Smoothing your tune with graphing
#48
while we're real technical i had another thought...
is the resolution shown in the 3d view the same as what's really in the ECU? If so then the ECU is going to be interpolating the values when it's between points anyway. Either that or it'll just use the value at the closes point. I'm sure smooth maps are happier than rough ones but as far as the ECU is concerned it doesn't matter wether it's interpolating between 2 similar values or 2 very different values. A flop is a flop if all that changes are the values of the numbers. Obviously changing from say 20% of max value to 80% of max value from one cell to the next would be terrible for the engine but the ECU itself doesn't care at all.
It's just as difficult to find the mid point between 2.2 and 2.3 as it is to find it between 2.2 and 88.8.
I can't wait to get a cable and start playing with the software. With the AEM you are so insulated from all this stuff. Just push + or - to adjust a point on the graph.
is the resolution shown in the 3d view the same as what's really in the ECU? If so then the ECU is going to be interpolating the values when it's between points anyway. Either that or it'll just use the value at the closes point. I'm sure smooth maps are happier than rough ones but as far as the ECU is concerned it doesn't matter wether it's interpolating between 2 similar values or 2 very different values. A flop is a flop if all that changes are the values of the numbers. Obviously changing from say 20% of max value to 80% of max value from one cell to the next would be terrible for the engine but the ECU itself doesn't care at all.
It's just as difficult to find the mid point between 2.2 and 2.3 as it is to find it between 2.2 and 88.8.
I can't wait to get a cable and start playing with the software. With the AEM you are so insulated from all this stuff. Just push + or - to adjust a point on the graph.
Based on the above example it does matter as the interpolated values are different. The interpolated value is in between the two cells not always in the middle if I am not mistaken.
YMMV
#49
Evolving Member
iTrader: (1)
Join Date: Jun 2005
Location: Gulfport MS
Posts: 213
Likes: 0
Received 0 Likes
on
0 Posts
Oh absolutely, the interpolated values are different! Sorry if i gave the wrong impression. The mechanical systems don't like big jumps and the jumps will be bigger with a map full of jagged cells. The ECU itself doesn't have to "try harder" to do the calculations was all i was trying to point out. It can interpolate big jumps just as easily as small ones. You might loose some percision but i doubt it. Your engine trying to deal with a 60% increase in injector pulse between cells would be another matter
I suppose it's splitting hairs but that's what i like to do, computers are cool
I suppose it's splitting hairs but that's what i like to do, computers are cool
Last edited by Rob_GPT; Aug 12, 2008 at 02:18 PM.
#50
Evolved Member
iTrader: (3)
I have been thinking about upgrading to 1.35, just for this feature, but concerned w/ 2 things...
1. W/ my limited comp. knowledge, I was hoping it would be an easy task, but was assuming it wasnt as easy as Jack_of_trades makes it out to be...
Could you possibly start a "how to" on this when you have some free time, for us, not so knowledgable people?
2. Im using 1.29 w/ most of the .xml "patches", minus the Tephra stuff. What Im concerned about is: Are all of the new patches in there? ex. (3D MUT table, and the new boost control tables. I would assume yes, but what do I know... And are there any other probs w/ this version, besides the freezing?
I think this is an awesome feature! Just need to know how to use it...
TIA
#51
Evolved Member
iTrader: (3)
My noob will be showing but...
How is the best way to even start the smoothing process? Is it better to smooth the low points "up"? Or smooth the high points "down"? Or to only smooth the areas between the high and low points? What is more important vertical smoothness? Or Horizontal smoothness?
Which is more critical in avoiding knock, smoothing the fuel or smoothing the timing? I would assume timing...
How is the best way to even start the smoothing process? Is it better to smooth the low points "up"? Or smooth the high points "down"? Or to only smooth the areas between the high and low points? What is more important vertical smoothness? Or Horizontal smoothness?
Which is more critical in avoiding knock, smoothing the fuel or smoothing the timing? I would assume timing...
#52
Evolving Member
but like I said, smoke ...
Last edited by Jumperalex; Aug 12, 2008 at 02:44 PM.
#54
Evolved Member
iTrader: (90)
Join Date: Jul 2005
Location: Roselle, IL
Posts: 1,917
Likes: 0
Received 0 Likes
on
0 Posts
Man, I just did some driving on this map and am reporting back via a PC. Off boost driveability seemed smooth, but WOT felt a LOT more smoother. I had a few occasions of sparatic knock >3cts which I will have to take a look at the log and fix, but other than that, smoothing makes a BIG difference.
#55
Wow, lots of people posted in here today,lol. The maps are in raw bit values and just use a formula to show a more "user friendly" value in the cells. Timing is in increments of 1 so its the easiest to tweak since you basically are seeing it the same way the ECU does. The Fuel maps are adjustable from 0 to 128 if memory serves me. Refining the increments just gives you more adjustment points so you can dial it in a little better.
I threw some new spark plugs in my car today and just took it for a ride. Wow, I'm in love with my car all over again! Thing pulls like a BEAST compared to before. I am now able to run 25* of timing at 7K RPM @22psi where I couldn't climb towards 20* before without knock coming in to join the party. It totally works and I love it. If you wonder why I can run so much timing, I'm running an Aquamist HFS-5 kit with 100% Denatured Alcohol.
I threw some new spark plugs in my car today and just took it for a ride. Wow, I'm in love with my car all over again! Thing pulls like a BEAST compared to before. I am now able to run 25* of timing at 7K RPM @22psi where I couldn't climb towards 20* before without knock coming in to join the party. It totally works and I love it. If you wonder why I can run so much timing, I'm running an Aquamist HFS-5 kit with 100% Denatured Alcohol.
#60
Evolving Member
Auto-smoothing
Ok I worked out a very basic autosmoothing version. Nothing more than a center weighted average of the cells surrounding the original cell. For non-edge cells that means average(x) = (x+x+x+a+b+c+d+e+f+g+h)/11). Except on the edge cells which are just the straight average of the original cell and its surrounding 5 cells.
A bunch of cells I left completely alone because the load/rpm is low and the resulting values just didn't aid in smoothing out the curve but seemed to give lower timing than I really would want anyway. They are shaded gray.
Obviously this is just a way to get a good start before you hand tweak. But boy does it take out the grunt work.
That was for timing. For fuel I did the same basic thing but instead of doing a direct link on some cells I just tested for it being 14.7 and kept it at 14.7otherwise it was center-weight averaged. Again the edge cells just averaged.
On both timing and fuel I did some conditional formatting to key in on changes both in the main tables as well as the "difference" tables.
Finally I've got them both set up so that you can compare stock, to modded, to smoothed.
I'm not really done playing around increasing the "logic" of the conditional smoothing but I can tweak this thing forever without ever being "done" and figured it is better just to release versions.
http://jumptronix.com/evo/EVO%20Map%...osmoothing.xls
EDIT: oh yeah and on the timing I rounded to an integer. I wanted to use the ceiling function to always increase timing but it doesn't seem to work with negative numbers and I got too lazy to set up a conditional or create my own version of a ceiling function. Another option I'm considering is limiting all changes to +/- 1 and then letting the user iterate till you get it at smooth as you want with as few changes as possible. Or allow for a settable threashold value below which there is no change. and and ... yeah it is 12:45am, Law and Order SVU is on (with Tank from The Matrix as the bad guy), and I have 10 different ideas running through my head. hehe and I could easily make those threshold values unique for each RPM and/or load but that just seems like silly over engineering.
A bunch of cells I left completely alone because the load/rpm is low and the resulting values just didn't aid in smoothing out the curve but seemed to give lower timing than I really would want anyway. They are shaded gray.
Obviously this is just a way to get a good start before you hand tweak. But boy does it take out the grunt work.
That was for timing. For fuel I did the same basic thing but instead of doing a direct link on some cells I just tested for it being 14.7 and kept it at 14.7otherwise it was center-weight averaged. Again the edge cells just averaged.
On both timing and fuel I did some conditional formatting to key in on changes both in the main tables as well as the "difference" tables.
Finally I've got them both set up so that you can compare stock, to modded, to smoothed.
I'm not really done playing around increasing the "logic" of the conditional smoothing but I can tweak this thing forever without ever being "done" and figured it is better just to release versions.
http://jumptronix.com/evo/EVO%20Map%...osmoothing.xls
EDIT: oh yeah and on the timing I rounded to an integer. I wanted to use the ceiling function to always increase timing but it doesn't seem to work with negative numbers and I got too lazy to set up a conditional or create my own version of a ceiling function. Another option I'm considering is limiting all changes to +/- 1 and then letting the user iterate till you get it at smooth as you want with as few changes as possible. Or allow for a settable threashold value below which there is no change. and and ... yeah it is 12:45am, Law and Order SVU is on (with Tank from The Matrix as the bad guy), and I have 10 different ideas running through my head. hehe and I could easily make those threshold values unique for each RPM and/or load but that just seems like silly over engineering.
Last edited by Jumperalex; Aug 13, 2008 at 09:49 PM.