. With E10 a larger amount of fuel must be added / removed to achieve a similar amount of change in combustion increase / decrease. This ties directly to the O2 prop table change I mentioned.

Quote:

---

Originally Posted by kur4o

Since it is really hard to find anything other than e10 can we benefit of it and take advantage in the tune like adding more spark advance.

---

All the sources I've found say that E10 doesn't offer enough evaporative cooling to allow pushing higher with spark advance. And I'm not certain ethanol blends are that widely available outside of North America. Back in the energy crunch of the 80's we started diluting gas with corn ethanol. Then in the late 90's / early 2000's when the tree-huggers decided the world was going to end if we didn't come up with a lower carbon footprint fuel source, the government started subsidizing corn production for the purpose of ethanol. Trust me, nothing makes economies shift rapidly like offering to give farmers "free government money".

I just figure that my current tune is POS. I guess tuning with stuck open evap valve is a real bad idea. Never tune an unhealthy engine.

I am sure the cyl trims are way off too. The ve tables are really lean on idle, 160+blms. Interesting now the closed loop idle with vette settings is really good and stable, You can`t tell a difference in running closed or open loop. They were good before but never stable enough.

I am sure EOIT target will also like some tweeking. I am starting all over again with the tune, so some fresh ideas will be posted.

It will be some e9 gas used, so I hope to take some advantage of it.

Quote:

---

Originally Posted by kur4o

The ve tables are really lean on idle, 160+blms. Interesting now the closed loop idle with vette settings is really good and stable, You can`t tell a difference in running closed or open loop. They were good before but never stable enough.

---

Wow, there's something really wrong there (160 blms). No amount of tuning is going to solve that unless your cylinder volume constant or VE table was accidentally scaled by -20% instead of -2%.

I've had some success tweaking the prop step size (0x270C) and found increasing the higher airflow cells improved things somewhat. Going from 24 to 28 generated more cross counts because the integrator resembles more of a square wave instead of requiring two or more integrator changes before the O2 voltage swings past the 0x26F1 threshold. And I've observed that too many integrator changes tends to overshoot the threshold - especially when coupled with too much (or too little) integrator delay. This is why I mentioned that we need to invest some thought on how to tune a PID controller because that's what this is.

I haven't experimented extensively with changing the lowest airflow cell for idle. The integrator "waveform" here still resembles a sawtooth.

Quote:

---

Originally Posted by kur4o

It will be some e9 gas used, so I hope to take some advantage of it.

---

Don't look for some miracle by running ethanol blends. My observation about surge going from 7 to 4 was premature. The only real difference is that a 10% ethanol blend will cause your trims to add 2-5% compared to pure dino juice.

I'm starting to think I might have a plug boot with a hole burned through. I drove it late the other night when the humidity was ridiculous and the temps had fallen down near the dew point. Then again the next morning in similar conditions and both times I was feeling an intermittent miss well before it would have switched to closed loop (I wasn't logging but the engine was cold and less than 30 seconds of runtime). I was assuming this was cam surge, but I'm not sure the actual surge is anywhere near as bad as I initially thought it was.

I'll have a play with some water in a spray bottle after dark as soon as time permits. But this week is essentially shot since my daughter decided (unbeknownst to me) to open a not-for-profit border collie rescue in my home two nights ago. Today, some rocket scientist of a veterinary doctor thought it would be a good idea to neuter the dog before we've had a chance to build trust and wash the cow manure out of his fur.

That evap valve was leaking unknown amount of fuel vapors mixed with air. It totally screw all the closed loop and idle tuning.

I did some cyl trim fix today. Somehow my last maf tune is almost perfectly dialled for the idle with fixed evap.

I put the engine in CL and there were some splits about 4-5 points. Rigth side running alitlle richer. I did some lowering on all of right side cylinders and the trims got stuck at 128 at some point. At least now bank to bank fueling is equal, not sure about the in bank trims.
I also confirmed that fuel robbing thing we suspect long time ago. Some lowering of the rigth side made slighly lean the left side. With perfectly dial blms you can monitor 1-2 point blm increase and int movement also and how they react to cyl trims.
I plan to run the cyl cut thing and see how much air each cylinder is drawing.

Map is still jumping around and I suspect some mechanical shit happens, since it is not consistent sometimes at idle it goes to 42-43, sometiomes jump to 46-47 with no apperant reason.

I also played with some of the closed loop controls and some setting make the up-down 02 swing much slower. I didn`t record the setting but also have no idea what we are looking there.

Quote:

---

Originally Posted by kur4o

I put the engine in CL and there were some splits about 4-5 points. Rigth side running alitlle richer.

---

Several weeks ago I dug through all my logging and found an hour long log from before I touched anything on the engine or exhaust (attached). There wasn't a bunch of idling to look at, but using an analysis script I wrote to look at the .csv data, the bigger picture showed that there were max BLM splits of as much as 10 in certain cells, and average BLM splits that are both fairly consistent with my current tune. I wouldn't call my current tune as good as it can get, but I think it's probably 95% there.

Another thing I've noticed about idle and low airflow BLMs is that they seem to vary wildly with load and weather conditions. Having just spent some time rehabbing my A/C system and tuning for it, I've noticed that simply turning the A/C on (increasing load) at idle causes my BLMs to jump several points positive and become much more balanced (remember this is at closed throttle). Looking for this pattern within normal driving I'm seeing the cells with the worst BLM splits are the very low load cells - bordering on decel.

Another thing I've finally been able to tweak is the idle air bypass size. With the 6.4L displacement I'm seeing that 7/32" is about the best. 15/64" may be better but it also may be too big for very low temps / high air density. This helped immensely with getting it to idle decently with the A/C at max load. It's still a little lumpy at idle with full A/C load, but I can live with it. The vibration of the lumpy idle doesn't bother me as much as the gear clatter from running a single mass flywheel.

Quote:

---

Originally Posted by kur4o

With perfectly dial blms you can monitor 1-2 point blm increase and int movement also and how they react to cyl trims.

---

I mentioned that I was trying to see the integrator (when graphed w/ eehack against O2 voltages) switching like a square wave. I only tried this because I noticed this pattern looking at the bone-stock log I mentioned above. I'll try to dig that log back up and post. When I compared to my running tune at the time, the integrator switched at least twice in most cells before crossing the rich/lean threshold.

Quote:

---

Originally Posted by kur4o

I also played with some of the closed loop controls and some setting make the up-down 02 swing much slower. I didn`t record the setting but also have no idea what we are looking there.

---

I'm beginning to think that tuning this integrator behavior at idle may be worth spending some time perfecting. It may be that it simply requires scaling to match injector constant. But I haven't gotten that far.

I'm playing grandpa to a pup until tommorow night but hope to get around to checking for a spark leak this weekend.

I just tested some tweaked cyl trims. Took the base from the last blm stable ones at idle and adjust the off idle based on them. The engine started to run much better.
I think that if the trims are not optimal in the long run of engine running, some cylinders will get heated more some less and the total balance will be screwed. Also at off idle and take off conditions the engine will be heat soaked differently and it will also affect that and the engine will run like crap all the time even though only idle trims are off.

After good warming I checked the blm at CL on idle and they were almost even. Than decided to give some play with EOIT target.
Tried all kind of it from 40 to 70 in hex. |First I thought stock was $50. It turned out it was $60, Than start increment by 1 and at $63 what a difference. I looked at the afr, map and feel for reference, if AFR goes lower the fuel is evaporated much better and the engine have excess fuel leading to richer mixture.

I got lucky from the first try to get the sweet spot on the first try.

The fuel evaporation test with EOIT must be done at fully heat soaked engine and increment it by 1 and wait at least 30-60 seconds to feel the change.
I guess the optimal settings is cam, temp, head, valve dependent, and the the time needed for best fuel evaporation must be found by feel and experiment.

Than I give it a short drive.

Now what a difference from small eoit move $60 to $63 and part dialing the cyl trims. The engine got quieter by a lot[I could hear the stereo again]. It felt some much better[from crap running to almost stock like running].
The interesting part is that you can not see any significant change in the logs. It is all by feel. It started to shift much better also and the overall throttle response increased.

And that is with guess dialed off idle trims. I still haven`t figure out reliable method to get this dialed unless there is a driver with and a tuner in car. Getting some steady conditions and monitor the blm closely for splits.

The blms are great for
dialing bank to bank difference[split blms are almost definitely from bad ind trims.]

I also noticed that in closed loop only couple of cells are used most of time. I am sure this is not optimal and I am looking for some tweaks there.
Anyone got any experience with it.

Now the wideband at closed loop goes from 14.6-15.1 all the time at stoich. At idle that changes to 15.4-16.??. I guess this could be from low airflow or it could be from not optimal CL settings.

Quote:

---

Originally Posted by kur4o

Than start increment by 1 and at $63 what a difference. I looked at the afr, map and feel for reference, if AFR goes lower the fuel is evaporated much better and the engine have excess fuel leading to richer mixture.

---

I've had mine at $69 since last summer. Haven't spent any time with this since installing the 170F thermostat so it might be worth revisiting.

Quote:

---

Originally Posted by kur4o

I also noticed that in closed loop only couple of cells are used most of time. I am sure this is not optimal and I am looking for some tweaks there.

---

This is what I'm currently running:

Code:

---

RPM        MAP
1200        42
1650        56
2100        68

---

But this is with a M6. I decided to put the main RPM split at 1650 because there's really only two cruising gears for me - 5th and 6th. Below 55mph in 6th is pretty sketchy around 1500 rpm so I generally run in 5th under 60mph. With my cam this hits a wider range of cells with typical driving. An A4 would probably behave quite differently with these thresholds.

Quote:

---

Originally Posted by kur4o

Now the wideband at closed loop goes from 14.6-15.1 all the time at stoich. At idle that changes to 15.4-16.

---

Is that leaner range idling in gear, out, or both?

I'm seeing a pretty consistent range from 0.97 to 1.06 λ in all closed loop operation. That's roughly 14.25:1 to 15.58:1 assuming 14.70 stoich. But that's unfiltered data - it's not using eehack's wideband functionality, just D27 adc conversion.

Quote:

---

Originally Posted by kur4o

I just tested some tweaked cyl trims. Took the base from the last blm stable ones at idle and adjust the off idle based on them. The engine started to run much better.
I think that if the trims are not optimal in the long run of engine running, some cylinders will get heated more some less and the total balance will be screwed. Also at off idle and take off conditions the engine will be heat soaked differently and it will also affect that and the engine will run like crap all the time even though only idle trims are off.

After good warming I checked the blm at CL on idle and they were almost even. Than decided to give some play with EOIT target.
Tried all kind of it from 40 to 70 in hex. |First I thought stock was $50. It turned out it was $60, Than start increment by 1 and at $63 what a difference. I looked at the afr, map and feel for reference, if AFR goes lower the fuel is evaporated much better and the engine have excess fuel leading to richer mixture.

I got lucky from the first try to get the sweet spot on the first try.

The fuel evaporation test with EOIT must be done at fully heat soaked engine and increment it by 1 and wait at least 30-60 seconds to feel the change.
I guess the optimal settings is cam, temp, head, valve dependent, and the the time needed for best fuel evaporation must be found by feel and experiment.

Than I give it a short drive.

Now what a difference from small eoit move $60 to $63 and part dialing the cyl trims. The engine got quieter by a lot[I could hear the stereo again]. It felt some much better[from crap running to almost stock like running].
The interesting part is that you can not see any significant change in the logs. It is all by feel. It started to shift much better also and the overall throttle response increased.

And that is with guess dialed off idle trims. I still haven`t figure out reliable method to get this dialed unless there is a driver with and a tuner in car. Getting some steady conditions and monitor the blm closely for splits.

The blms are great for
dialing bank to bank difference[split blms are almost definitely from bad ind trims.]

I also noticed that in closed loop only couple of cells are used most of time. I am sure this is not optimal and I am looking for some tweaks there.
Anyone got any experience with it.

Now the wideband at closed loop goes from 14.6-15.1 all the time at stoich. At idle that changes to 15.4-16.??. I guess this could be from low airflow or it could be from not optimal CL settings.

---

Hi guys.
Kur40: My test data the other year showed also 163 (#63) to be about the best for fuel atomisation.
I went back to some stock values with the newpipes & TB but will try the #63 again.

Some thoughts, I need to check previous data logs.
Question: If I technically have bigger injectors than required the time for injection would be reduced? Ie double the injector size would need half the time to inject. Would this affect the atomisation?
Also ( I'm still not 100%in my head which way the target is moving... #60 is 180 degrees , #63 is 163 degrees ) so the EOIT is actually earlier giving more time to evaporate. This would move the window for injection back slightly... IF the injectors we not big enough you might not have enough window spare.
My 42'# had a Duty cylce of around 70%. At the track last week I actually managed to turn 7Kover the 6400 previously, so need to recheck what DC I was acheiving.

I relogged recently using cylinder cut to measure changes in individual cyl trims.
OTTOMH, 0% TPS was doable as the results were within the Wideband range. Results averaged & calced accordingly.
>0% I think I ran about 2.5% TPS which is where my surge seems & logged again. On several cylinders the WIDEBANDS maxed out so the data was inconclusive other than some cylinders needed more fuel.
After calculating and setting the trims ( I didn't like the size of difference between cylinders like up 7%) the car seemed to run pretty well, until the BLM's relearned and the surge came back. I've since cut the trims back.

Scott: I see you mentioned BLM differences of upto 10 points, I've got down to around 6ish ie from -6 to +6 which seems pretty good.
I'm currently idling about 124 128 but with 2 widebands running I notice 1 side reads lower than the other ie in EEHaack graph, 1 O2mV trace is slightly higher than the other (Voltage Offset maybe or Difference in O2 Output ??)

I also note with the 2 WB, It's rich on the left down low but WOT up top they finish up the other way round, lean on the left ( upto 1.0 AFR )

Kur4o: I'm interested how you calculate the individual trims.


Mitch

Quote:

---

Originally Posted by Terminal_Crazy

Also ( I'm still not 100%in my head which way the target is moving... #60 is 180 degrees , #63 is 163 degrees ) so the EOIT is actually earlier giving more time to evaporate.

---

Correct, I think.

$60 = 180 ATDC = 540 BTDC (before cylinder fires)

$63 = 163 ATDC = 557 BTDC - more time for evaporation

Quote:

---

Originally Posted by Terminal_Crazy

This would move the window for injection back slightly... IF the injectors we not big enough you might not have enough window spare.
My 42'# had a Duty cylce of around 70%. At the track last week I actually managed to turn 7Kover the 6400 previously, so need to recheck what DC I was acheiving.

---

I calculated it way back when were originally discussing this. I don't remember the exact number, and it changes with EOIT, but at injector duty cycles larger than roughly 40% you're going to be injecting fuel during the intake stroke. And at large duty cycles you'll be injecting fuel through the entire intake stroke. So evaporation time isn't going to make much difference at WOT / large throttle openings. For all intents and purposes, at large injector duty cycles it's one big fog of fuel.

Quote:

---

Originally Posted by Terminal_Crazy

Scott: I see you mentioned BLM differences of upto 10 points, I've got down to around 6ish ie from -6 to +6 which seems pretty good.

---

I'm talking about PEAK BLM SPLITS, not average which is what eehack's analyzer gives. My averaged splits are generally never more than 4 or 5, and the lean condition moves from right to left as the throttle opens.

The point I was trying to make is that BLM splits change within the same BLM cell based on engine load, baro, humidity, air temp. This is most apparent in cells 0, 1 and 2, but also can be witnessed in 4, 5 and 6. In my opinion, achieving perfect balance by looking at BLMs is an impossible task. Even the factory tune with a completely unmodified engine showed similar behavior with largest average splits around 4-5 and peak splits as high as 8-10. Looking at the logs I uploaded with eehack may make it more self-explanatory.

Bigger injectors spitting fuel fast will need more time to evaporate even though the end of the injection is at the same place the start of the injection will move and there will be less time for evaporation. Increasing fuel pressure will help with atomization alot.

So the injector size must be included in the EOIT variables.

Low res signal can be seen as ms for each 90* rotation. Multiplying it by 8 will give total time of 2 crank revolutions. Compare that by the injector opening time, and you will get the duty cycle of them. If you want to make sure the injectors are not spraying in the intake stroke, multiply the low res ms by 6 to exclude the intake stroke from the duration. Adding the eoit target to this you can still overlap events unless eoit matches the start of intake stroke.

To get some initial cyl trims I used some variables like exhaust shape and length of individual tubes. Intake passage holes modified or stock uneven size.
And make some chart of the passages and how the air is drawn for the firing order of the engine. Made some initial matrix but have no idea how close I am now with the initial guessing. I can`t find them anywhere.

For some reason right side always needs less fuel. It could be because the stock exhaust merges near right side and the left side exhaust is longer that the right side making the more efficient. Higher length = better exhaust scavenging = higher vacuum in the cylinder and better filling with air.
I also fine tune later for map and afr change with cyl cutting. You need to take into account and fuel robbing by other cylinders. Adding to one cyl can make others run richer.

Last technique I used was to monitor blms at idle and making them 128 even on both sides. I am sure in the end you can dial it too so there is no split.
7% difference shouldn`t bother you. You need to inject fuel matching the air that enters in the cylinder. I think the biggest split I have now is from 136 to 119.

I closely monitor the behavior of the INT. On the right side it is really some subdued wave line. On the left side it is at zig-zag pattern, that matches the wideband fluctuations. This gives me some thoughts that left side is not in bank balanced. I will do some more experiments with adding and cutting inbank.

I did the map test and the results are #1 the strongest, and 5 and 6 the weakest. I hope to use that data and see how it relates with the trims.

Quote:

---

The point I was trying to make is that BLM splits change within the same BLM cell based on engine load, baro, humidity, air temp. This is most apparent in cells 0, 1 and 2, but also can be witnessed in 4, 5 and 6. In my opinion, achieving perfect balance by looking at BLMs is an impossible task. Even the factory tune with a completely unmodified engine showed similar behavior with largest average splits around 4-5 and peak splits as high as 8-10. Looking at the logs I uploaded with eehack may make it more self-explanatory.

---

I also noticed per cell splits, evens and opposite splits. That is the reason I want to reconfigure the cell arrangements. Some explanation how to modify for closer denser cells between 1000-1500 rpm where is most of the time the engine runs.

I think that in cell splits can be due to different cyl trims needed at off idle, based on the airflow, map and rpm.

I am sure there needs to be at least 3 different off idle trim ranges. That can be easily patched once I figure how to tune them and nightmare to add real time control on all of them.

I am starting to think how to add a button in the eehack controller window that will save a screenshot with date, time stamp of it. Now some ideal settings can be lost really fast. I will be really glad if some of you guys can encode something like that.


I got the leaner wideband in park, no load at the engine. With perfect blms at 128, I suspect the wideband not reading right at low airflow.

Quote:

---

Originally Posted by kur4o

I also noticed per cell splits, evens and opposite splits. That is the reason I want to reconfigure the cell arrangements. Some explanation how to modify for closer denser cells between 1000-1500 rpm where is most of the time the engine runs.

---

It seems fairly self-explanatory.

Attachment 15926

The rpm thresholds are in table 0x273f and the map thresholds are in table 0x2742.

I think for giggles I'm going to try stacking my map thresholds a little closer together and see if anything changes.

Quote:

---

Originally Posted by kur4o

I think that in cell splits can be due to different cyl trims needed at off idle, based on the airflow, map and rpm.

I am sure there needs to be at least 3 different off idle trim ranges. That can be easily patched once I figure how to tune them and nightmare to add real time control on all of them.

---

This is why I mentioned that adding load at idle / 0% TPS causes my BLMs to add fuel and become almost perfectly balanced. I'd hate to see you go to the trouble of patching in another individual trims table based off of TPS when it might work better to have different tables based on load or airflow. For example, without the load of A/C my idle is around 45kpa and 9g/s. With A/C it's more like 48kpa and 11.5g/s.

Edit: However, I'm not optimistic more trimming will "fix" the issue. It seems like these imbalances are most prominent at very low manifold pressures. I'm beginning to wonder if there could be something to your theory that O2s are less accurate at low air velocity.

Quote:

---

Originally Posted by kur4o

I am starting to think how to add a button in the eehack controller window that will save a screenshot with date, time stamp of it.

---

There's probably a 3rd party utility you can install on windows that will do that for you without messing with coding it into eehack. On most linux distros the [printscr] key mapping is already configured to do this.

I finally had a chance to check for spark leak last night and found absolutely nothing. I think I'm going to beef up my MAP vs RPM multiplier in the ignition controller and see if a little fatter spark makes any difference with what I initially thought was surge.

I did some runs with fixed EOIT target and guess what the idle cyl trims are off again. A slight adjust of the off idle trim gave positive results and some of the nasty noise on high load low rpm take off is almost gone now.

Now the car have much more crisp throttle, and a increase in power can be felt at very low rpm. Much more tamed and quieter also.

I did change the curve of the eoit target during warm up. Now it leans much earlier and for longer. I guess the curve must be dialed pretty good for a good warm up.
The main transition is around 50*C where you get some lean spark. I am thinking keeping the straight in cyl spray longer might be needed longer, since the head is colder and can`t evaporate fast enough the fuel.


Right now I am having 650rpm at idle with 7.5agfs and map varies 42-47 for no reason, sometimes it is 42-43, sometimes 45-46.
Now all surging is gone and the blms are almost perfectly balanced around with stock maf table.

I think INt can give some very good tuning data about the ind trims.
I need some completely stock engine log in good health to see how the blm and int act there.

After close looking at some of the logs, the Int acts like that, goes down 1-2 points, wait for 02 feedback, if the mv are still high and not moving, goes down 1-2 more points and so on until the 02 are lean enough around 50-100mv than the action is reversed. With some timer it adds to blm if the response is too slow. Looking at code doesn`t give a clue how that pid system works, so guessing the basic algorithms of the control and than finding them in code can be better. Playing with the real time control can show what goes where in the inputs.

I am not sure eehack can be draw realtime the int. Anyone tried the graph at realtime logging.

Quote:

---

Originally Posted by kur4o

After close looking at some of the logs, the Int acts like that, goes down 1-2 points, wait for 02 feedback, if the mv are still high and not moving, goes down 1-2 more points and so on until the 02 are lean enough around 50-100mv than the action is reversed.

---

As I understand it, the amount that's added to / subtracted from the integrator is derived primarily from the 0x270c Base O2 Proportional Step Size table. I've increased the higher airflow cells by 4 points over stock, and the result was more cross counts for the same amount of driving vs stock. Also less number of integrator changes per swing / cross cycle.

Edit: It's worth nothing that the stock Y-body values for this table are 24, and the stock F-body are 32. Having 42lb injectors, I expected to have to reduce this value.

The amount of time it waits is in the 0x26fb Base O2 Integrator Delay table. I ended up with the table below after graphing the integrator and finding example frames at different airflows, then adjusting the delay to match the amount of time between the integrator change and the moment the O2 starts to swing. If your Prop Step isn't big enough there may not be enough fuel being added and so the integrator will need to to be adjusted again before a cross happens.

Code:

---

AF    Stock    Mine
00    0.40    0.53
16    0.18    0.35
32    0.13    0.25
48    0.11    0.20
64    0.10    0.15

---

This is for equal mid-length headers.

The threshold, whether going rich or lean is the 0x26f1 Fast O2 Rich/Lean Threshold table. I haven't touched this. My assumption is if after the integrator delay, if the O2 voltage has crossed this threshold the integrator is then moved in the opposite direction.

The idea is to get the O2 voltages to oscillate around the 0x26f1 threshold without excess over or under-shoot. None of this is really new information. I'm sure there are a bunch of other things that can be tweaked such as the slow O2 filter coefficient and the integrator and proportional step gain tables.

Ultimately my goal is to get to the best running tune possible in all weather conditions, and for whatever fuel blend being burned. Unfortunately instead of being out logging and testing, I'm sitting here letting the "grandkids" (two exhausted border collies) get their afternoon nap in.

There is one more component in the CL control that is not logged for some reason. The CORRCL. It is changing too fast to log I guess. I only know that it makes immediate changes to the mixture for really fast corrections.

Maybe after I gather some more thoughts about how the CL control will look again in the dissasembly to figure the missing pieces.

For me CL settings off idle never give any troubles. It was at idle that was terrible but after copying all the vette parameters[ THANKS for making that discovery] now CL idle is great, especially after fixing the evap valve.

I think that cyl trims can greatly affect INT behavior at low airflow. If tuned and balanced properly the INT is like subdued wave. Going up down slowly. Untuned it is like zig-zag pattern with some peaks. Also the oscillation between stoich on the wideband can give an idea how good the CL settings are.

To get really good understanding you need perfectly tuned engine so it doesn`t mix or cover some of the output.

I suggest you keep stock settings and very slowly work any changes from there. The best will be to make the wave resemble as close to stock engine wave as possible.

Closed loop control at idle on lt1 is not perfect, that is the reason gm upgrade it on later pcms[ls1 code] with some special settings for idle.
ON 96-97 code there is also some tables for swing voltages.

So what will be the general consensus. At idle subdued wave, off idle some zig-zag pattern with wide or short peak?

Really interested to see some stock logs to compare.

Quote:

---

Originally Posted by kur4o

So what will be the general consensus. At idle subdued wave, off idle some zig-zag pattern with wide or short peak?

---

I think (solely my personal opinion) for all ranges other than idle, a square wave without massive overshoot of the O2 voltage is preferable. There is some scientific theory behind my opinion, but for now let's call it "redneck premonition".

Quote:

---

Originally Posted by kur4o

Really interested to see some stock logs to compare.

---

The log I uploaded back on post # 14 is the best I have [link]. If I get time tomorrow I'll dig through and see if there are any others. But this was on an engine with 160,000 miles on it. So it would be good to have some other examples to look at.

I played with my BLM thresholds this week and ended up here:

Code:

---

RPM    MAP
1400    38
1650    50
1900    62

---

I've only driven this on interstate (primarily in 6th gear @ 75mph) so it's a little soon to declare it "good". But I'm seeing a much wider range of cells being touched with typical driving. Normally my "boring" commute on interstate results in cells 2 and 6 seeing the majority of action. I'd be interested to see how these thresholds work in an A4 car.

I'm hoping to make time to work on EOIT and individual cylinder trims this weekend. I'm out of ideas on fixing the surge / bucking I'm getting between 2-6% TPS and < 1400 rpm.

After a couple hundred miles logged I'm liking the BLM thresholds / boundaries above. Results in a much wider number of cells being touched.

Over the weekend I messed around with individual trims and EOIT. After quite a bit of fiddling I was unable to see or feel a difference between 0x63 and 0x69 so moved it to 0x63 as that seems to be the consensus.

Individual trims seem to be a tough nut to crack. Based on header tube temps I found that #2 was quite a bit hotter at closed throttle than the other holes. So I started by adding to it (was 134, now 142). I noticed immediately the clutch is much less "grabby" when feathering and the idle is much more stable with the A/C on. I still haven't found a magic set of trims to eliminate what I was diagnosing as surge at cruise. I did however manage to move it to another load / speed range which helped me find an instance of it in a log and now think it's actually lean misfire caused by integrator overshoot (actually undershoot, if that's a word).

Anyway, one thing I noticed is that adding fuel to #2 didn't seem to change the tube temp. So I conducted a little experiment in the parking lot just now. I'd been looking solely at max temp on my IR gun and wondered what the min/max/avg would look like if I left the engine running in closed loop so the AFR was oscillating and measured for roughly 10 seconds each. After looking at the data in a spreadsheet I find #2 has the smallest deviation - 2*C vs #8 with the highest at 35*C. I don't know if this means anything, but I'm going to experiment further with it this evening. My hope is that managing the temp deviation is the key to getting the trims just right.

After numerous hours trying to improve on individual trims I'm throwing in the towel and going back to what I was running from two weeks ago.

I'm not sure header temps can tell us anything. Repeated min/max testing showed no patterns that could be repeated, either in closed or open loop. Once the hood is opened everything seems to change and it takes 10 minutes or more before temperatures moderate. And then if you're like me the cooling fans switch to high, which causes further inconsistency.

The only hang-ups I have with the current tune (assuming it runs the same as it did two weeks ago) are:

1) low speed bucking at closed throttle or very low throttle opening, open or closed loop
2) a missing / surging sensation at cruising speeds (closed loop) with low throttle openings - mostly prevalent with the a/c on

After digging through logs from that last few days I think my second issue is being exacerbated by the IAC ramping down when the A/C clutch disengages, and that happens to coincide with instances where both integrators are swinging negative. The IAC closing causes a temporary rich condition, but the integrators continue to go negative trying to drive the O2s below the threshold, resulting in a severe lean condition that is felt as either a miss or detonation. Whatever it is, I think I can live with it as long as it otherwise runs excellent.

I'm not sure I'll be able to do anything with the low speed bucking. I suspect it's either the closed-loop idle spark fighting the car's inertia, or the IAC similarly oscillating. Whatever the cause, it hates the idea of allowing me to get to work without spilling my effin coffee all over the console / myself. Not to mention the stress it must be putting on the driveline.

Quote:

---

Originally Posted by spfautsch

After numerous hours trying to improve on individual trims I'm throwing in the towel and going back to what I was running from two weeks ago.

I'm not sure header temps can tell us anything. Repeated min/max testing showed no patterns that could be repeated, either in closed or open loop. Once the hood is opened everything seems to change and it takes 10 minutes or more before temperatures moderate. And then if you're like me the cooling fans switch to high, which causes further inconsistency.

---

That's why I asked Kur4o how he tuned his. I had coated headers which didn't help. The new long tubes have 4 layers so No way is it going to read any where near.

Fuel cut to each cylinder did seem to give fair results but I was unhappy with the weird mismatch of fueling. I'd prefer to have them all getting a similar shot at idle.

Quote:

---

The only hang-ups I have with the current tune (assuming it runs the same as it did two weeks ago) are:

1) low speed bucking at closed throttle or very low throttle opening, open or closed loop
2) a missing / surging sensation at cruising speeds (closed loop) with low throttle openings - mostly prevalent with the a/c on

After digging through logs from that last few days I think my second issue is being exacerbated by the IAC ramping down when the A/C clutch disengages, and that happens to coincide with instances where both integrators are swinging negative. The IAC closing causes a temporary rich condition, but the integrators continue to go negative trying to drive the O2s below the threshold, resulting in a severe lean condition that is felt as either a miss or detonation. Whatever it is, I think I can live with it as long as it otherwise runs excellent.

---

I've invested in MLVHD so you can track a lot more data at a time.
IAC didn't really move that much.
I think It seemed to tracked the O2's mV rather than anything else, not sure about the Integrator, will have to re look.

Quote:

---

I'm not sure I'll be able to do anything with the low speed bucking. I suspect it's either the closed-loop idle spark fighting the car's inertia, or the IAC similarly oscillating. Whatever the cause, it hates the idea of allowing me to get to work without spilling my effin coffee all over the console / myself. Not to mention the stress it must be putting on the driveline.

---

I've hacked lots of air tables without much luck. One day it's great, then it's back on the next flash.

I'm currently down atm
Took the car to the track a couple of weeks ago.
Sticky tyres ate the clutch, it was smelling after the first pull
Best was 12.6 and 112mphand there's a bit left so an 11.0 isn't far off

Quote:

---

Originally Posted by Terminal_Crazy

That's why I asked Kur4o how he tuned his. I had coated headers which didn't help. The new long tubes have 4 layers so No way is it going to read any where near.

---

Mine are unfinished 304L stainless, so the reflectivity could be an issue for the IR gun. Long story short without 8 separate O2s or EGT sensor bungs welded in, I think the methodology of measuring tube temperatures is at least two orders of magnitude less efficient than the enterprise of trying to herd cats.

Quote:

---

Originally Posted by Terminal_Crazy

Fuel cut to each cylinder did seem to give fair results but I was unhappy with the weird mismatch of fueling. I'd prefer to have them all getting a similar shot at idle.

---

I suspect the mismatch you're seeing (you're not alone btw) is due to baro, humidity and trims. Read on to see my comment about trims.

Quote:

---

Originally Posted by Terminal_Crazy

I've invested in MLVHD so you can track a lot more data at a time.

---

I have no idea what MLVHD is. But the word "invested" scares me to a degree that I'm not sure I want to.

Quote:

---

Originally Posted by Terminal_Crazy

IAC didn't really move that much.
I think It seemed to tracked the O2's mV rather than anything else, not sure about the Integrator, will have to re look.

---

IAC should only track idle speed, roughly, at closed throttle with the car stopped and at running temperature. Outside of those conditions IAC can be anywhere depending on how fast you were just going when you (for example) downshifted from 4th to 2nd. O2 readings should be very, very, very distantly related to any of those parameters. Certainly quite disconnected from IAC values (IAC is an output, O2 is an input btw).

Quote:

---

Originally Posted by Terminal_Crazy

I've hacked lots of air tables without much luck. One day it's great, then it's back on the next flash.

---

Are you resetting BLMs after every flash, or do you have your BLM keepalive value set to 128? Whatever the case, long term trims are applied all the time except possibly to prime pulse fueling. So when you park the car and idle BLMs are 140, they're going to be 140 the next morning when you spank it over for the first time. And if you happen to reflash the night before and reset trims to 128 you'll be experiencing a 12 point (BLM) lean shift based on current weather conditions. This is what I was talking about earlier with trims. Long term trims are meant to last a while, hence the word "LONG". This is not to be taken lightly. Weather, as we assuredly all know, is a bitch that changes as her moods dictate.

With my current setup targeting a slightly lean condition with E10 fuel, a large ambient temperature shift can produce a very sketchy, slobbery (pig rich, or billy goat lean) start because BLMs are at whatever they were from the evening before. This is why I've been making it clear I'm driving the car like a daily driver / grocery getter. I want to be able to turn my wife loose with it and not get a call because the S>O>B won't start.

Quote:

---

Originally Posted by Terminal_Crazy

I'm currently down atm
Took the car to the track a couple of weeks ago.
Sticky tyres ate the clutch, it was smelling after the first pull

---

You're preaching to the converted with burned clutch smell woes. Ultimately I'd like to go to a dual disc setup just to get a smooth engagement again. I can smell the clutch burning going down a 3% grade in 3rd gear, below 70% TPS. It clearly can't take what the engine has to offer.

Quote:

---

Originally Posted by Terminal_Crazy

Best was 12.6 and 112mphand there's a bit left so an 11.0 isn't far off

---

There are no tracks near me (at least ones that don't do double-duty as public roads) so I can't tell you what my ETs are.

Long story short, if this is as good as it gets I'm happy with it. If I can make closed loop run well enough that I'm not melting my cats down on a regular basis, I'll be thrilled.

My next goal is to find a chassis dyno and put some numbers to "what it is". I've seen at least two different intraweb discussions from folks running the same cam I have and similar porting that it should make 450-500 at the wheels "no problem" that causes me to struggle with a reason to worry about the minutia of how it feels when cruising with the A/C on. But everybody needs a hobby, and this is mine.

I just wish the auto guys understood the feeling of 900 rpm parking lot buck. They just don't know how good they've got it.