VE Tune by statistics, calculating Lambda delay

[se7ensport]

Hi All

My fuel map is really getting there, but being an alpha-N set up it's taking a while.

I'm using the VE Tune by statistics to correct the fuel map, but I'm unsure what lambda delay to use. 

1. one figure for the whole table based on lambda location (i.e. mine is 29inch from head, I'm using 1 3/4 headers on a 2.0 4 cylinder engine)
2. RPM dependent figure as higher RPM the higher the speed of the exhaust gas resulting in a lower delay figure
3. view logs and calculate from there... if so what should i look at as I though big changes in tps then see how soon lambda follows but this seems to vary as well depending on rpm but its not very consistent.

any guidance appreciated.

Alex

[z0tya]

I think the bigger problem to filter the trasient: acceleration deceleration. I examined more log, I have found the lambda follow the pw with around 4-500ms

[mattias]

Even if the "VE tune by statistics" is a workable method, I find that manually tuning it both live and from datalogs is quicker. The statistics method works real well for fine-tuning after a long drive, with many visits to the same load site. If you analyse the data from a complete race you will come to understand it, check for high variance and low hit counts and ignore those recommendations, those are probably transients from acc/dec.

[GintsK]

Does anyone can get similar results like with MLV?
I had no luck until now. MLV alows 4 different coarse-fine tuning settings. And to get good tables takes just few attempts. Even if initial table is totally wrong shape. Two 5...10 minutes driving with "normal" and one with "hard" and you are within 2% of target - it is usual.
But it seems VEMSTUNE analyser gives spiky table even on close to optimal table.
Any tricks here?

[GintsK]

Even if the "VE tune by statistics" is a workable method, I find that manually tuning it both live and from datalogs is quicker. The statistics method works real well for fine-tuning after a long drive, with many visits to the same load site. If you analyse the data from a complete race you will come to understand it, check for high variance and low hit counts and ignore those recommendations, those are probably transients from acc/dec.

These words means that VEMSTUNE tool is still far from optimal. I can't imagine any circumstances how tuning by hand could be faster than MLV. Especially if we talk about whole data area not few cells.
On dyno I usually switch road simulation and table is ready within a minutes. Then you have a lot of time to search for optimal timings, work on boost settings and do other hand adjustments - there is a lot of them.
I am very disappointed with dropped MLV support until VEMSTUNE allows do datalog review, live view and analysis in same handy and fast way

[Jamo]

I am very disappointed with dropped MLV support until VEMSTUNE allows do datalog review, live view and analysis in same handy and fast way

support isn't dropped I still use it, it's a bit long winded but I convert my vemslogs to csv and export an msg and do analysis in megalogviewer.

Once I have the new VE values I use the import button on the VE table select all files and select the msq.

It imports the new values

[mattias]

GintsK: Don't read between the lines, this is how *I* work and perhaps "quicker" was the wrong wording but to me it's just that - quicker.

I have to start out by saying that  I _always_ datalog.  I also tried the MLV method (it's still working afaik) but only found it takes too long to go from a blank VE table to a working one, I'd rather spend the time  live tuning instead and get a feel for the engine and with a warmed up N/A engine, once on the road this is done in a matter of minutes. I don't know, maybe that sounds incredibly fast to some, but with experience and a safe ignition map to start out with it has become the quickest method for me. And the way the VE tune 3D view and all the gauges (lambda history graph) and other things work while live tuning with VemsTune, it's just gotten better.
Once I feel I've gotten as close as I can with manual intervention I resort to datalogging and analysis. I see what you mean by VemsTune not being optimal,  you should get your point across (in a structured manner)  of how you would like it to work,  and it just might improve and change the way me and others use the program.
If you're operating a dyno you can, and are bound to use, other methods than you do while tuning on the road, so that is a big factor here. I only have a few engine dynos nearby, nearest rolling road is 3-4 hours away so when I do get there the car is usually tuned on the street to near perfection anyway, maybe some ignition timing left to find.

As with many other things, there are as many methods of tuning as there are tuners out there, with different experiences making their way of working towards a final tune unique.

[se7ensport]

Mattias, how do you calculate lambda delay when using tune by statistic?

[gunni]

You can review your datalogs and see how it behaves.

i.e for instance when you let go of the throttle and the pw is off, how longer does it take for the lambda value to react to no fuel at all.
this will vary with engine speed and sensor location.

I found it to be 2-3 log instances when I use mlv .

However don´t waste your time chasing lambda values to hard. There is little to no real world difference in lambda being 0.8 or 0.82 or even 0.78 , or lambda 0.99 and lambda 1.02 and so on,

I also just manually live tune and then review logs if a customer wants me to verify his engine is still running right and I´m far away.

[GintsK]

I am very disappointed with dropped MLV support until VEMSTUNE allows do datalog review, live view and analysis in same handy and fast way

support isn't dropped I still use it, it's a bit long winded but I convert my vemslogs to csv and export an msg and do analysis in megalogviewer.

Once I have the new VE values I use the import button on the VE table select all files and select the msq.

It imports the new values

Where can be the mess: last point - import from msq do not work for me (at least on previous VT 2010-09-08 version)? In worst case it just smash table. In best - do nothing. I tried both: drag and drop and Import button in VE table.
What VT works for that?

[GintsK]

Mattias, you are right. It is hard to adopt to VT It is just different. It has a lot of very handy features. I can't live anymore w/o PgUp, PgDwn funcionality, importing. Lambda target gauge, history graph. But datalog window simply kills me! Blow it! It just strike trough everything what can be done with just a single mouse or keyboard touch in MLV!

I agree there is many different kinds of tuning strategies. My strategy is hurt strongly.

Lambda delay depends from too much factors. At idling lag can be even seconds. At WOT/highRpms - almost always just small fraction of second. It would be good if analyzing software takes this into account.
Usually it is well readable from datalog.

[peter_jensen]

This is a new way to make a rough tuning and a lot of people like´s it 
http://www.youtube.com/watch?v=qwkqjotkDJQ

[GintsK]

This is a new way to make a rough tuning and a lot of people like´s it 
http://www.youtube.com/watch?v=qwkqjotkDJQ

I always said offline analyzing method is better. Better because it does not create uneven spikes, hills and valleys on map. You start with even table. And as result Ego correction stays even with no spiky jumps. Then analyzer just do their job. Less chance of over-adjusted cells. Less chance of bad jerking when you go over adjusted -> non-adjusted cells.
Holleys software looks doing good, but this video did not show 3D graph for purpose. On 3D it will not look so promising.
BTW TunerStudioMS Registered Edition also perform quite good autotuning.

[Pet]

I have to agree with GintsK; when my very experienced tuner saw VemsTune logging/tuning cappability, he wanted to close that SW and don't use it anymore. His simply way: MT or TunerStudio to create log + then MLV offline to analyse logs, then back export .VEX tables; it's more then enough for very fast, simple and powerfull tuning.
 Because logging is used very often instead changing basic settings, it means it's very important how is it user friendly....
I think at least .VEX import to VS should be enough for all old school tuners.
Nice looking GUI and gauges etc. are fine, but imho tuning tools are on the 1st place....

Note: sorry for my english....

[gunni]

VT is alot better then megatune or tunerstudio for live tuning.

a whole lot better then alot of tuning programs even.
The log review side just needs more work.