Smooth Throttle for UpRev
 

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.

http://www.the370z.com/members/wstar...0-rpm-line.png

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.

http://www.the370z.com/members/wstar...rottle-pos.png

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" :)

http://www.the370z.com/members/wstar...s-vvel-pos.png

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:

http://www.the370z.com/members/wstar...-data-view.jpg

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

http://www.the370z.com/members/wstar...rface-plot.jpg

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:

http://www.the370z.com/members/wstar...rottle-map.png

I should add a request too: does anyone out there understand what the "Driving Force" vs "Flow Potential" axes really mean?

I don't at the moment but I'm getting my uprev tune on sunday and I'll ask the tuner for ya

I'll check this out tomorrow.

This is nice! I have the cable that came with the GTM turbo kit, but no software :-/

BTW, I created this thread below and wonder if you can chime in. I miss the quicker response of the DE/HR. I wonder if it's possible to compare the tables.

http://www.the370z.com/engine-drivet...hr-vs-vhr.html

Nice work! I'm taking my car in to uprev to get my maps fixed and retuned soon, so I might ask Jared to drop this table into one of my maps and test it out.

Awesome, thanks -- I will pass this on :tup:

"Driving Force" appears to = Load and "Flow Potential" is either VE or MAF counts (or a calculated perecentage based on both).

Load is based on RPM, VE, and g/sec of metered air, so that's probably about right.

Yeah the table still doesn't make intuitive sense to me though. The tuner I used (as in, the dyno-tuning professional) seemed to know what he was doing and set all "3800" values in the last few columns though, and that seems to correspond to WOT at full pedal.

It's clear from experimenting that this table does affect how the throttle reacts to pedal input, but both of the non-obvious labels seem like secondary effects. The position of the throttle butterfly will determine airflow, and that combined with RPM is pretty much going to define load or driving force or whatever. It just seems like one of these axes must be mislabeled for this to be a throttle table.

Very cool geeky stuff! luvit

Interesting stuff. Doesn't the VVEL vary the valve lift and duration to control power and the throttle butterfly stays wide open nearly all the time? If so, then the power varies more by VVEL rather than butterfly position.

I've heard that myth often, but I'm pretty sure that's not the case. VVEL very likely does vary the valve lift over different operating conditions, but that's no real substitute for actual throttle plates (which is why you still need ITBs even on the most advanced engines).

The easy way to tell is to log the throttle position over OBD-II, which is what I do on Torque while playing with this stuff. The basic Throttle Position PID comes back as angle data, and it ranges from around 1.0 to 89.6 in my testing so far. Generally speaking, if anything the throttle position has a tendency to stay more closed than you'd expect on stock-ish maps, and only opens up fully when you're at full throttle combined with high RPM.

Later today I plan to test some "extreme" maps that will get to the bottom of the mysteries of this table I think. I'm thinking one that ramps to 3800 after the first 4-5 columns or so, and one that never gets higher than about 1500 even at the end. Results should be "interesting", and will confirm whether the table is really just a pedal -> butterfly vs RPM map, or there's something more complicated going on.

Ok, so I did my crazy extreme map testing. I need to sit down and think through the results, but they should be informative.

Based on my smooth map from the top post, crazy map #1 was altered so that all of the columns from 37.5% flow onwards (to 100) were all pegged to the value "3800". The result from a driving perspective is that when I reached somewhere around 1/3 gas pedal position, the car would suddenly jump like I nailed the pedal to the floor, the OBD-II throttle position values would jump out to high values (wide open), but the whole jump would be very brief. Basically if I held onto that pedal position, the car would pulse on the throttle between 1/3 and full. It was like it was trying to take off at full throttle, but something else in the ECU held it back a short moment later.

Crazy map #2 was the opposite. From the 37.5% column onwards, all values were pegged to the number "1310". The reaction was somewhat similar. At around 1/3 throttle, the car would suddenly jump in power like I tried to nail the gas hard. However, the OBD-II throttle position sensor would *not* jump to wide open when this happened, just the engine power surged, and power got retarded back pretty quickly like the previous test, although smoother.

I imagine all the throttle jumping around and acting retarded in both scenarios is because there are other tables involved (e.g. VVEL) and the ECU in the overall is either confused by or just flat refuses to follow the contradictory commands with crazy values. But it does seem to confirm that raising the numbers in the table leads the ECU to at least try to open the throttle wider.

I need to think more on all of this to make sense of it though.

It's because VE and MAF counts should go up with higher TPS and RPM -- but, the load tables must also activate different cels in the VVEL maps, which is where all the actual throttling takes place. Finally, there are target AF maps that also correpsond to those load values.

Basically, no matter how you scale the tables, actual physical VE will not reach maximum levels unless RPM is approaching redline and throttle (as controlled by VVEL) is wide open.

If the values in the table don't corerespond well to physical reality you will most likely get herky jerky engine responsiveness as it tries to maintain whatever the target AF ratio is for the corresponding load cell.

Can you adjust timing or fuel on tip-in? That's a great way to improve throttle response!

Also -- do the values in the map correspond to injector duty cycle? I'm guessing that they do, although exactly what 3800 means isn't clear to me... uprev doesn't use intuitive cell values for a lot of the maps...

You know how the VHR ramps up pretty slow under 3k RPM compared to DE/HR? Does this mean it still has the capability to ramp up pretty quick? I just want to verify the lag or slowness is really in the software than physical or mechanical.