Originally Posted by wstar

Update - the old top post was wiped out in favor of new, better data and a better map. If you saw/used the map that was originally here, that map is what's called "Linear" below.

This post is all about the Throttle map in UpRev. You'll need the Tuner license (or take this data to a tuner) to use this. Keep in mind this is all from an NA vehicle with a fairly maximal set of bolt-ons and a dyno tune. YMMV on a bone stock car, and I especially wouldn't blindly follow this on FI.

After a bunch of testing (and posting, and thinking, and testing more, etc) I've finally settled on a throttle map I really like. I've done some drivability testing on this map, and more importantly I've done datalogging of several reasonable candidate maps with Cipher to add some numeric credibility to my subjective opinions.

Background:

The stock throttle map uses lower values at the upper end of the pedal range when at lower RPMs. This doesn't let you get on full power in the lower part of the range, instead opening up a bit gradually as the RPMs rise. This is sometimes perceived as throttle "lag" at low RPM. UpRev tuners have figured out that you can "fix" this by putting a solid line of slightly-over-maximum values (3800) in the uppermost column of the data, and then trying to smooth that into the existing factory data a bit for drivability. The idea is to give you a similar-looking throttle curve in the useful parts at all RPMs, which reaches the maximum.

However, "simple" or "naive" attempts to just set a few top columns fixed at 3800 and step off of that in linear amounts (which is what my tuner did. I don't fault him for that, it took hours of mathing around with curve fitting algorithms, plotting software, and testing to do better), you get a jerky throttle response map, and you can feel it places when driving.

My first attempt to smooth this out (while retaining the benefits) was a linear map from low-end stock values out to a couple columns of 3800, which was better, but still not ideal. I found that a simple 2nd order polynomial curve going through 3 target data points (stock 0% values, a fixed value at 50%, and 3800 at 100%) seemed to give better results, given some tuning on the middle fixed value. A reasonably magic value for the 50% mark seems to be 2300. The polynomials were derived using the perl module Algorithm::CurveFit from CPAN.

The Data:

In all of the charts below, the maps in use are:

Linear: This was my first smoothing attempt, the one that was posted here at the top originally. It's a straight shot from around 18.8% up to a couple columns of 3800's at the end.
Curve 2700: This is the good polynomial curve, but with the middle value set to 2700, as an example of what happens when it's a bit out of range.
Curve 2300: This is the good curve, with the middle value at 2300. You'll notice it tends to approximate the stock curves in Cipher data graphs, which is a really good sign.
Stock: This is the bone stock throttle map
Tuner: This is the simplistic "jam some 3800s into the top of the chart and roughly smooth down from there" approach that came from my tuner. I didn't datalog this for the data charts since I threw it out of the running long ago, but it's in the first chart just for comparison sake.

These charts were generated from the raw data using Plot.

This first graph is of the raw table data itself, just the line of values for the 4800 RPM row, to give you an idea of the shape and smoothness of each of the candidate maps. Note that axis labels are what they're called in the UpRev Rom Editor, but even UpRev acknowledges that we don't really know what these numbers really mean. "Flow Potential" seems to roughly correspond with the accelerator pedal position, and "Driving Force" has something to do with engine throttling via the butterflies and/or VVEL.

Note that "Curve 2300" makes a good smooth curve approximation of the Tuner values, and the "Curve 2700" is the same shape but roughly bumped up to enclose the Tuner values.



The next two charts are based on Cipher data logging of the first 4 candidate maps above. The log snapshot was taken with me rolling through the accelerator pedal from a low position to full throttle at a reasonable approximation of "linearly" in 3rd gear from around 3000 RPM to redline. Since I knew I wouldn't be able to reliably reproduce the same pedal curve for each run, the time-axis was thrown out and the data has been re-mapped strictly versus the accelerator position itself, to provide clean comparisons.

The first datalogging chart is Accel Pedal vs Throttle Position Sensor. The important takeaways here are that (1) the Linear map actually has a big non-linear bump in the real world versus the stock curve, (2) the 2700 curve does crazy things up top and doesn't even sustain full throttle, and (3) the 2300 curve does a good job approximating the stock response curve.



The other is Accel Pedal vs "VVEL Position Sensor". Don't ask me what that means, but it's still interesting to observe diffs in the shapes. Note again the 2300 curve is the closest to stock, the 2700 curve is kinda screwy at the bottom, and the linear curve is a bit off at the top in the opposite direction. I take the stronger divergences away from the stock shape to mean the engine is telling me "you're doing it wrong"



So I've settled on the 2300 Curve above. It seems to provide the best blend of (a) getting those 3800 values across all RPMs at the top of the pedal, so you're not limited by "throttle lag" at lower RPMs, (b) having a smooth graph shape with no sharp, unpredictable transitions, and (c) having smooth results that closely mirror stock in the important related Cipher datalogs during a smooth pull through 3rd gear.

The 2300 Curve Data:

You can copy this straight from here and paste it into a throttle table in UpRev:

The results should look like this in the editor when displaying decimal values:



And this is the UpRev view of the whole map surface for it:



If you're concerned about compatibility between my stock ECU setup in general and yours, this is what my stock throttle map looks like for comparison: