Welcome to the North American Subaru Impreza Owners Club Tuesday September 30, 2014
Home Forums WikiNASIOC Products Store Modifications Upgrade Garage
NASIOC
Here you can view your subscribed threads, work with private messages and edit your profile and preferences Home Registration is free! Visit the NASIOC Store NASIOC Rules Search Find other members Frequently Asked Questions Calendar Archive NASIOC Upgrade Garage Logout
Go Back   NASIOC > NASIOC Technical > Engine Management & Tuning > Open Source Reflashes

Welcome to NASIOC - The world's largest online community for Subaru enthusiasts!
Welcome to the NASIOC.com Subaru forum.

You are currently viewing our forum as a guest, which gives you limited access to view most discussions and access our other features. By joining our community, free of charge, you will have access to post topics, communicate privately with other members (PM), respond to polls, upload content and access many other special features. Registration is free, fast and simple, so please join our community today!

If you have any problems with the registration process or your account login, please contact us.
* Registered users of the site do not see these ads.
Reply
 
Thread Tools Display Modes
Old 04-20-2010, 09:47 PM   #1
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default MAF Scaling - Smoothness Theory (Some Experiments)

So I've been wondering how "smooth" a maf scaling has to be. So I did some homework and made a tool that "best" fits some high order polynomial equation to the maf scaling. Taking into account suggested corrections for romraider's maf tool and airboy's spreadsheet.

So here's my original scaling that kept all my fuel trims <4%

Maf Scaling:



first discrete derivative to show smoothness or slopes between cells




Not so good. Looks like it could be smoother. So I used the smoothing tool and got:



and the smoothness looks like:





Loaded the map and tried it out. Took the CL part for a spin and found out I have 18% - 7% corrections all over the place. So then I applied the RR suggested corrections to my maf scaling and I'm back at:


for smoothness.







So, in conclusion, it looks like the maf curve doesn't have to be "super smooth" From what I can tell, as long as there are no jagged spikes or huge dips in the scaling, it's all good.

-BN
* Registered users of the site do not see these ads.
Bad Noodle is offline   Reply With Quote
Old 04-20-2010, 09:49 PM   #2
SubieFiesta
Scooby Newbie
 
Member#: 149426
Join Date: May 2007
Vehicle:
2002 Subaru WRX
and 2005 Subaru STi

Default

nice job
SubieFiesta is offline   Reply With Quote
Old 04-21-2010, 12:18 AM   #3
NSFW
Scooby Guru
 
Member#: 140444
Join Date: Feb 2007
Chapter/Region: NWIC
Location: Near Seattle, WA
Vehicle:
05 Stage Free LGT
ATP 3076, 6MT, AVO FMIC

Default

I still think smoother would be better. Try setting up a scatter plot in Excel, and smooth it manually, that's what I do. I've been accused of being obsessive though and I can't argue with that.
NSFW is offline   Reply With Quote
Old 04-21-2010, 02:34 AM   #4
arghx7
Scooby Specialist
 
Member#: 232940
Join Date: Dec 2009
Location: cold
Default

As a control, why don't you use this method on the stock calibration for several different models? Like 02 WRX, 04 STi, 05 LGT just download the stock roms and pull the data from there.
arghx7 is offline   Reply With Quote
Old 04-21-2010, 06:57 AM   #5
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Quote:
Originally Posted by NSFW View Post
I still think smoother would be better. Try setting up a scatter plot in Excel, and smooth it manually, that's what I do. I've been accused of being obsessive though and I can't argue with that.
I've been obsessive about it too, but am starting to realize that it might not be beneficial. I smooth manually, run the auto smoothing algorithms, then manual smooth again.

I ran the last smoothness scaling I posted this morning and I'm back within 4%.

I agree that smoother would be better, in theory, but practice isn't coming to the same conclusion. I guess the lumpiness of the first derivative is just a way of saying that the MAF sensor isn't a perfect sensor and the deviations from the perfectly smooth graph are it's shortcomings. I'm sure everyone's would look different on the fine scale but show the same patterns on a large scale.

Quote:
Originally Posted by arghx7 View Post
As a control, why don't you use this method on the stock calibration for several different models? Like 02 WRX, 04 STi, 05 LGT just download the stock roms and pull the data from there.
I've checked them out and they're somewhat smooth. They have dips in places where the voltage step isn't consistent. I've also checked out other scaling people have posted and some from protunes I've gotten. Same result, they're not all super smooth. The first derivative is usually "lumpy" to a limit
Bad Noodle is offline   Reply With Quote
Old 04-21-2010, 09:27 PM   #6
fastblueufo
Scooby Specialist
 
Member#: 212154
Join Date: May 2009
Chapter/Region: South East
Location: Ed@fastperformancetuning.com
Vehicle:
02 billet crank
pauter rods, pt5857, E85

Default

I found the same thing you have. I try to keep some form of a smoothness to the curve but most of the time the maf scale wont be a perfect curve.
fastblueufo is offline   Reply With Quote
Old 04-21-2010, 09:43 PM   #7
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

I worked out my maf curve to <4% again. Car runs smooth, good power, no knock but the curve looks like:



and the smoothness looks like:



it ain't pretty, but it works
Bad Noodle is offline   Reply With Quote
Old 04-22-2010, 12:17 AM   #8
NSFW
Scooby Guru
 
Member#: 140444
Join Date: Feb 2007
Chapter/Region: NWIC
Location: Near Seattle, WA
Vehicle:
05 Stage Free LGT
ATP 3076, 6MT, AVO FMIC

Default

I just realized that when you say "smoothness" you mean the derivative, not the curve itself. I take back what I said. I thought those wiggly lines were your MAF scaling.

Yeah, I wouldn't lose sleep over the wiggly lines in the derivative of the scaling. I basically just want the "actual AFR / desired AFR" scatter plot to be symmetrical and not too wide.
NSFW is offline   Reply With Quote
Old 04-22-2010, 10:30 AM   #9
the suicidal eggroll
Scooby Guru
 
Member#: 51961
Join Date: Jan 2004
Chapter/Region: RMIC
Location: Broomfield, CO
Vehicle:
2005 STi
2012 WRX

Default

Mine is definitely not perfectly smooth. I don't think it's a problem with the sensor, I think it's turbulence in the MAF pipe that creates a non-linear relationship between total mass air flow and air velocity at the sensor. The MAF sensor is essentially reading air velocity, in a perfect world this would be directly proportional to the mass air flow rate in the pipe at all flow rates, however in the real world it's HEAVILY dependent on the air flow pattern through the pipe, and how that changes with varying air speeds. At low speeds you might have a high velocity node right at the sensor, so the sensor reads higher than what a smooth curve would dictate and your MAF scaling has to be reduced at that voltage. At higher speeds you might have a high velocity node in another part of the pipe, and a low velocity node right at the sensor, so the sensor reads a little lower than what a smooth curve would dictate and your MAF scaling has to be increased at that voltage.
the suicidal eggroll is offline   Reply With Quote
Old 04-22-2010, 10:42 AM   #10
arghx7
Scooby Specialist
 
Member#: 232940
Join Date: Dec 2009
Location: cold
Default

You have to remember that the factory scaling is done in expensive simulation software using proprietary data. If it's not perfectly smooth then it doesn't surprise me that a decent custom scaling isn't perfectly smooth either.

Also, are you using an Excel macro to calculate these derivatives?
arghx7 is offline   Reply With Quote
Old 04-22-2010, 10:49 AM   #11
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Quote:
Originally Posted by arghx7 View Post
You have to remember that the factory scaling is done in expensive simulation software using proprietary data. If it's not perfectly smooth then it doesn't surprise me that a decent custom scaling isn't perfectly smooth either.

Also, are you using an Excel macro to calculate these derivatives?
na, the derivatives are slopes between adjacent cells.
Bad Noodle is offline   Reply With Quote
Old 04-22-2010, 01:18 PM   #12
NSFW
Scooby Guru
 
Member#: 140444
Join Date: Feb 2007
Chapter/Region: NWIC
Location: Near Seattle, WA
Vehicle:
05 Stage Free LGT
ATP 3076, 6MT, AVO FMIC

Default

Quote:
Originally Posted by arghx7 View Post
You have to remember that the factory scaling is done in expensive simulation software using proprietary data.
The more factory tunes I see, the more I'm convinced that they don't use any better tools than we do.

Maybe here there, to check some aspect of their tune... But in general I suspect they were done by people who get paid to churn out as many "good enough" tunes as possible, as cheaply as possible. Not by people who get paid to produce perfect tunes using amazing tools, and are given enough to time to do it right.

There's a little bit of hyperbole in what I just wrote, but not much.
NSFW is offline   Reply With Quote
Old 04-22-2010, 01:30 PM   #13
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Is each tune custom to the car or do they just make one tune and slap it on all factory vehicles?
Bad Noodle is offline   Reply With Quote
Old 04-22-2010, 01:35 PM   #14
the suicidal eggroll
Scooby Guru
 
Member#: 51961
Join Date: Jan 2004
Chapter/Region: RMIC
Location: Broomfield, CO
Vehicle:
2005 STi
2012 WRX

Default

Quote:
Originally Posted by Bad Noodle View Post
Is each tune custom to the car or do they just make one tune and slap it on all factory vehicles?
One tune for all vehicles of a given year/model. Sometimes they'll revise the tune mid-year, but that's about it.
the suicidal eggroll is offline   Reply With Quote
Old 04-27-2010, 10:09 PM   #15
shutupnshift04
Scooby Specialist
 
Member#: 101710
Join Date: Nov 2005
Chapter/Region: Tri-State
Location: Jersey. not shore.
Vehicle:
BOSS

Default

Shouldn't you get the maf scaled then smoothed and then change the other tables ie timing and afr to get the car to run its best?
shutupnshift04 is offline   Reply With Quote
Old 04-28-2010, 06:35 AM   #16
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Quote:
Originally Posted by shutupnshift04 View Post
Shouldn't you get the maf scaled then smoothed and then change the other tables ie timing and afr to get the car to run its best?
yeah, but that's a whole different thing.

This thread is about defining or finding the meaning of "smooth" in terms of the the allowable changes between cells on your maf table.
Bad Noodle is offline   Reply With Quote
Old 04-28-2010, 06:40 PM   #17
Phatron
Scooby Guru
 
Member#: 36033
Join Date: Apr 2003
Chapter/Region: SCIC
Location: Tuning Lab
Vehicle:
CEO PhatBottiTuning
2006 STi GTX3582 + Meth

Default

i still dont really understand what the hell im looking at.....
Phatron is online now   Reply With Quote
Old 04-28-2010, 07:20 PM   #18
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Quote:
Originally Posted by Phatron View Post
i still dont really understand what the hell im looking at.....
You're looking at the maf scaling as a graph, then the corresponding slopes between cells. Slopes between cells can be used to describe "smoothness". When a function or graphed line is "smooth", the changes between two values are predictable. Therefore you can judge how smooth something is by taking the derivatives of the equation. In this case, there is no equation for the maf curve. So you look at the slopes between cells to determine if the maf scaling is changing in a gradual and predictable way. If it is, then it's "Smooth"

The more discrete derivatives you take of the data, the more you can tell how smooth a data set really is.

In this case, my maf scaling got bumpy after the first derivative showing that my scaling isn't smooth. I fitted polynomial equation to the maf scaling to make sure it is smooth, then tried it out. It didn't work, reverted to original unsmooth scaling, and everything was ok again. Conclusion is that maf scaling doesn't have to be super smooth, so it's probably best to just use the romraider maf tool suggested value.
Bad Noodle is offline   Reply With Quote
Old 04-28-2010, 07:23 PM   #19
flatthump
Scooby Specialist
 
Member#: 102749
Join Date: Dec 2005
Chapter/Region: SCIC
Location: Fullerton, Ca
Vehicle:
2005 impreza STi
OBP'

Default

I'd much rather have the AF corrections linear or flat than i would want the MAF scaling smooth.
flatthump is offline   Reply With Quote
Old 04-28-2010, 07:34 PM   #20
flatthump
Scooby Specialist
 
Member#: 102749
Join Date: Dec 2005
Chapter/Region: SCIC
Location: Fullerton, Ca
Vehicle:
2005 impreza STi
OBP'

Default

Quote:
Originally Posted by arghx7 View Post
Also, are you using an Excel macro to calculate these derivatives?
Use a Delta command to find corresponding deltas between cells then make another command using those Deltas to calculate the slope between cells. Then just graph it.
flatthump is offline   Reply With Quote
Old 04-28-2010, 07:46 PM   #21
Bad Noodle
Scooby Specialist
 
Member#: 174442
Join Date: Mar 2008
Chapter/Region: Tri-State
Location: NNJ
Vehicle:
2007 SuperWagon
I don't spell well

Default

Quote:
Originally Posted by flatthump View Post
Use a Delta command to find corresponding deltas between cells then make another command using those Deltas to calculate the slope between cells. Then just graph it.
you gotta use the slope formula because the voltages are not all evenly spaced.

So just use

slope = (gs[x+1] - gs[x])/(v[x+1] - v[x])

and I agree about the zeroed learning view over smooth scaling.


It always bothered me that cobb's tuning guide says the maf scaling should be smooth, but doesn't specify how to judge that.
Bad Noodle is offline   Reply With Quote
Old 04-28-2010, 08:11 PM   #22
flatthump
Scooby Specialist
 
Member#: 102749
Join Date: Dec 2005
Chapter/Region: SCIC
Location: Fullerton, Ca
Vehicle:
2005 impreza STi
OBP'

Default

Quote:
Originally Posted by Bad Noodle View Post
you gotta use the slope formula because the voltages are not all evenly spaced.

So just use

slope = (gs[x+1] - gs[x])/(v[x+1] - v[x])

and I agree about the zeroed learning view over smooth scaling.


It always bothered me that cobb's tuning guide says the maf scaling should be smooth, but doesn't specify how to judge that.
That would work. But I just evenly space them. In 2007 Excel I can do it easily. I'll try your way next time.
flatthump is offline   Reply With Quote
Old 04-28-2010, 08:18 PM   #23
Phatron
Scooby Guru
 
Member#: 36033
Join Date: Apr 2003
Chapter/Region: SCIC
Location: Tuning Lab
Vehicle:
CEO PhatBottiTuning
2006 STi GTX3582 + Meth

Default

Quote:
Originally Posted by Bad Noodle View Post
It always bothered me that cobb's tuning guide says the maf scaling should be smooth, but doesn't specify how to judge that.
you judge it visually......you'd be surprised at what some peoples maf scales look like....

i use the MAF tab to collect data....but i input the data into ecuflash while im looking at the graph to make them smooth....

Heres 5 MAF scales.....i did 4 of them.....1 of them is "not smooth"...its pretty apparent which one.

Phatron is online now   Reply With Quote
Old 04-29-2010, 01:00 AM   #24
arghx7
Scooby Specialist
 
Member#: 232940
Join Date: Dec 2009
Location: cold
Default

Quote:
Originally Posted by NSFW View Post
The more factory tunes I see, the more I'm convinced that they don't use any better tools than we do.

Maybe here there, to check some aspect of their tune... But in general I suspect they were done by people who get paid to churn out as many "good enough" tunes as possible, as cheaply as possible. Not by people who get paid to produce perfect tunes using amazing tools, and are given enough to time to do it right.

There's a little bit of hyperbole in what I just wrote, but not much.
Have you ever heard of tools like Matlab Simulink? Are you familiar with hardware-in-the-loop simulations?
arghx7 is offline   Reply With Quote
Old 04-29-2010, 01:45 AM   #25
NSFW
Scooby Guru
 
Member#: 140444
Join Date: Feb 2007
Chapter/Region: NWIC
Location: Near Seattle, WA
Vehicle:
05 Stage Free LGT
ATP 3076, 6MT, AVO FMIC

Default

No hands-on experience, but sure I'm passingly familiar with the theory.

Why do you ask? Do you have reason to believe that Subaru's (Denso's?) tuners use them?

I have a hard time believing it makes economic sense to invest that much in something that can be done sufficiently well. with trial-and-error. Or maybe that's the problem... I mean, a hastily brewed simulation would go a long ways toward explaining the wacky timing tables I've seen in stock 2002 WRX tunes.
NSFW is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Anyone have the maf scaling values for an Perrin Big MAF. Stock WRX injectors NITROS Open Source Reflashes 1 02-05-2008 06:57 PM
probably elementary questions - maf scaling, afr targets, UTEC wg control gpatmac UTEC 36 04-05-2006 05:11 PM
UTEC MAF Scaling? Ski Man Engine Management & Tuning 1 10-10-2005 08:48 AM
a/f learning and maf scaling deluxsti Engine Management & Tuning 2 06-06-2005 04:22 PM
what is MAF scaling? erggy Engine Management & Tuning 0 11-29-2004 11:36 AM


All times are GMT -4. The time now is 02:43 PM.


Powered by vBulletin® Version 3.7.0
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Powered by Searchlight © 2014 Axivo Inc.
Copyright ©1999 - 2014, North American Subaru Impreza Owners Club, Inc.