Issue with Trim Error Tool

zkemp82

New member
Weird stuff keeps breaking on my car, so every time I think I have it dialed in, I find something wrong and fix it, having to start all over again.

I'm working off of HolyCrapItsFast's guide and the trimerror tool. Since I'm back to square one, I'm trying to log the latencies again. This time, after using the tool, i get a graph that has a nearly perfect parabola. I.e., it starts out at 0, then dips negative to around -4/-5, jumps up to just over 10, then back down to -5 before returning back to 0.

The line is in no way, shape or form flat, but by reading the instructions, I'm supposed to have a nice evenish slope. This is not nice, even, or even a single slope.

Anyone out there who can help?
 
Weird stuff keeps breaking on my car, so every time I think I have it dialed in, I find something wrong and fix it, having to start all over again.

I'm working off of HolyCrapItsFast's guide and the trimerror tool. Since I'm back to square one, I'm trying to log the latencies again. This time, after using the tool, i get a graph that has a nearly perfect parabola. I.e., it starts out at 0, then dips negative to around -4/-5, jumps up to just over 10, then back down to -5 before returning back to 0.

The line is in no way, shape or form flat, but by reading the instructions, I'm supposed to have a nice evenish slope. This is not nice, even, or even a single slope.

Anyone out there who can help?


Don't waste your time logging latencies.

1. The basics
a. How much data are you using (how many lines of data in your log)?
b. Is the car FULLY warmed up when you start logging?
c. I assume you're using access tuner?
d. Did you zero out load compensation tables and cylinder compensations (not required but I find it easier)?
e. Did you make any adjustments to inj latency or scaling?

2. Can you post a screenshot of your graph?

3. If you insist on using that tool then sometimes you can make a copy of the graph and eliminate the trailing 0's and then add a trend-line or 2, I usually add 1 for linear and 1 for polynomial. It's usually not anything close to perfect, don't get stuck with anything it is 'supposed' to be like in your head. I've attached a couple screen shots.
a. In trendlines since so much of it is flat except the beginning I might just make some adjustments to those high positive values in the beginning and then log again and fine tune then smooth out my MAF Calibration table and then adjust my load compensation table.
b. In trendlines2 I tried to make it like you're describing again depends on several factors but I might rough adjust for the extremes and then log again and see where I'm at or...maybe increase latency by 2.5% and then lower scaling by 5% or something of the sort, depends on any other adjustments made, mods, etc.

4. Think about making some of your own spreadsheets once you get the hang of it and start noticing any repetitive actions you perform. See my Speed Density Volumetric Efficiency Automatic Smoother (VBA) and my MAF Voltage Scaling Smoother.

5. Switch to this tool, much better and can chew through 250,000 lines of logging data within very specific parameters in seconds. See screenshot
https://github.com/vimsh/mafscaling

6. Remember shit data in = shit data out. Log LOTS of GOOD DATA. Pedal fluctuations to a minimum, go very slow and very smooth. You want cell values after they stabilize not the initial change so in general more data the better.
 

Attachments

  • trendlines.png
    trendlines.png
    18.3 KB · Views: 316
  • trendlines2.png
    trendlines2.png
    19.4 KB · Views: 281
  • auto_ve_smooth.jpg
    auto_ve_smooth.jpg
    28.4 KB · Views: 344
  • maf smoothing.jpg
    maf smoothing.jpg
    31.6 KB · Views: 273
  • maf scaling.jpg
    maf scaling.jpg
    32.4 KB · Views: 340

zkemp82

New member
Don't waste your time logging latencies.

1. The basics
a. How much data are you using (how many lines of data in your log)? - This last one was 48,000, the max you could use on the tirmerror tool
b. Is the car FULLY warmed up when you start logging? - yes
c. I assume you're using access tuner? - yes
d. Did you zero out load compensation tables and cylinder compensations (not required but I find it easier)? - I usually do, but I did forget to zero out the load comp tables. Cylinder comps have been zeroed since I started
e. Did you make any adjustments to inj latency or scaling? - Not during the log, and I set them back to factory calibration specs before I ran the logs.

2. Can you post a screenshot of your graph? I don't have permissions to attach files yet.


3. If you insist on using that tool then sometimes you can make a copy of the graph and eliminate the trailing 0's and then add a trend-line or 2, I usually add 1 for linear and 1 for polynomial. It's usually not anything close to perfect, don't get stuck with anything it is 'supposed' to be like in your head. I've attached a couple screen shots.
a. In trendlines since so much of it is flat except the beginning I might just make some adjustments to those high positive values in the beginning and then log again and fine tune then smooth out my MAF Calibration table and then adjust my load compensation table.
b. In trendlines2 I tried to make it like you're describing again depends on several factors but I might rough adjust for the extremes and then log again and see where I'm at or...maybe increase latency by 2.5% and then lower scaling by 5% or something of the sort, depends on any other adjustments made, mods, etc.

4. Think about making some of your own spreadsheets once you get the hang of it and start noticing any repetitive actions you perform. See my Speed Density Volumetric Efficiency Automatic Smoother (VBA) and my MAF Voltage Scaling Smoother.

5. Switch to this tool, much better and can chew through 250,000 lines of logging data within very specific parameters in seconds. See screenshot
https://github.com/vimsh/mafscaling

6. Remember shit data in = shit data out. Log LOTS of GOOD DATA. Pedal fluctuations to a minimum, go very slow and very smooth. You want cell values after they stabilize not the initial change so in general more data the better.


My logs were taken during mostly highway driving, and on low speed limit surface streets with minimal stoppage from lights, traffic, etc. I've thought about adding a throttle delta monitor as well to try to flush out any weird changes as well, in addition to eliminating anything in open loop. I'll try the other tool you posted too. I've been learning for about the past 2 years, but didn't have any major changes until recently to test out what I've learned yet. Thanks again for your help.
 
Last edited:
My logs were taken during mostly highway driving, and on low speed limit surface streets with minimal stoppage from lights, traffic, etc. I've thought about adding a throttle delta monitor as well to try to flush out any weird changes as well, in addition to eliminating anything in open loop. I'll try the other tool you posted too. I've been learning for about the past 2 years, but didn't have any major changes until recently to test out what I've learned yet. Thanks again for your help.

If you're tweaking your tune you don't have to zero out load comps. If you're following HCIF guide it is stated you only need to rescale MAF if you did something to alter flow across the MAF sensor I believe with examples otherwise you should only need to tweak load compensations depending on the mods.

Use that tool I specified, it makes tweaking load compensations soooo easy and MAF adjustment, though smoothing it accurately takes a little bit to do in excel.

Also log minimal things for highest resolution, when LOGGING on the AP in the car look at the bottom of the AP and it will say Hz it is logging at (samples per second). I like to have 18 bare minimum but 20+ is preferred depends whats going on and what needs to be logged. You can get some things yourself with excel, see my attached screenshot and you would just copy those formulas down (when you type them it, it won't give formula it will show an actual value I just did that so you can see the formula).

If you have trouble PM me and you can email me screenshot and or log or something.
 

Attachments

  • excel example.jpg
    excel example.jpg
    28.6 KB · Views: 334
  • excel example2.jpg
    excel example2.jpg
    30.5 KB · Views: 317
  • excel example3.png
    excel example3.png
    5.8 KB · Views: 269
Top