Tuning MAF and VVE at the Same Time

[Capt Chewy]

I ignore these outliers that clearly point to something else going on and focus on a visually smooth graph and transition in the area of the anomaly. If the area in question is not repeatable on the next log I usually discount it. Also I have found that increasing the cell count sometimes gets rid of them. Obviously the exception is if it is in multiple cells and signifies a pattern.

Thats what i sarted doing. Any where it would just pass over the cell during a transient i just copy around it then interpolate between the cells in the vve i skipped to keep the table smoother.

[Cringer]

Cringer sent pm with formula

Thank you Capt Chewy!

[Cringer]

So something was off with my original GMVE_CL formula. Thanks to Capt Chewy, the results I am seeing are now much closer to what I calculate (and what I assert is the correct way to do this) using Dynamic Air.

Perhaps the confusion for some people is that in the HPT Editor descriptions are questionable?
Engine > Airflow > Dynamic
[ECM] 3003 - Dynamic Airflow High RPM Disable
[ECM] 3033 - Dynamic Airflow High RPM Re-Enable
On the surface this would seem to indicate that VVE is the Dynamic Airflow. But it is not.

However, in the HPT Scanner, you can log [Virtual Volumetric Efficiency Airflow] , [Dynamic Airflow] , and [Mass Airflow] and they all show unique and different numbers. And again, the [Dynamic Airflow] PID is result of the Kalman filter and that is the only airflow number used by the ECM that determines injector pulse width which in turn is effecting the fuel trims. So we should be using [Dynamic Airflow] and the fuel trims to calibrate both the MAF and VVE.

[Capt Chewy]

I'm logging Maf, Dynamic, Ve, and Gm air flow all in g/sec. The closer i get with the gmve in both maf and vve the closer those lines come together to be the same. Awesome stuff.

[DCZ]

Anything besides DFCO or COT need to be disabled when using VVE Assistant?

[FC3S Murray]

I have been tuning the MAF and VVE the "old school" way since 2017 on my cammed L9H Denali 1/2 ton.

The E38 ecu is so good at processing data (compared to my old 0411 in my swapped RX-7) that I firmly believe sperating the two fueling models is key to "perfection".

I have my VVE within +/- 0.5% LTFT in CL and my MAF is +/- 1%...sometimes +/- 0.25% consistently even with baro changes.

When blended back together, the truck is snorty, linear and shifts like a dream (I run a pretty aggressive 6L80E trans tune I configured using BlueCat's program).

Fun fact, I often switch between running CL or OL for months at a time. My mileage never changes much (19.5 mpg city... lead foot haha!) and the two feel damn near identical but my point is, it just takes TIME and lots of data/road time to get the airflow models perfect. I am very OCD and I don't even want to calculate how much time I spend datalogging out of curiosity but regardless the proven way works.

Wash, rinse, repeat

It's that simple.

Been tuning since 2013, always willing to pick up new tricks IF they quicken the process effectively.

My dynamic airflow is damn near spot on with my MAF airflow doing it the old way.

Not trying to be nay sayer, just want to ensure that beginners know the fundamentals still work.

[Cringer]

Anything besides DFCO or COT need to be disabled when using VVE Assistant?

The tool itself does not care what you have enabled. But it just comes down to being diligent about collecting quality data. As such, I would suggest disabling DFCO/CFCO, LTFT, Desoot, and COT, as well as driving with slow and steady throttle inputs.

[smokeshow]

So we should be using [Dynamic Airflow] and the fuel trims to calibrate both the MAF and VVE.

Dynamic airflow is a state estimator. It provides feedback to the ECM about transient airflow behavior that the controller is blind to in real time. It exists because direct measurement of the dynamic behavior is not possible.

A state estimator should never be used to correct the airflow model.

[Cringer]

Dynamic airflow is a state estimator. It provides feedback to the ECM about transient airflow behavior that the controller is blind to in real time. It exists because direct measurement of the dynamic behavior is not possible.

A state estimator should never be used to correct the airflow model.

So I just did a test. I suggest you do the same to challenge and test your assumptions. You may prove yourself right, or you may learn something, either way the entire community benefits.

I edited my MAF table:
4650 Hz - 6900 Hz: multiplied by .9 (leaner)
7050 Hz - 8400 Hz: multiplied by .8 (leaner)

VVE Table:
RPM columns 1400 - 2400 RPM: multiplied by 1.2 (richer)
RPM columns 2600 - 4000 RPM: multiplied by 1.4 (richer)


I went for a drive and and using Dynamic Air as the benchmark, I got the results reflected back exactly as predicted. And when you look at the chart vs time, plotting MAF g/s, VVE g/s, and Dynamic Air g/s you can see these hard change overs. Furthermore, and MOST IMPORTANTLY, when I took the fuel trims and multiplied it with the Dynamic Air, I got the original values from my previous tune (where the fuel trims were near zero).

Bottom Line:
Is Dynamic Air a state estimator? Maybe? Who cares. What is important is that Dynamic Air is what is in control of the injector pulse width and thus this is what the fuel trims are in direct response to.

It does not make any logical sense for GM software engineers to create a complex Kalman filter to take into account MAF, VVE (which is augmented with transient fueling) only to ignore it most of the time, let alone force the ECM to adopt either/only a MAF only air model, then switch to VVE air model, and maybe sometimes hop over to a Dynamic Air model all on a whim.

[sgod1100]

So something was off with my original GMVE_CL formula. Thanks to Capt Chewy, the results I am seeing are now much closer to what I calculate (and what I assert is the correct way to do this) using Dynamic Air.

Perhaps the confusion for some people is that in the HPT Editor descriptions are questionable?
Engine > Airflow > Dynamic
[ECM] 3003 - Dynamic Airflow High RPM Disable
[ECM] 3033 - Dynamic Airflow High RPM Re-Enable
On the surface this would seem to indicate that VVE is the Dynamic Airflow. But it is not.

However, in the HPT Scanner, you can log [Virtual Volumetric Efficiency Airflow] , [Dynamic Airflow] , and [Mass Airflow] and they all show unique and different numbers. And again, the [Dynamic Airflow] PID is result of the Kalman filter and that is the only airflow number used by the ECM that determines injector pulse width which in turn is effecting the fuel trims. So we should be using [Dynamic Airflow] and the fuel trims to calibrate both the MAF and VVE.

what was incorrect in the original formula that you posted?

[Cringer]

what was incorrect in the original formula that you posted?

Here is my GMVE_CL formula history. To be clear, I do not use these anymore.


The better version from Capt Chewy
([50040.71]*(1+((.01*[50116]*100)+(.01*[50114]*100)))-(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4))/(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4))*100


Original Version I had (I think it was from someone's Excel spreadsheet I found on the forum?):
(([16.71.avg(500)]*(1+(.01*([6.156.avg(500)]+[7.156.avg(500)])+.01*([8.156.avg(500)]+[9.156.avg(500)]))/2)-[50070.56.avg(500)]/60*4*[11.92.avg(500)]*[2312.avg(500)]/[2126.240.avg(500)])/([50070.56.avg(500)]/60*4*[11.92.avg(500)]*[2312.avg(500)]/[2126.240.avg(500)]))*100

[sgod1100]

Here is my GMVE_CL formula history. To be clear, I do not use these anymore.


The better version from Capt Chewy
([50040.71]*(1+((.01*[50116]*100)+(.01*[50114]*100)))-(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4))/(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4))*100


Original Version I had (I think it was from someone's Excel spreadsheet I found on the forum?):
(([16.71.avg(500)]*(1+(.01*([6.156.avg(500)]+[7.156.avg(500)])+.01*([8.156.avg(500)]+[9.156.avg(500)]))/2)-[50070.56.avg(500)]/60*4*[11.92.avg(500)]*[2312.avg(500)]/[2126.240.avg(500)])/([50070.56.avg(500)]/60*4*[11.92.avg(500)]*[2312.avg(500)]/[2126.240.avg(500)]))*100

ok, thanks. You are using those shorter formulas that you created taking the dynamic air error and applying it to the maf correct? Then using the same formula but comparing the VE to the dynamic error

[Cringer]

ok, thanks. You are using those shorter formulas that you created taking the dynamic air error and applying it to the maf correct? Then using the same formula but comparing the VE to the dynamic error

No, those are the standard GMVE formulas. I do not use them. I use this instead:
(([2320.71]+([2320.71]*([50116.156]+[50114.156])/100))-[2311.71])/[2311.71]*100


EDIT: I also updated the original post...as those formulas had copy/paste errors on my part.

[TheMechanic]

No, those are the standard GMVE formulas. I do not use them. I use this instead:
(([2320.71]+([2320.71]*([50116.156]+[50114.156])/100))-[2311.71])/[2311.71]*100


EDIT: I also updated the original post...as those formulas had copy/paste errors on my part.

What specific channels would you have to best use your VVE tool? I was wanting to slim it down to get the data rate up. Want to eliminate the "stuff" that is not needed.

[sgod1100]

No, those are the standard GMVE formulas. I do not use them. I use this instead:
(([2320.71]+([2320.71]*([50116.156]+[50114.156])/100))-[2311.71])/[2311.71]*100


EDIT: I also updated the original post...as those formulas had copy/paste errors on my part.

Gotcha, what you are shown above actually what I was referring to

[Cringer]

What specific channels would you have to best use your VVE tool? I was wanting to slim it down to get the data rate up. Want to eliminate the "stuff" that is not needed.

For Gen4 bare min using MAF and VVE, it would be:

• RPM
• MAP (or MAP/Baro Ratio Pressure Math for Gen 5)
• DynAir
• MAF g/s
• VVE g/s*
• STFT
• LTFT (although if you are concerned about polling rate you can elininate this and then remove the reference to it in the formula)
• Commanded EQ
• WB Lambda

* You are on the custom 2 bar OS though right? So you might need to alter some of these channels. Since you don't have to calc coefficients, the only benefit here is that you can average multiple logs into one.


Then additionally, I can't remember if I posted about this on the forum or not, but I create a MAF tool, named MAF Assistant (creative I know, but it pairs nicely with the aptly named VVE Assistant). You will need to watch this video to learn to use it:
https://youtu.be/ccF7pv1B-Xg
Then to use this, you will need to add the MAF Hz in addition to the list above.

[smokeshow]

So I just did a test. I suggest you do the same to challenge and test your assumptions. You may prove yourself right, or you may learn something, either way the entire community benefits.

I edited my MAF table:
4650 Hz - 6900 Hz: multiplied by .9 (leaner)
7050 Hz - 8400 Hz: multiplied by .8 (leaner)

VVE Table:
RPM columns 1400 - 2400 RPM: multiplied by 1.2 (richer)
RPM columns 2600 - 4000 RPM: multiplied by 1.4 (richer)


I went for a drive and and using Dynamic Air as the benchmark, I got the results reflected back exactly as predicted. And when you look at the chart vs time, plotting MAF g/s, VVE g/s, and Dynamic Air g/s you can see these hard change overs. Furthermore, and MOST IMPORTANTLY, when I took the fuel trims and multiplied it with the Dynamic Air, I got the original values from my previous tune (where the fuel trims were near zero).

Bottom Line:
Is Dynamic Air a state estimator? Maybe? Who cares. What is important is that Dynamic Air is what is in control of the injector pulse width and thus this is what the fuel trims are in direct response to.

It does not make any logical sense for GM software engineers to create a complex Kalman filter to take into account MAF, VVE (which is augmented with transient fueling) only to ignore it most of the time, let alone force the ECM to adopt either/only a MAF only air model, then switch to VVE air model, and maybe sometimes hop over to a Dynamic Air model all on a whim.

It doesn't make sense because you don't understand the 'whim'... Dynamic air compensates for transients...meaning not steady state. Airflow models must be calibrated at steady state, period.

Post the data from your test.

[Cringer]

It doesn't make sense because you don't understand the 'whim'... Dynamic air compensates for transients...meaning not steady state. Airflow models must be calibrated at steady state, period.

Post the data from your test.

When you watch dynamic air, it sure seems to track steady state just fine. Stuff attached...the log file for the initial tune that demonstrates acceptable trims is too large for a forum upload. So I guess you are going to have to take my word for it. I am happy to email it to you if you really want it though. But I did post a screenshot showing just the simple STFT+LTFT results (no dynamic air or GMVE calcs, just the actual trims only).

This is going to take a few posts for the attachments I think...I keep getting a forum error stating "The text that you have entered is too long (155228 characters). Please shorten it to 15000 characters long."

Screenshot 2023-05-23 234132.jpg

[Cringer]

Attempting to attach the data...

1_Init_Tune_Good_Trims.hpt
2_Testing_Intentional_Rich_Lean.hpt
2_Testing_Intentional_Rich_Lean.hpl

[smokeshow]

When you watch dynamic air, it sure seems to track steady state just fine. Stuff attached...the log file for the initial tune that demonstrates acceptable trims is too large for a forum upload. So I guess you are going to have to take my word for it. I am happy to email it to you if you really want it though. But I did post a screenshot showing just the simple STFT+LTFT results (no dynamic air or GMVE calcs, just the actual trims only).

This is going to take a few posts for the attachments I think...I keep getting a forum error stating "The text that you have entered is too long (155228 characters). Please shorten it to 15000 characters long."

Screenshot 2023-05-23 234132.jpg

If you filter for steady state, then of course dynamic airflow gives you acceptable data. It's equivalent to the MAF airflow, or VE airflow if MAF is disabled. The flow isn't exhibiting dynamic behavior at steady state. The inherent risk with dynamic airflow for error calculation is mistaking a transient for steady state and baking that error into the MAF and/or GMVE which need steady state data.