yes it's percentage
in default configuration it rejects tps < 1 so it isn't going to help you with idle (aka cell 16) (assuming it found that field, otherwise it wont add the filter for you)
stand by for next version with hugely improved analysis interpolation!
tricky part is, there's so much math now that we have to tune the tuner. there are all these scalars and multipliers to make the analysis behave correctly without adding data to the wrong cells. i think in the end it'll be awesome.
this one is more better:
http://fbodytech.com/files/trimalyzer-0.2a.zip
uses data point distance scaling to feed a weighted average system, with high pass filtering to remove noise and clipping at the top to ensure maximum strength of data points near the target.
also added a UI for the advanced tuning settings, so users can experiment with different parameters to try to figure out how to optimize this thing for me.
it has a checkbox so you can disable the interpolation and try a traditional cell-bounded analysis.
feedback please
edit: fixed that link .....
analyzed a bunch of logs and tuned it a bit more
http://fbodytech.com/files/trimalyzer-0.3a.zip
if you had used that last alpha version, once you install this, use advanced settings -> reset all
also has knock event map
the math in the above version has a flaw. it works alright... but...
will fix it later tonight.
well a week of progress is great so far,
this version fixes the math in previous versions (which was an attempt to demonstrate an idea) with a bit more consistent math that actually works.
i was overcomplicating it.
it kinda ends up making each data point relationship into a vector, the magnitude of that vector defines its weight within the average. it is maturing quite a bit and finally giving good trim maps, making this an awesome linear analyzer that may end up being good at relating tons of different kinds of data to its target tuning cell.
so basically, the closer the data point (your trim vs rpm vs map) is to the target 'observer' point (the dead center of the ve table cell), the more effect it has on that observer. it uses a proper slope calculation with noise filter highpass (meaning data from too far away of cells has zero effect) and clipping at the top (meaning stuff close by is always 100%)
we don't have to worry about a hysteresis or anything near cell boundaries so data near a boundary will affect neighboring cells, since it doesn't care about literal cell boundaries, it just tries to gather surrounding data in a way which makes sense, and each data point can affect any cell 'within its reach' with varying effect.
it also has a visualizer now, so you can see how this interpolator is actually mapping its data points to cells,
veinterp.png
i also fixed a one-off bug in the knock analyzer so the points are in the right spot... added a few other small tidbits.
enjoy
http://fbodytech.com/files/trimalyzer-0.3.1a.zip
Seem like a great idea to me! Not only for narrowband, I thing the same algorithm could be used for wideband data. Will make tuning much more precise.
This corresponds perfectly with the idea I've been developing for a while: a standalone datalogger that would plug into ALDL port like small ELM327 adapter and log to .csv file on SD card. You could collect a LOT of data just driving normally without any cables or laptops cluttering the car!
So I'll be glad to test your software as I progress with this datalogger device :)
I tested the software on Saturday and it seems like it will be great. Ill try .3.1a.when the weather breaks. Great work steveo!
i did that with linux and a raspberry pi once.Originally Posted by dzidaV8
ncurses interface, cheapo ebay backup camera as a screen (optional).
it worked as soon as you turned the igntion on, added small wifi adaptor and some scripts so it automatically uploaded logs when i parked in my driveway
http://fbodytech.com/misc/raspberry-...arddatalogger/
the software is old but you could probably screw around and make it work?
Posting Permissions
Reply With Quote
Bookmarks