Reading MLV Data Logs

[whittlebeast]

This is the sort of thing that intrigues me. Almost all high performance motorcycles use ITB intake manifolds fed by a fairly large air box. Almost all modern inline 4 cylinder engines have a single fairly large throttle body feeding a log style intake and four fairly long intake ports. Here is screen shot of what one of those "typical automotive intakes" looks like when plotted with MAPxRPM vs Duty Cycle and TPS as the color. Note that this is a Miata motor.

http://www.nbs-stl.com/tuning/MAPxRPM%2 ... 0Motor.png

Notice how you get a nearly straight line.

Now, here is my typical personal ITB motor plotted the same way.

http://www.nbs-stl.com/CrxTypeR/MAPxRPM ... 0Motor.png

Notice the Duty Cycle (and resulting power) is very much a function of throttle position. If I make the color RPM, I also get a pattern. The required fueling is very much dependent on throttle position and RPM. Both motors are 4 cylinder, 4 valve motors of similar size. Mine happens to be a VTec motor but I switch to high lift any time above 5450 RPM.

http://www.nbs-stl.com/CrxTypeR/MAPxRPM ... 0Motor.png

Note that in this screen shot, I filtered out RPM below 3500 RPM. This happens to be about 100 MB of data coming off my new MS3 Ultimate collecting data at 200 samples per sec.

I find this stuff amazing.

Andy

[whittlebeast]

Here is a great log from a turbo drift car. The blue line indicates where the motor passes thru the VE map on a typical pass. In the field is how the motor needs to have the VE adjusted to better get the motor to hit AFR targets. This is all done with a MS3 Ultimate logging at 320 samples per sec and MLVHD histograms.

http://www.nbs-stl.com/tuning/Turbo%20D ... %20Adj.png

To create this custom field, I used this formula.

(([AFR 1 Error]>1?1:[AFR 1 Error]<-1?-1:[AFR 1 Error])*.07)+1

Have fun tuning.

Andy

[whittlebeast]

Here is how I personally find lambda delay. I find a place in the log where the driver suddenly lifted out of the throttle and use the time measuring function built into MLV to see how long it to for the wideband to register a change/stabilize. In this motor, the delay is about .2 sec as seen at the bottom of the screen in the white box.

http://www.nbs-stl.com/tuning/Turbo%20D ... 0Delay.png

Notice how quickly the motor the motor goes from full throttle and hard accel to backing down as seen in the RPM per Sec trace. This ability is rare in a turbo car.

Have fun tuning

Andy

[whittlebeast]

I then look at the Target AFR vs the Actual AFR and in the right screen I look at the percentage that the the AFR is missing the targets.

http://www.nbs-stl.com/tuning/Tutbo%20D ... %20AFR.png

[Rick Finsta]

Andy, what stuck out to me in srb94's post was that they found the delay to be tied to a number of crankshaft rotations. Would you be able to take a few of the logs you have for finding sensor delay and seeing about creating a function that reduces it to crank degree rotation? That could also maybe be added to the code later on since it would be a single value for an engine rather than a 2D or 3D plot.

In fact, now I wonder if Phil didn't write the original MLV auto-tune code delay to be number of crank degrees rotation, since it seems to not need much tweaking relative to RPM once you have a close number picked out with the slider?

[whittlebeast]

My bet is Lambda delay is fairly close to a function of MAPxRPM. It will work out to the raw volume of exhaust being processed by the exhaust tubing. Writing a formula to work with all the different data rated we are dealing with now makes things more difficult. I am fairly sure the existing Lambda delay filters are number of records based.

Andy

[whittlebeast]

Here is what a near textbook data log looks like when a motor is ready to be tuned.

http://www.nbs-stl.com/tuning/Textbook%20Data%20log.png

Notice that I have a little filtering turned on as most of the traces have a factor of 2 activated. I am getting 192 samples per sec out of my MS3 Ultimate.

The MAP, RPM and TPS are all smooth and the readings look reasonable.

CLT, MAT and Battery voltage are all steady but system voltage looks a little low as 13.5 to 14.4 would be more typical.

You can see in the PW vs PW Base That AT is setting in and looks reasonable. This also looks similar on the Duty Cycle vs Duty Cycle Base

At the bottom, you can see the Engine code jumping up and down as AE turns on and off. I happen to have DE turned of in this tune.

Time to start tuning on this motor as all the basic sensors look good.

Andy

[whittlebeast]

I was doing a little high speed data logging testing and got bored so I tested the absolute response time of of the map sensor at two different RPMs and two different MAP filtering settings. I chose 1000 RPM and 8000 RPM and filtering of 50 and 100. The faint line is with the filtering set at 50. Notice at the bottom is the trace is the number of samples per sec that the data logging was running at and the number of ms per engine cycle. Way cool stuff bet I have now idea how to take this to the bank. All testing was done on my MS3X setup on a stim. I had about 1 foot of hose on the MAP sensor. Here is what I got for results.

At 1000 RPM
http://www.nbs-stl.com/tuning/MAP%20Res ... %20RPM.png

At 8000 RPM
http://www.nbs-stl.com/tuning/MAP%20Res ... %20RPM.png

Have fun tuning

Andy

[dll67]

[quote="whittlebeast"

The way I verify the lambda delay is I hold the motor at light accel and then stab the throttle. I look in the logs for how long it took for the AFR to notice the sudden change in airflow and then recover.

http://www.nbs-stl.com/CRX/AE%20Testing.png

On this install the AFR will be ignored for about .15 sec.

Andy[/quote]

Saw this back at the beginning of page 3 while reading thru your techniques.
How many different spots throughout the rpm range do you gather delay info?
Do you notice the delay changing much thru the rpm range?

[whittlebeast]

Most of the stuff I tune are race cars so the critical range where throttle response has to be spot in is relatively constant. I really have not tested many motors over a wide range of flows.

Andy

[Lemon]

Hello,
Brian .H from the MegaSquirt support sent me a link to this post, so I could contact Andy, Whittlebeast, on MLV tuning with a 2015 Harley FLH.
Ok so here we go.
I got the software an tried to figure out as much as possible, also by searching the net for answers.
I do understand what the software shows although im not sure yet if I understand it right.

Now there are some really big question marks.
The settings of the application.
If I dont punch in the RPM and KPA numbers in the Histogram / Generator I can't really past that back into PV2
I found out that I can change them, to what I get out go the PV2 I hope thats right.
As soon as I press the wrong button or choose "Don't generate YX Axis" its gone, to be entered again.

Im recording almost all outputs, since im not sure which ones are the important ones.

I read, that in the Scatter Plot I should choose Map x RPM x AFF or CLI if AFF is off.

Here:
https://www.hdforums.com/forum/ignition ... ost8775845

you wrote:
"If you have MAP based VE tables then plot
x = MAP or MapPV if you imported my settings file
y = RPM
z = Front Rear AFF if adaptive fueling is turned on
or z = Front CLI if adaptive fueling is turned off
Just keep sneaking up on it with about 3% changes to VEs on every test.
Beast"


From here I dont get it how to use all this, how to generate a "correct" VE Table.
which signal must I use.

Sneaking up by 3% changes ?
it made some more than -3% difference with new VE.
And about + 50% if CLI was used to generate the table.

If I undestand right, in the Histogram / Generator view, is the place to generate a VE Table, using the "New VE" signal.

That made sense although the VE seemed a bit to low.
Using the CLI signal gave way to high VEs 100 in idle is quite a bit on the high side I`d say.

I dont know at all what is the correct way.

One other point, its not directly related to the MLV,
the CDE on the Rear of a Rushmore FLH looks rather weird to me.
How can it be that, one generation before (2012) they used identical CDE in Front and Rear.
now Front is some where close to the previous model, but the Rear is a pretty flat Z shape.
Compressed the Rear VE below 1800 RPM and is expanded above 2600 RPM.
In the 1800 - 2600 range its the same.
I suspect many heat and low RPM running problems are related to that.
In my case, roll on is terrible, if im not really careful about the throttle im getting some marble in a case noise. I suspect its the un balance of the front and rear VE.
Increasing VE manually, solved the problem some, but it is not the correct way.

I found out, that if CDE is increased, VE is increased after a follow up Autotune from about 60KPA,
each ten KPA steps down more, how much is depending on the setting. There is a formula which is actually quite close.

So im stuck at:
What are the correct steps tuning with MLV ?
How can we tune CDE correct ?
Shouldn't we get the VE table right by a AT first if CDE was changed,
since a log and generated table will just lower the VE again ?

You probably wrote it down some where, I could not find it.
Often your links end up in a website for construction.
Also that PV2 file to load in to MLV is not around anymore, the same Website shows up.
I dont know what is outdated now.

I hope its not to much at once.
Thanks in advance.
PM is fine too, not to dilute the tread.
Best regards from Europe.

Lemon

[jsimmons]

Op, if you like I can email you my tune and latest log.. if you would like to use as an example on here.. The tune is a work in progress but I'm sure there would be plenty of spots in the tune you could use as an example of something amiss.. Send me a pm with your email if your interested.. The motor is a coyote with vvt enabled.. pure race car

[jsimmons]

So i just got off the phone with andy and wanted to give him a shout out for helping me understand some of this.. he showed me that my car needs some ae adjustments..

correct me if im wrong andy but this is what i got from our call.. a lot of info was being tossed around there in a short amount of time so forgive me if i am wrong on any of this or didnt retain everything we talked about..

you can see the top set of data fields, the red line is map or engine load.. where it starts to level out near 100 kpa.. ae enrichment should be basically off at that point. 2nd colum shows my afr in the 10s.. clearly rich.. 3rd row is injector pulse width and injector pulse width base.. this is where the log shows ae enrich is staying on to long. the white line shows what injector pulse width would be without any ae enrichment.. the red line is injector pulse width with ae enrichment. the red line should jump like it is and basically level off white the top of the white line in the 3rd row of data sets.. looking at afr shows that ae is too aggressive and needs to be dialed back.

[whittlebeast]

Saw this back at the beginning of page 3 while reading thru your techniques.
How many different spots throughout the rpm range do you gather delay info?
Do you notice the delay changing much thru the rpm range?

A guy contacted me recently and a moment of brilliance stuck (mostly brilliance on his part) on how to figure out Lambda delay. This is the big reason I like doing threads like this. I forces me to think things thru the logic....

When a controls system works well, it does a good job at holding it's targets. When things get out of sync, the control starts chasing itself, constantly overshooting and then undershooting the target. At any given airflow, there should be a best compromise value for Lambda delay. If you think about it, at any given exhaust flow, it takes time to get from the exhaust valve to the o2 sensor.

go to <Fuel Settings> <AFR/EGO Control> <Use EGO Delay Table> <Use Delay Table> in the drop Down

Then go to <AFR/EGO Delay Table> becomes available. I simply never noticed this option in TS.

You should be able to have a friend play with the numbers wile watching the GEGO gauge in the background looking for a best setting for any given box. This is just theory at this point but give it a try and report back.

Andy

[dll67]

Saw this back at the beginning of page 3 while reading thru your techniques.
How many different spots throughout the rpm range do you gather delay info?
Do you notice the delay changing much thru the rpm range?

A guy contacted me recently and a moment of brilliance stuck (mostly brilliance on his part) on how to figure out Lambda delay. This is the big reason I like doing threads like this. I forces me to think things thru the logic....

When a controls system works well, it does a good job at holding it's targets. When things get out of sync, the control starts chasing itself, constantly overshooting and then undershooting the target. At any given airflow, there should be a best compromise value for Lambda delay. If you think about it, at any given exhaust flow, it takes time to get from the exhaust valve to the o2 sensor.

go to <Fuel Settings> <AFR/EGO Control> <Use EGO Delay Table> <Use Delay Table> in the drop Down

Then go to <AFR/EGO Delay Table> becomes available. I simply never noticed this option in TS.

You should be able to have a friend play with the numbers wile watching the GEGO gauge in the background looking for a best setting for any given box. This is just theory at this point but give it a try and report back.

Andy

I am hoping when I get it dialed in It will work. Once I get everything right between the delay table, authority table ego sensor response time among others. Granted not all tunes will require such in depth tuning methods but in my case the car wants as little AFR variance as possible. Right now I am having trouble getting the motor to run smooth anywhere below where the cam comes on.

[whittlebeast]

I looked at some data I had off a couple of race cars I tune that I have huge data logs from. Both cars run full time GEGO correction. My thought was to get an idea of just how much time the GEGO spends adding or pulling fuel based on the number of dots I see above or below 100 or 0 correction. I am thinking the goal of this GEGO tuning is getting as many dots in the field as possible and keeping the dots off the rails. Keep in mind that MAPxRPM is essentially air flow and as a result, is should be a fairly good estimate of exhaust flow and exhaust velocity.

Here is a VW Rabbit motor.

http://www.nbs-stl.com/tuning/MAPxRPM%20GEGO%20VW.png

and here is my CRX B18C VTec motor

http://www.nbs-stl.com/tuning/MAPxRPM%20GEGO%20CRX.png

Andy

[dll67]

Hers a couple MLVHD screen shots of my last drive.

This first one is a log view at steady cruise of the EGO oscillating.

Oscilation Log View.png

This second one is a scatter plot view of the competition between EGO 1 and EGO 2. Theres also another problem thats showing up here with the tune.

Oscilation Scatter Plot.png

That problem will show better in this final scatter plot screen shot

Scatter. Plot MxR 3 different speeds.png

3 different vehicle speeds (3 different loads) with essentially the same MAPxRPM (Same airflow). To make the car more pleasant, but not perfect to drive I might be able to get the VE table close enough for EGO Correction to handle it.
This car will never be something many people would consider "driveable" by any definition of the word, accept when its on the cam.
I can try to add resolution in the VE table around those 4 cells. The problem with tat is the unsteady MAP from camshaft overlap or I might have to go ta a blended table, adding a TPS blend.
I could feel the oscillating EGO, but couldnt make it stop. It would do it with EGO Correction set to simple and PID. Ive got some new PID settings to try next drive along with a little more refinement to the VE table.

Theres even a 3rd possibility .. I could be completely wrong and surely Andy will chime in and help me see the light.

[whittlebeast]

I figured out a new way to look at engine codes. What I did was create new custom fields in MLV. The idea is to be able the "See" any of the engine codes as they pop in and out.

Engine Running
Engine Cranking
Engine ASE
Engine WUE
Engine TpsAE
Engine TpsDE
Engine MapAE
Engine MapDE

http://www.nbs-stl.com/mlvhd/Engine%20Code.png

Here is the settings file that can be imported into any licensed version of MLV or MLV HD

http://www.nbs-stl.com/mlvhd/Engine%20Code.settings

I plan to do the same thing for the Status Codes soon. This is to make is less painful for people to understand what is going on under the covers of EFI tuning.

Andy

[dll67]

Excellent .. Now when my data log goes wonky all of a sudden I'll know what just came on or went off that may have just caused said wonkyness

[whittlebeast]

Here is a calculated I like to use when glancing at a data log in scatter plot view.

Create a new field called

Closed Loop

and define it as

[Field.egoCorrection]=100?0:1

Closed Loop areas end up with red dots, open loop will be blue.

http://www.nbs-stl.com/tuning/Closed%20 ... ueling.png

Andy