Hondata

Well, my little 'proof of concept' application for analysis of kmanager datalogs to aid in tuning for power (cam angles) has had some pretty fruitful results so far. I started off with a pretty choppy and moderatly useful system based on tune points and decided to go to a more 'trend curve' based system using some smoothing and interpolation routines across multiple runs.

I'm not so sure that ignition can really be done using this since the power differences would be so small that I imagine a dyno would almost be required to tune ignition. My methodology has been to leave the ignition where it is on the base maps and retard for knock when encountered and adjust a/f and then pick cams for power. Afterall, it's the air flow that people change the most from the base maps.

I started off with a pretty sound permis:
http://forums.clubrsx.com/showthread.php?t=292428

And later after using the datalogs of a few people who actually have tuned a few cars and had real dyno results to compare against datalogs they gave me AND considering the real world results I got with my first few attempts to make a composite cam angle map from acceleration results, the process was refined and the following is the most up to date result set. It came out better than I had hoped:
http://forums.clubrsx.com/showthread.ph ... ost6320287

I would really like to hear something from hondata on this in regards to my theory, practice, and actual results. Any advice or caveats you know of would be great.

Sorry to interupt here, but I commend you for youre research enthusiasm. I am currently an undergrad student in Gainesville hoping to achieve a degree in mechanical engineering from UF. I enjoyed reading your thread on Club rsx as well as those who gave their input. Keep at it, I'm very interested to see the result of this.


Thnx,
Evan

Update: Results of bracketing, before and after calibrations.
http://forums.clubrsx.com/showthread.ph ... ost6374415

Update: Documentation now done, system is ready for some testing / use by anyone that wants to give it a whirl.

http://forums.clubrsx.com/showthread.ph ... ost6714001

Razathorn wrote:Update: Documentation now done, system is ready for some testing / use by anyone that wants to give it a whirl.

http://forums.clubrsx.com/showthread.ph ... ost6714001

Razathorn, great work!! I really really like this kind of stuff! Inspired by your work I wrote a piece of software to do the exact same thing to be used off-line and just to get a better understanding of it all. Anyway, I was out last night to get the composite map done for the lowcam, and got nice results.

The following graph is the result of my code, post-processed in Excel with some 6:th order polynomial trendlines, they actually fit quite nicely:




And this is the result of the same datalogs ran through your online software:




Man, this is just great! I really like the whole concept of tuning my car on the street, and this is just the tool I've been waiting for. Keep up the good work!

How many runs did you use per line? I needed to average out a few runs to get rid of road noise, but I was looking for some number magic to remove that need since 3 runs is a lot for just one change .

I did the suggested 3 runs per cam angle, so 3*6 runs in total for the lowcam. There is still quite a lot of "noise" in the runs like you said, but a friend of mine is currently looking at into more sophisticated ways to do the smoothing.