Karter16's Silbergrau E46 M3 Journal

[karter16]

This is a good comparison of the difference.

This is the resultant VE table error of just mapping the correction to the standard interpolation points:



And this is exactly the same data, but with interpolation points optimised - I have also specified certain areas of the surface as having a greater weighting (e.g. low dr_rel/rpm combos the relative error matters more).

[Bry5on]

That is too cool. And on a Mac no less! Guess I’m glad I sprung for the beefcake M4 in my latest machine.

[ac427]

Well we seem to finally be coming out of the months of winter sickness in our house. I've been very patchy the last couple of months in terms of car stuff - hopefully I haven't missed replying to anyone!

Few quick updates: I'm likely taking the car on a trip at the weekend. In preparation for this I thought I'd better revert to the standard community patch v2 software (I was running a customised version for testing something) and took the opportunity to take the car for a decent drive and do a long VE tuning run to capture some more "spirited" driving. Very much enjoyed the grip of the wider tires up front, and the rigidity in the rear from the RACP brace makes a really big difference. I'm looking forward to getting Heinz's front brace installed as it's definitely noticeable that the front isn't quite as focused now in comparison to the, improved, rear.

Doing the VE tuning reminded me to have a play with something I've been meaning to do for a while. With high rate CAN logging there's enough resolution in the logs to go more granular with the VE tuning process. I'm just playing around at this point, but have tried things like building a correction map at a higher resolution, and then essentially brute-force solving (my M1 Max seems to be able to do about 150,000 fits per minute) an optimal 20x24 table.

In this quick example below, the best-found table has an RMS error of 0.0046 vs 0.0109 on the starting corrected table (e.g. standard interpolation points) - a 57.95% improvement, and a reduction in the max error point of about 20% - not bad.

Lots more playing round with this to be done.

Great stuff. Is the higher resolution logging enabled in the community patch?

[karter16]

That is too cool. And on a Mac no less! Guess I’m glad I sprung for the beefcake M4 in my latest machine.

I must say it's nice to finally have a reason to need such a powerful processor haha. I don't think I've really stretched its legs all that much!

[karter16]

Great stuff. Is the higher resolution logging enabled in the community patch?

It's not released yet - the code is done, but I'm still finishing documenting all of the variables (there's thousands). I'm actually making some progress on that so hoping to have it out there in the not too distant future.