VEMS Logger

[hilly]

I have finally started to drive my car about and have been recorded a couple of logs.

When veiwing the logs in VEMS logger I have notice that it does not display the current ve bin value used........
Everything else is displayed, just not the ve values which I find very useful when analising the logs.
I guess I could work it out from the other parameters, but there are a few dispalys I would gladly loose ( such as fuel pressure, EBP etc) to have it.

Is there any way to make VEMS logger display the current ve map value

Hilly

[rob@vems.co.uk]

You are correct, it doesnt VE value as its calculating a load of factors from it.  Normally we read the points of interest on the Lambda feedback and correlate the RPM and KPa values to work out which load point its reading from, then you adjust the values in that area.
VEMS interpolates across a number of cells so its often the case that you need to adjust 2 or 4 values.

[hilly]

Thanks for the speedy reply Rob

It sounds like it can't display it then.
Odd as it chucks a load of other stuff on screen and there is a handy space next to Lambda and target Lambda.

I am fine with working it out, it would just be nice to see what value it based the ego correction on.

Hmmmmmmmmmmmm, perhaps I'll have to generate a conversion xls to allow me to put VEMS logs through MSLVV as that does display ve, plus that has better controls to skip through the log.

Hilly

[antonch]

Does MSTweak3000 work with VEMS? It basically takes your logs of observed AFRs and makes adjustments to your VE table to set it to stoich.

[dnb]

VEMS has a built in VE learn mode which does much the same thing as MSTweak3000.

It has the advantage that you don't need a computer connected to it while it's learning, and it all works on the fly.  Have a look at the VE learn options in MT.  Setting it up can be a bit tricky, but is worthwhile as it's quite an interesting to see.

I have a couple of excel macros for adjusting the VE map and such like for when I don't want to use VE learn (there's no way to have VE learning working on a different area of map to the "closed loop" area - possible mod suggestion here Rob!)  I keep meaning to get them coded up in Perl so I can run them on batches of logs from some of the long trips I've done recently.

MT can display the currently calculated VE as far as I remember, although I tend to not use it, prefering instead to calculate average GEgo from a RPM/LOAD cell (from my excel macro) in steady state (this is very important) and apply this as a correction to the current VE map.

[hilly]

I have a had a quick play with VE learn, but I couldn't get it to do anything, probably because the settings were wrong.
The ego control is working fine as I can see that, it just doesn't adjust the ve map.

I'll just have to do it the old way of looking at the values as I drive and fiddling with the maps.

Hilly

[multiplex]

can anyone post workable numbers for VE Learn?  I've tried it with little success.

I've been using megalogviewer to recalculate my maps, but it is a little time consuming switching files back and forth all the time.

[dnb]

OK, I'll write something about how to use VE learn over the weekend. 

A couple of points worth noting: 

It makes a difference map that is stored in the ECU and applied to the main map at run time.  It is not written across to the "main" VE map until a command is issued to the ECU (via terminal or PS2 keyboard)

There are confidence values in the params.  If these are too high, then the ECU is never sure it is allowed to change the tables, so it leaves them alone.  Setting these is a balancing act between stability and speed of learning.

[antonch]

OK, I'll write something about how to use VE learn over the weekend. 

A couple of points worth noting: 

It makes a difference map that is stored in the ECU and applied to the main map at run time.  It is not written across to the "main" VE map until a command is issued to the ECU (via terminal or PS2 keyboard)

There are confidence values in the params.  If these are too high, then the ECU is never sure it is allowed to change the tables, so it leaves them alone.  Setting these is a balancing act between stability and speed of learning.

Any updates? Hopefully I will have some time to put my new injectors in this weekend.

[dnb]

Here's version 1.  I need to dig a few more things up

VE learning happens when the car is in closed loop mode and the following conditions (specific to VE learn) are met:

Min coolant temperature:  Fairly obvious.  This is the temperature when you want VE learn to work.  Best set it a degree or so under stat temperature so that the engine is thoroughly warm otherwise you'll see odd results creeping through.

Max power:  I have absolutely no idea what this does.  IIRC I set it to 100 and ignore it!

There are scaling factors applied to the inputs to the VE learn algorithm.  Inputs are RPM, Pressure (kPa) and EGO.

I'm not 100% sure what these do.  I think they are specifiers of how accurate the measurements are for RPM, kPa and EGO.  I put 100 in each of them, which seems to work.  (I will do a bit more digging into these) 

Inside the algorithm, other parameters are used for specifying confidence thresholds, how quickly values can change and what the maximum change can be:

Min weight for modifying VE entry:  The algorithm calculates a new VE estimate and a "confidence" of how good it thinks the new estimate is.  This parameter allows you to filter out the "bad" estimates (eg from acceleration).  Too low and you'll let any old rubbish modify the table.  Too high and nothing will ever change.   

Speed of changes:  As far as I know, this controls how quickly the VE values are changed.  Too fast and the algorithm will never settle on an answer.  Too slow and you'll be waiting until the oil runs out... 

Max deviation from reference table:  Fairly staight forward - the largest correction to the VE table that the algorithm can do.  I usually use 10% to 20% on the grounds that the algorithm isn't perfect, and may go crazy (with +ve feedback effects) if there are misfires etc.

Be careful where you let the algorithm work.  If you use decel fuel cut etc, then it's best to disable EGO correction in these areas where practical otherwise the learning algorithm can try to "drill holes" in the map as the fuelling is turned back on.  As always, it's best to do some tests yourself and see what it does. 

Remember to back up your config before you start, and that you need to use the command line interface/PS2 keyboard to merge the VE map and the set of generated changes.

[antonch]

How can I save the values to the ecu with the keyboard? My VEMS unit did not come with a slot for one. Can I send command through mini-term in Megasquirt?

PS. Some more (yet confusing) information here:
http://megasquirtavr.sourceforge.net/manual/Detailed.Control.VE.html#Operate

[rob@vems.co.uk]

The keyboard commands are given here:
http://www.vems.hu/files/refcard.pdf

The terminal program that we use is Bray Terminal which is in the Tools directory of your firmware& Megatune package

[antonch]

The keyboard commands are given here:
http://www.vems.hu/files/refcard.pdf

What command would I use from Bray?

VE autotune (mv)
reset learning (discards all changes) r
copy learned ve table to reference ve table s