O2 sensor tuning with Headers

[JBZ]

Barum, trying to wrap my brain around this and having a hard time at how the multiple different values correlate with rpm vs map and what im having a hard time with is that in your above screenshots each cell has a diffierent value, but with hpt we would only have one millisecond value for the given airflow mode. So i would imagine on the lower portion of that certain cell our integrator delay may be too fast (lower airflow) and before it changes cells at the higher airflow we may Have to long ot high of an integrator miliseconds...if i said the right??? Lol

Man it would be great if we had a way to calculate exhaust flow to calculate our integrator delay. Im sure we can and looks like you have used maf.

How does fueling and injectors play into this? Looks like all these screenshots relate to integrator delay and just wondering what you found regarding porportional and error cells?

[Michael_D]

I did a couple more tests today.

Proportional multiplied by 1.8, Delay x 1.8, Error x .5. Basically just cut the error table in half from my last test with it stock. Didn't feel or notice any improvement from the last O2 settings.

7-23-16- O2 CHANGE-p x 1.8 err x .5.hpl

I then put the proportional back to when the engine felt best, and used the same error / halved settings. Engine was smoother, but still did not feel as good as it did when the error table was stock.

Proportional x 1.8, Delay x 1.8, error x .5

7-23-16- O2 CHANGE-p x.8 err x .5.hpl

I think I'm done until I have a better understanding of exactly how these tables work, especially the error table.

[topher455]

I dropped these in last night and what Ghuggins mentioned for the purge ON/OFF and LTFT purge values. seem pretty good to me so far, but every time I say that it goes to shit on my next drive.


this is how I figured out to see what table it was pulling from and how I determined that airflow mode break down I post a while back where I drew the lines in there based on MAF LB/HR. you can see with the swing speeds exactly when it changes for the most part.

Yes! That illustration helped a ton!! Once I understood the relationship between all the tables it was much easier to begin to modify them!

[barum]

I suggest this for bed time reading.

Especially these are important:



Note how the roll transport delay is very similar to O2 sensor delay.



When defining optimal parameters (by this method) for the PI regulator they are calculated as follows:



What this means in practice, is that when ever the transfer delay (L) increases
- the time constant (Tp) is 5.7-fold compared to transfer delay => controller will be able to make slower control actions than before
- the suitable integration time (Ti) is three times the transfer delay
- the suitable total gain decreases directly proportionally to the increase of the transfer delay

So, by increasing the integrator delay (= transfer delay) AND by increasing the proportional gain (as we can't access any of the integrator parameters), you are actually going to wrong direction by increasing the total gain of the controller leading to more unstable control = more overshoot, more uncontrolled oscillation.

[barum]

JBZ, I used VE table as template since the original study containing the transfer delay measurement used that format.

The procedure to calculate what integrator delay the PCM used in each cell is as follows:

1) calculate the corresponding mass air flow for each RPM/MAP pair based on the VE table, RPM, # of cylinders, one cylinder displacement and density of air at 25 degC:




2) calculate the integrator delay for each cell in the MAF vs Airflow mode table:




3) calculate the integrator delay for each cell in the MAF vs RPM vs MAP table:




Now the real world measurements and PCM selected integrator delay values are in the same format allowing them to be compared to check where the integrator delay is adequate and where it is not.

The engine used in the test was a Gen3 6.0 liter unit and I wanted to see "what the PCM was thinking" compared to the measured transfer delays, nothing more for the time being. It's not perfect but it's not bad either.

To do this effectively in both formats (MAF/Airflow mode indexed and RPM/MAP indexed) would require some VBA code to be written, it's too labor intensive to do it manually.

However, the main task would be to predict the transfer delay as well as possible. By comparing the GM values, real world measurements and calculation model, we might be able to get a better guestimate for the transfer delay required by the LT headers than what is simply based on the distance delta to OEM O2 location.

[topher455]

So, by increasing the integrator delay (= transfer delay) AND by increasing the proportional gain (as we can't access any of the integrator parameters), you are actually going to wrong direction by increasing the total gain of the controller leading to more unstable control = more overshoot, more uncontrolled oscillation.

Im having to break this down into chunks because it is a lot to digest. I have noticed great results by decreasing the integrator delay table which corresponds with what you are saying. I find it odd that increasing the proportional would be moving in the wrong direction though. Increasing the value in the table increases the proportional amount of fuel being added or removed to drive the engine across stoich. But possibly with the o2 sensor being further away from the head more proportional change is needed?

Now that I have the integrator table modified and a proper swing pattern throughout the rpm range Im starting to look at other aspects of this trying to figure out how much we can really get from this. My wideband swings from 14.3-15.3 afr at idle and lower rpm. Im not sure if a 1 point afr is an acceptable amount for oscillation or if I can tighten this up more. What should be the first steps to tightening this up?

<I figured adjust my min/max rich lean and b1,b2 r/l tables all to 450mv so the engine is always targeting exact center stoich? I have been slowly tightening those values but haven't seen any difference.

<I guess start lowering the proportional fuel a little at a time to see if it tightens the afr swing during oscillation?

<Or would this be where I need to start modifying the error table?

[schpenxel]

You guys are making my head hurt

[JBZ]

Barum again thank you, that is some very valuable info and some of the best info posted yet, between you and ghuggins you have both brought some very good info for us to try and use to back calculate each and every one of our own situations. Although we are all struggling with the same problem, our engine combos are all different so thus making everyones unique and this i hope is the data that we can pick apart to get this figured out.

To barum and topher regarding proportional corrections, i may be wrong here but if we had left stock injectors in the car wouldnt proportional fueling say the same? No less or more fuel would be needed to drive oscillation? But what im thinking barum is saying (and this is a guess...lol) is that we need to look at decreasing the amount of proportional fueling due to larger injectors? Im pretty sure we are all going larger injectors so we all should be decreasing this right?

[schpenxel]

I'm working on a twin turbo Camaro next too, so that's one more combo we can mess with. The O2's are after the turbos on this setup, so quite a bit further downstream than stock. I need to spend a few hours reading this thread, a lot of good info, but I haven't had time to digest it just yet.

[JBZ]

I'm working on a twin turbo Camaro next too, so that's one more combo we can mess with. The O2's are after the turbos on this setup, so quite a bit further downstream than stock

There was another thread where ghuggins had some values for a post turbo set up, ill see if i can dig it up and post up the link. Or maybe hr can chime in.

[schpenxel]

He sent me a few tunes, so one of those may have some good info for those tables.. if not, a link would be awesome.

I'm trying to figure out another issue right now on this car.. DFCO is getting stuck "on" for some reason... Inj PW stays at 0 all the way to 0 RPM's. It's really weird.. fuel system status stays OL - Decel/Accel, inj PW = 0, and all O2 sensors are maxed out lean.. You can literally push in the clutch at 3000 RPM's and it just falls from 3K to 0, inj PW 0.000msec the whole way. It starts and idles hot or cold just fine.. push in clutch while driving and you're screwed.

I'm just disabling it for now but owner wants the gurgling sound that it creates during decel, so trying to figure out why it's doing that. That's like last in the list of priorities.. would like to get it figured out at some point though

Anyways... back to GM patents and theory

[GHuggins]

VE-table of the engine used in test I linked earlier:



Predicted transfer delay according to Kaidantzis et al. [1993] is calculated as follows:

For those of us who are idiots - what do the letters in the equation stand for?

I know some of them - or at least think I do - such as Transfer Delay, Volume for intake and exhaust flow - Would have to figure out a way to calculate the volume for exhaust, otherwise just not sure on the rest of the letters?

Thinking this could be made much simpler, if I just knew what everything stood for

[barum]

JBZ, my thinking is that since the transfer delay with LT headers is longer (without any doubt), we shouldn't be increasing the total gain of the regulator. I could try to estimate the speed at which the exhaust gas travels at different MAF readings which would enable us to estimate that part of the transfer delay. Once this is done, figuring out the corresponding change for the LT headers shouldn't be too difficult.

The proportional base doesn't (probably) behave like the proportional part of the normal PI regulator as its main purpose is to "go over the switching point". Our tests have shown that proportional base is responsible for the big step-wise change in the pulse width which seems to confirm that assumption.
O2 error table, on the other hand, might be much closer the proportional gain part.

Anyway, it would be really good if someone (like the original poster of this print screen ) with access to "3rd tuning software" could show us the contents of these CL tables (starting from Fast O2 filter coeff... all the way to the CL proportional term gain vs. airflow mode - rich):



Preferably from a 2010 C6Z E38 so that I could try search and locate those directly from the binary.

But what im thinking barum is saying (and this is a guess...lol) is that we need to look at decreasing the amount of proportional fueling due to larger injectors? Im pretty sure we are all going larger injectors so we all should be decreasing this right?

I hadn't thought of that yet, but I'm guessing that all these CL calculations are eventually multiplied to the (feedforward based) pulse width which already takes into account the higher flow rate of the bigger injectors => injector size shouldn't matter.

[topher455]

He sent me a few tunes, so one of those may have some good info for those tables.. if not, a link would be awesome.

I'm trying to figure out another issue right now on this car.. DFCO is getting stuck "on" for some reason... Inj PW stays at 0 all the way to 0 RPM's. It's really weird.. fuel system status stays OL - Decel/Accel, inj PW = 0, and all O2 sensors are maxed out lean.. You can literally push in the clutch at 3000 RPM's and it just falls from 3K to 0, inj PW 0.000msec the whole way. It starts and idles hot or cold just fine.. push in clutch while driving and you're screwed.

I'm just disabling it for now but owner wants the gurgling sound that it creates during decel, so trying to figure out why it's doing that. That's like last in the list of priorities.. would like to get it figured out at some point though

Anyways... back to GM patents and theory

Stupid question.. It's not the cfco settings is it? Next to the dfco?

[schpenxel]

Stupid question.. It's not the cfco settings is it? Next to the dfco?

Nah I had CFCO disabled but I'll check to make sure it's not that.

[Stuffs]

nice thread
can can i implement this info on a 2011 camaro ss 113 lsa cammed with long tubes?

[Da-Grump]

For those of us who are idiots - what do the letters in the equation stand for?

Page 2, lower left:

http://homepage.usask.ca/~tkw954/content/MISC-0167.pdf

[GHuggins]

Page 2, lower left:

http://homepage.usask.ca/~tkw954/content/MISC-0167.pdf

Hmmm... Should have read it more closely or at least when it wasn't so late at night

Next stupid question... How is the ecm actually calculating this or rather how did the OE come up with the delay times - was it how it was done in testing (from the moment the injector fired) - this just seems to put way too much into the mix since injection soi is all over the place depending on engine load and size of injector - OR is it simply based off of when the cylinder was supposed to fire (I would assume it would make more since to base the "root" numbers of the time frame off of this instead of the injectors, since a power stroke is more or less set in stone with engine rotation)

Even then there are a lot of other things coming into the equation, such as when the exhaust valve is opening - this would inevitably change when the exhaust pulse hits the O2... If it's off of cylinder firing, it would be extremely easy to dial in an O2's delay function... All one would need to do is set the airflow mode table to one mode (a mode that corresponds directly with the proportional and integrator tables) / run the engine at that airflow / kill a cylinder (preferably do one at the front and then the rear most of each bank - not sure on this - I know the OE bases most off of cylinder one, so wondering if it wouldn't be best just to use the front cylinders?) / average those out in the measured ms from time of noted miss to the time the O2 saw a lean transition... Wash, rinse, repeat for corresponding modes and airflow levels...

It's just the way it was done in the testing, didn't seem to yield numbers correctly correlating with what's in the tune - they follow the correct curve - just give different numbers - unless I was looking at the wrong thing posted above?

[JBZ]

nice thread
can can i implement this info on a 2011 camaro ss 113 lsa cammed with long tubes?

Absolutely, but at this point in time...such as in my case it may be a guessing game. I think we are getting closer though.

[schpenxel]

Hmmm... Should have read it more closely or at least when it wasn't so late at night

Next stupid question... How is the ecm actually calculating this or rather how did the OE come up with the delay times - was it how it was done in testing (from the moment the injector fired) - this just seems to put way too much into the mix since injection soi is all over the place depending on engine load and size of injector - OR is it simply based off of when the cylinder was supposed to fire (I would assume it would make more since to base the "root" numbers of the time frame off of this instead of the injectors, since a power stroke is more or less set in stone with engine rotation)

Even then there are a lot of other things coming into the equation, such as when the exhaust valve is opening - this would inevitably change when the exhaust pulse hits the O2... If it's off of cylinder firing, it would be extremely easy to dial in an O2's delay function... All one would need to do is set the airflow mode table to one mode (a mode that corresponds directly with the proportional and integrator tables) / run the engine at that airflow / kill a cylinder (preferably do one at the front and then the rear most of each bank - not sure on this - I know the OE bases most off of cylinder one, so wondering if it wouldn't be best just to use the front cylinders?) / average those out in the measured ms from time of noted miss to the time the O2 saw a lean transition... Wash, rinse, repeat for corresponding modes and airflow levels...

It's just the way it was done in the testing, didn't seem to yield numbers correctly correlating with what's in the tune - they follow the correct curve - just give different numbers - unless I was looking at the wrong thing posted above?

I wonder if the formula can be simplified since the volume between injector and intake valves isn't changing. The volume from exhaust valve to O2 sensors is the big one that is changing, and obviously air density on FI applications. Seems that on an average longtube setup the only real change would be exhaust volume basically... I need to look at the formulas to see if that can be used to simplify things some

I really need to sit down and read the whole paper..