Boost/AFR smoothing

Want the ECU+ to make you breakfast in the morning? Did the software light your basement on fire? Chime in here.
Post Reply

If you had a pie to give to Tom... would you?

Poll ended at Sat Nov 10, 2007 3:26 pm

yes
3
75%
no
0
No votes
Are you kidding? I would give my first born son!
1
25%
 
Total votes: 4

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Boost/AFR smoothing

Post by honki24 »

Could we get like a right-click ability to smooth the boost/AFR graphs? Something like how DSMlink does it? I think it is a moving average over very small time increments that smooths it out. Logworks has it too. Very very handy for conistant readings. It doesn't need to log with smoothing on, but could we get an option to smooth the log file later? One generic function could probably be used to smooth all of the graphs and we could just right click and choose the smoothing for whichever graph we need. I think this would be most helpful for me because I'm getting tired of looking at 5% changes between data points because sample rate is so fast. (not that fast is a bad thing). Thanks!
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

tlcoll1
Site Admin
Posts: 1512
Joined: Sun Jun 13, 2004 1:01 am
Location: Odenton, MD
Contact:

Post by tlcoll1 »

Describe *exactly* how it would work. Things to think about:

1. Would the amount of smoothing be configurable in the ECU+ Win settings?
2. Would the same smoothing be applied to all windows, or would it be different for different windows?
3. Would ECU+ Win remember whether smoothing was on between opening and closing the application? Between opening and closing a log?
4. What would the right-click menu look like?
5. How would it work on the all-in-one graph?

See how the other apps you mentioned do it, and come up with a "just code it up" kind of specification. This sounds like something I could do if given enough detail that I don't have to figure out the nuances of it.

Tom

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

tlcoll1 wrote:Describe *exactly* how it would work. Things to think about:

1. Would the amount of smoothing be configurable in the ECU+ Win settings?
2. Would the same smoothing be applied to all windows, or would it be different for different windows?
3. Would ECU+ Win remember whether smoothing was on between opening and closing the application? Between opening and closing a log?
4. What would the right-click menu look like?
5. How would it work on the all-in-one graph?

See how the other apps you mentioned do it, and come up with a "just code it up" kind of specification. This sounds like something I could do if given enough detail that I don't have to figure out the nuances of it.

Tom
1: configurable? how about like 4-6 settings. I'm not sure the increments but something like logworks would be nice. (I've attached pictures of the logworks method in sequential order)
2: different smoothing for each window (or graph if you like the all-in-one)
3: I don't think ECU+ needs to remember smoothing. It should be an easy enough process to resmooth if necessary.
4: Maybe the right click menu could look like the ones from logworks that I've attaced. Then again if you want to make it simple it could be like an option in the right click window like "Smoothing >" and the subwindow would have something like ".25sec, .5sec, ...etc"
5: Not sure how it would exactly work in the all in one because I never use that. I suppose if you can currently click on individual curves then it could work the same as logworks. If you cant click on the curves... maybe you can click on the names on the right? Or perhaps just right click in the background and the the menu would go somethign like "smoothing >
  • > [smoothing values (.25sec/.5sec/etc)]
Attachments
pictures.zip
(304.54 KiB) Downloaded 439 times
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

rogersmith
Posts: 84
Joined: Fri Nov 24, 2006 5:28 pm
Contact:

Post by rogersmith »

You wouldn't lose accuracy with smoothing?

tlcoll1
Site Admin
Posts: 1512
Joined: Sun Jun 13, 2004 1:01 am
Location: Odenton, MD
Contact:

Post by tlcoll1 »

In an engineering sense, no, but in a practical sense yes. That is, the smoothing would be the mathematically correct average, but you wouldn't be able to see the detail you see now. I guess it depends on the signal as to what kind of broad brush you'd like to use to paint the signal.

Tom

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

plus high level detailed accuracy isn't always helpful. At some point it becomes noise since you can't do anything about it. In the case of boost and AFR for me, this is the case. These small fluctuations can't be corrected and are a result of the logging/sensor refresh speeds, so there is no need to see them. They make it difficult to see the "big picture" of what is going on. I suggested an "after" modification, if you will, because some people might like the pinpoint accuracy that the ECU+ is feeding back, but personally I'm not tuning individual cylinders nor am I tuning to partial degrees of timing or exact thenths of a psi of boost. I just want to scroll the mouse across the graph and get the numbers I'm looking for, not have to make a best-fit line in my head. Plus, if you smooth it afterward, you're not really degrading the integrity of the data, you're just puting an adjustable filter on it.
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

rogersmith
Posts: 84
Joined: Fri Nov 24, 2006 5:28 pm
Contact:

Post by rogersmith »

This seems like it would be a nice feature. I'm for it.
I wonder sometimes if my AFR curve could get any flatter but maybe it can't.

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

rogersmith wrote:This seems like it would be a nice feature. I'm for it.
I wonder sometimes if my AFR curve could get any flatter but maybe it can't.
Yay! How does it go? "Where two or more are gathered in ECU+'s name, there Tom is."? lol
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

rogersmith
Posts: 84
Joined: Fri Nov 24, 2006 5:28 pm
Contact:

Post by rogersmith »

Honki, have you ever thought maybe those fluctuations in AFR are real fluctuations? Like the ECU is bouncing back and forth between two load columns? If that's true then it could just as well bounce back and forth between two AFR values.
I'm going to see if it's true once I get my MUT logging working.

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

well.. no I don't think that's the case since the ECU interpolates between cells and my cells are all smoothed in every direction so I know it shouldn't be "jumping" at all. I'm relatively sure that this is just a normal case of over-accuracy. In any good experiment you can become so detailed in your measurement that you drown out the real importance in the data. With sample rates so fast I can see why we're getting a "noisy" feedback from the WB. I think a normal smoothing factor should make the trend much more apparent and make tuning a bit easier.
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

tlcoll1 wrote:In an engineering sense, no, but in a practical sense yes. That is, the smoothing would be the mathematically correct average, but you wouldn't be able to see the detail you see now. I guess it depends on the signal as to what kind of broad brush you'd like to use to paint the signal.

Tom
How's it goin? need any more input? :wink:
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

malibujack
Posts: 15
Joined: Sun Feb 18, 2007 2:25 pm
Contact:

Post by malibujack »

This reminds me of the old question..

Whats the difference between accuracy and precision..

Sometimes too much precision just clouds your data, but accuracy is always important.. Not to mention there are plenty of occasions where the values are precise, but not accurate (space shuttle o-rings for example)

I agree, there should be some user configurable way to normalize the data a little, but if your getting jumps that are large enough for you to see it as noise and not useful info, then you should zoom out of the graph a bit, as that has a similar effect.

If your seeing jumps between samples that are on the order of 0-100 (from mut values that is) then there may be a problem with dropped data, and then you have to send those logs to Tom so he can see if its the case.

malibujack
Posts: 15
Joined: Sun Feb 18, 2007 2:25 pm
Contact:

Post by malibujack »

OOOOHHHH I get it. In your case the problem is the boost control solenoid, I used to see this fluctuation on my software because the sample rate was faster than the duty cycle of the solenoid, and the map sensor was picking this up..

honki24
Posts: 525
Joined: Thu Nov 17, 2005 9:59 am

Post by honki24 »

lol, you're on a page of your own sometimes! I'm not even talking about MUT values, just the voltage inputs from boost and wideband. I agree, accuracy is key... that's what calibration tables are for, but precision is open to determination based upon the usage.
Advoacate of the Theocratic Reign.
http://www.FlatlineMotorsports.net
03 Evo 8

malibujack
Posts: 15
Joined: Sun Feb 18, 2007 2:25 pm
Contact:

Post by malibujack »

Yeah I realized what you were looking at after I read it the second time, but felt it was better to just post a new one than edit/delete my older post.

Post Reply

Return to “Feature Requests and Bug Reports”