What is dynamic airflow

[SiriusC1024]

Yeah it's the same thing as saying tune VVE based on MAF. It only works in steady state.

The difference between dynair, MAF, GMVE isn't linear. It's weighted based on VE correction factor. Can't simply apply the %error between say dynair and GMVE for all conditions. Pay attention to the plots in the screenshot. Every elbow in the MAF trace is matched by a movement in the opposite direction in dynair. So can't just say:

If MAF was reporting [40 g/s] at this Hz, then what do we need to multiply it by in order to get it up to the correct airflow?
[40 g/s] * [1.27875%] = [51.15 g/s]
Copy Paste Special [1.27875]

If VVE was reporting [65 g/s] at this RPM:MAP, then we need to do in order to get it up to the correct airflow?
[65 g/s] * [0.787%] = [51.15 g/s]
Copy Paste Special [0.787]

As you can see this is very simple.

Theoretically you could populate fuel trims for all VVE cells by holding steady state for each, but theoretically you could also blow your motor.

I can't figure out if Cringer's success with this method, where conventional ones failed, was a result of smoothing the MAF signal from dynair filtering MAF itself, dynair including VE correction, or a combination of both.

I'm not convinced of simultaneously tuning VVE and MAF. The qualification that this has to be done in steady state points to the question of how we're tuning the part that's there to address non-steady state conditions. A transient to me is due to delay in signal to O2 sensors or what's handled in the Transient tab. Things like rapid throttle blips. A condition isn't a transient simply because it makes an assumption fall apart. That assumption would be that MAF and MAP are polling data from the same conditions.

I'd still be interested in seeing a dynair vs GMVE trace in SD mode.

[rabbs88]

For what it's worth, I originally tuned my car (2013 Camaro w/headers + intake) doing MAF and VVE separately with my wideband in open loop. Fuel trims back in closed loop after were within 1-2%. All was well. About a month ago I found Cringer's method using dynamic air to do MAF and VVE simultaneously so I flashed my stock MAF curve and VVE table back in and tried his way out of curiosity. Within 2 logs (Steady state with minimal transients) my fuel trims are all within 1-2% again and took significantly less time to dial in.

Smokeshow, to be fair I haven't tried your method to tune MAF and VVE simultaneously yet but at this point I almost feel obligated to see how well it works. If I get some time and good weather next week I might give it a shot and report back. Regardless, if both methods work well enough then who cares? Unless someone is smart enough to figure out exactly how Dynamic Air is calculated then there won't really be a right way to do anything.

I'm no pro tuner by any means. Just figured I would share my experience so far.

[smokeshow]

Yeah it's the same thing as saying tune VVE based on MAF. It only works in steady state.

The difference between dynair, MAF, GMVE isn't linear. It's weighted based on VE correction factor. Can't simply apply the %error between say dynair and GMVE for all conditions. Pay attention to the plots in the screenshot. Imagine drawing the average between GMVE and MAF. That's not the same line as dynair is it? So can't just say:


Theoretically you could populate fuel trims for all VVE cells by holding steady state for each, but theoretically you could also blow your motor.

I can't figure out if Cringer's success with this method, where conventional ones failed, was a result of smoothing the MAF signal from dynair filtering MAF itself, dynair including VE correction, or a combination of both.

I'm not convinced of simultaneously tuning VVE and MAF. The qualification that this has to be done in steady state points to the question of how we're tuning the part that's there to address non-steady state conditions. A transient to me is due to delay in signal to O2 sensors or what's handled in the Transient tab. Things like rapid throttle blips. A condition isn't a transient simply because it makes an assumption fall apart. That assumption would be that MAF and MAP are polling data from the same conditions.

I'd still be interested in seeing a dynair vs GMVE trace in SD mode.

Dynamic airflow and VVE airflow are identical in SD mode, as expected.

sd.PNG

Lot of different transient-type conditions can occur in an engine. O2 readings, port wall wetting...definitely. But this means the airflow itself is in a transient. And it isn't calibratable with fuel trims. The calibration is the found by tuning a virtual induction system via a stand-alone math model. Without being able to measure the real time cylinder air mass, this is the next best thing.

[smokeshow]

Smokeshow, to be fair I haven't tried your method to tune MAF and VVE simultaneously yet but at this point I almost feel obligated to see how well it works. If I get some time and good weather next week I might give it a shot and report back. Regardless, if both methods work well enough then who cares? Unless someone is smart enough to figure out exactly how Dynamic Air is calculated then there won't really be a right way to do anything.

My original post here explains precisely how it works, with references to the algorithm patents. And I can verify that further, having seen virtually all of this source code while working for GM myself. You may not be able to get good results with the simultaneous approach I described - I'd say about half of the people that try it can't get it right. It is sensitive to post-process filtering, which you are ultimately responsible for. And it makes no consideration for changes to MAT versus the filtered estimate, nor O2 sensor feedback issues. However what it won't do is obtusely bury external error into your calibration, for better or for worse.

[SiriusC1024]

Dynamic airflow and VVE airflow are identical in SD mode, as expected.

I was hoping dynair would smooth it out. Oh well. So not a Kalman filter?

However what it won't do is obtusely bury external error into your calibration, for better or for worse.

I think maybe Cringer's method is accurate enough that the end result is within the acceptable 5%, 2%, etc. fuel trim error bars that most tuners use as a rule of thumb. Technically incorrect, but it gets the job done. That's where I'm at with this.

[smokeshow]

I was hoping dynair would smooth it out. Oh well. So not a Kalman filter?

I'm sure it did do some smoothing. Dynamic air allows for substantial noise rejection when tuned.

There is a Kalman filter component of dynamic airflow - the prediction coefficients. Not all applications use them though.

[hjtrbo]

I still use smokeshows method for both, not that it matters for this conversation. Everyone agrees MAF at steady state will line up with dynamic air. Both smokes and cringers methods work if you know what you're doing. The secret sauce is the filters which I've posted many a times before. Only going off a bad memory but cringers filters looked familiar.

[sgod1100]

Insert "eating popcorn meme here"...continue on guys...lol

[SiriusC1024]

In the spirit of Cringer's quest for The Truth there can be only one. To find consensus I think we need to appeal to a higher authority. Lord kingtal0n I beseech thee...

[Alvin]

I purpose doing this in a friendly way..

It would be neat to have a contest among the methods.

Start out with a completely stock VVE table with a medium size cam. Everyone gets a shot at the basefile and log and then see how close we get on the first file..

[gtstorey]

In the spirit of Cringer's quest for The Truth there can be only one. To find consensus I think we need to appeal to a higher authority. Lord kingtal0n I beseech thee...

Better yet, both Cringer and Smokeshow can both tune cars (Vette, Camaro, I'm not picky just needs to be the same) and I'll drive the crap out of them (shouldn't take more than a year or two) and let everyone know which drives better. I'll even pay for some dyno time to help clarify things. Send some gas money though.

[hjtrbo]

I purpose doing this in a friendly way..

It would be neat to have a contest among the methods.

Start out with a completely stock VVE table with a medium size cam. Everyone gets a shot at the basefile and log and then see how close we get on the first file..

Gen 4 all stars

[Cringer]

Here is a great example of why using DynAir is so great. This is my latest log and my fuel trims are skewed a little since I have been playing with EOIT, but that gives a good opportunity to inspect this.

We are zoomed to 30sec and we can see throttle is slightly increasing, but overall pretty smooth and steady input.
MAF and DynAir agree with each other 124 g/s (so the ECM says we are in steady state)
Fuel trims are both negative at a -5% average between the two banks (LTFT is disabled btw). This tells us that 124 g/s is too much air...
And we can see VVE is reporting in at 113 g/s

So if MAF & DynAir are rich we need to reduce MAF by the fuel trims:
124.558 g/s * .95 = 118.33 g/s

And since VVE is still less than this newly identified corrected airflow, the DynAir VVE_CL formula at the top says you need to ADD air to VVE!
113.625 g/s * 1.0547 = 119.84 g/s

We can now see that both MAF and VVE are converging around 119 g/s. Pretty good agreement, which is what we want.

YET...WHAT DOES GMVE TELL US TO?! It says VVE is rich and to REMOVE air to the tune of 3.76%! Why would I want to remove air when the fuel trims are rich for MAF and VVE g/s is lower than MAF?
113.625 g/s * .9624 = 109.35 g/s
This will make fueling worse by 9% (from 119.84 g/s target)

Here is the GMVE formula I am using. If there is an error then please enlighten me:

([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

chartVtime.jpg

[Cringer]

And to add to that post, maybe you are thinking, "well sure that is chart vs time and you don't have transients filtered out like you would in the chart."

Here is the filter I am using:

[2517.161.avg(1500)]=0 and [2517.161.avg(-200)]=0 and (abs([50090.156.slope(1500)])+abs([50090.156.slope(-500)]))<2 and ([6310]=9 OR [6310]=14 OR [6310]=15)=0

And here is the aggregated data with a zoom range only applied for that same 30sec window. If I do the full range the results are nearly identical btw.

DynAir method showing I need to add fuel to VVE
VVE_CL.jpg

GMVE method showing I need to remove fuel from VVE
GMVE_CL.jpg

[Cringer]

For what it's worth, I originally tuned my car (2013 Camaro w/headers + intake) doing MAF and VVE separately with my wideband in open loop. Fuel trims back in closed loop after were within 1-2%. All was well. About a month ago I found Cringer's method using dynamic air to do MAF and VVE simultaneously so I flashed my stock MAF curve and VVE table back in and tried his way out of curiosity. Within 2 logs (Steady state with minimal transients) my fuel trims are all within 1-2% again and took significantly less time to dial in.

Smokeshow, to be fair I haven't tried your method to tune MAF and VVE simultaneously yet but at this point I almost feel obligated to see how well it works. If I get some time and good weather next week I might give it a shot and report back. Regardless, if both methods work well enough then who cares? Unless someone is smart enough to figure out exactly how Dynamic Air is calculated then there won't really be a right way to do anything.

I'm no pro tuner by any means. Just figured I would share my experience so far.

This is great feedback and the part I love most is your statement: and took significantly less time to dial in
as that also part of my quest (accuracy first, speed second).

I would love it if you could try the same experiment using the GMVE method and report your findings since you would be a non-biased 3rd party.

[smokeshow]

I don't need to say any more. That's the algorithm, with the reference. For every gasoline engine controller for the past 25 years. If you can't get a proper method right, don't use one. Use whatever method you like....install a carburetor, makes no difference to me lol. As Dave would say..."not a Dave Steck problem". Same here Dave...same here. Carry on, everyone.

[Cringer]

I don't need to say any more. That's the algorithm, with the reference. For every gasoline engine controller for the past 25 years. If you can't get a proper method right, don't use one. Use whatever method you like....install a carburetor, makes no difference to me lol. As Dave would say..."not a Dave Steck problem". Same here Dave...same here. Carry on, everyone.

You see? This is my problem. I posted very detailed information above from my log with the GMVE formula looking for feedback, but you have provided nothing in return to help me or anyone else. If you are going to go on the offensive and educate us about Dynamic Air voodoo, then please do not dismiss my post. I would love to get to the bottom of this, but if you cannot rectify the data, formula, and logs, then we are at an impasse and I am afraid this is all the more reason for me to keep raping DynAir. I will take a current log over a 20 year old dusty patent any day.

[GHuggins]

Here is a great example of why using DynAir is so great. This is my latest log and my fuel trims are skewed a little since I have been playing with EOIT, but that gives a good opportunity to inspect this.

We are zoomed to 30sec and we can see throttle is slightly increasing, but overall pretty smooth and steady input.
MAF and DynAir agree with each other 124 g/s (so the ECM says we are in steady state)
Fuel trims are both negative at a -5% average between the two banks (LTFT is disabled btw). This tells us that 124 g/s is too much air...
And we can see VVE is reporting in at 113 g/s

So if MAF & DynAir are rich we need to reduce MAF by the fuel trims:
124.558 g/s * .95 = 118.33 g/s

And since VVE is still less than this newly identified corrected airflow, the DynAir VVE_CL formula at the top says you need to ADD air to VVE!
113.625 g/s * 1.0547 = 119.84 g/s

We can now see that both MAF and VVE are converging around 119 g/s. Pretty good agreement, which is what we want.

YET...WHAT DOES GMVE TELL US TO?! It says VVE is rich and to REMOVE air to the tune of 3.76%! Why would I want to remove air when the fuel trims are rich for MAF and VVE g/s is lower than MAF?
113.625 g/s * .9624 = 109.35 g/s
This will make fueling worse by 9% (from 119.84 g/s target)

Here is the GMVE formula I am using. If there is an error then please enlighten me:

([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

chartVtime.jpg

OK, I think I get why this works so well and why the other methods don't work especially for certain individuals and people can say I'm wrong or chime in however they like.

Here's the deal or at least how it "clicked" with me just reading through all of this... The airmass calcs and other calcs not using fueling requires almost everything to be correctly dialed in as directly shown in Cringer's example believe it or not. They also require the temp readings and everything to be spot on for the related setup. BY THIS, what I'm referring to are the ERRORS - ERRORS that fueling is compensating for... O2's for example to be reliable have to be used with stock switching settings combined with an unaltered induction tube so as the MAF is in it's original orientation and location for it's g/s airflow input along with an unaltered engine so airflow through it is unaltered as well so the (timing) from MAF to O2 is inevitably the exact same. Keep in mind everything changed WILL CHANGE the airflow controlled corrections in the calibration. Well, all of these are usually different on a modified engine right? We've installed cams, CAI's, LT's, Different Heads, Entirely different engines/whatever into the puzzle. Now we have the O2's completely off and we can dial them in as best as possible but honestly they'll never be 100% because we don't have the background controls or lab tools to do so 100% so now we have "errors" in our fuel reporting. Well these Errors from Trim reporting are the whole reason why the Airflow calcs no longer work and why the Airflow Calcs can sometimes be WAY off... ESPECIALLY if the calibrator doesn't bother to try and dial a lot of the necessary things in such as the O2's...

Cringer's formula WORKS WITH these ERRORS. Encompassing them straight into both models, balancing and forcing them to be the same and work with the problems... I like the idea of this, because honestly I don't care how much time you put into the calibration modeling - We just don't have everything to be able to 100% dial it in nor do we have the knowledge necessarily to do so... Heck if we can honestly realistically get them within 60 to 70 percent considering what little we can adjust we're probably doing good. I've been playing with some throttle tables on genV's for instance for boost control and have learned a butt load on how they work with the other tables and more importantly how they relate to one another themselves. I would love to have better access to the transient and fueling controls in the gen 4's, but I know it would take a lot of experimenting and reading on patents to even begin to slightly grasp how they work and that's me, so for everyone to have access to something like that - well I could understand some of the problem.... In fact I would most likely have to turn to someone like Smoke or Cinger to even get an idea myself. I'm still trying to wrap my head around Sonic Airflow.

Either way this is what I see happening and why Cringer's may work better is simply because the errors involved with fueling are already included into it. I know using a calculated output from the ECM as an input isn't necessarily the best solution and sometimes can require applying it to different tables. I do this myself with torque modeling. Doesn't work 100% in every situation, but it does work really well for the majority.....

Anyway, this is what I observed. If it's wrong then it's wrong...

[Cringer]

OK, I think I get why this works so well and why the other methods don't work especially for certain individuals and people can say I'm wrong or chime in however they like.

Here's the deal or at least how it "clicked" with me just reading through all of this... The airmass calcs and other calcs not using fueling requires almost everything to be correctly dialed in as directly shown in Cringer's example believe it or not. They also require the temp readings and everything to be spot on for the related setup. BY THIS, what I'm referring to are the ERRORS - ERRORS that fueling is compensating for... O2's for example to be reliable have to be used with stock switching settings combined with an unaltered induction tube so as the MAF is in it's original orientation and location for it's g/s airflow input along with an unaltered engine so airflow through it is unaltered as well so the (timing) from MAF to O2 is inevitably the exact same. Keep in mind everything changed WILL CHANGE the airflow controlled corrections in the calibration. Well, all of these are usually different on a modified engine right? We've installed cams, CAI's, LT's, Different Heads, Entirely different engines/whatever into the puzzle. Now we have the O2's completely off and we can dial them in as best as possible but honestly they'll never be 100% because we don't have the background controls or lab tools to do so 100% so now we have "errors" in our fuel reporting. Well these Errors from Trim reporting are the whole reason why the Airflow calcs no longer work and why the Airflow Calcs can sometimes be WAY off... ESPECIALLY if the calibrator doesn't bother to try and dial a lot of the necessary things in such as the O2's...

Cringer's formula WORKS WITH these ERRORS. Encompassing them straight into both models, balancing and forcing them to be the same and work with the problems... I like the idea of this, because honestly I don't care how much time you put into the calibration modeling - We just don't have everything to be able to 100% dial it in nor do we have the knowledge necessarily to do so... Heck if we can honestly realistically get them within 60 to 70 percent considering what little we can adjust we're probably doing good. I've been playing with some throttle tables on genV's for instance for boost control and have learned a butt load on how they work with the other tables and more importantly how they relate to one another themselves. I would love to have better access to the transient and fueling controls in the gen 4's, but I know it would take a lot of experimenting and reading on patents to even begin to slightly grasp how they work and that's me, so for everyone to have access to something like that - well I could understand some of the problem.... In fact I would most likely have to turn to someone like Smoke or Cinger to even get an idea myself. I'm still trying to wrap my head around Sonic Airflow.

Either way this is what I see happening and why Cringer's may work better is simply because the errors involved with fueling are already included into it. I know using a calculated output from the ECM as an input isn't necessarily the best solution and sometimes can require applying it to different tables. I do this myself with torque modeling. Doesn't work 100% in every situation, but it does work really well for the majority.....

Anyway, this is what I observed. If it's wrong then it's wrong...

Thank you for offering that extra bit of clarification and putting into words that which I cannot. Welcome to the Dark Side!

[hjtrbo]

Here is a great example of why using DynAir is so great. This is my latest log and my fuel trims are skewed a little since I have been playing with EOIT, but that gives a good opportunity to inspect this.

We are zoomed to 30sec and we can see throttle is slightly increasing, but overall pretty smooth and steady input.
MAF and DynAir agree with each other 124 g/s (so the ECM says we are in steady state)
Fuel trims are both negative at a -5% average between the two banks (LTFT is disabled btw). This tells us that 124 g/s is too much air...
And we can see VVE is reporting in at 113 g/s

So if MAF & DynAir are rich we need to reduce MAF by the fuel trims:
124.558 g/s * .95 = 118.33 g/s

And since VVE is still less than this newly identified corrected airflow, the DynAir VVE_CL formula at the top says you need to ADD air to VVE!
113.625 g/s * 1.0547 = 119.84 g/s

We can now see that both MAF and VVE are converging around 119 g/s. Pretty good agreement, which is what we want.

YET...WHAT DOES GMVE TELL US TO?! It says VVE is rich and to REMOVE air to the tune of 3.76%! Why would I want to remove air when the fuel trims are rich for MAF and VVE g/s is lower than MAF?
113.625 g/s * .9624 = 109.35 g/s
This will make fueling worse by 9% (from 119.84 g/s target)

Here is the GMVE formula I am using. If there is an error then please enlighten me:

([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

chartVtime.jpg

Your formula is wrong. You're incorrectly normalising your fuel trims and there is an erroneous bracket.

([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

Here is a 3% discrepancy between your implementation of smokes formula, which would account for the error you are seeing between the 2 methods.
Screenshot 2023-08-19 094129.png

Try this
([50040.71]*(1+(.01*[50116.156]+.01*[50114.156]))-(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4))/(([50030.92]*[2312]/[2126.240])*[50070.56]/60*4)*100

Considerations:
Increase your data polling rate
Try average function