Page 3 of 5 FirstFirst 12345 LastLast
Results 41 to 60 of 100

Thread: MAF Smoothing Tool for most Gen IVs

  1. #41
    Senior Tuner
    Join Date
    Nov 2006
    Posts
    2,605
    Quote Originally Posted by redhardsupra View Post
    @Alvin, don't get too carried away with high degree polynomials, at one moment you're overfitting. You gotta understand why the polynomials make a good approximation of the underlying physical model, and exactly what limits they come with.

    Ain't that what I beat you up on a while back? Extrapolating is just a guess and what to me chooses the degree is a low r squared. Almost always the higher the degree polynomial the closer r^2 is to 1.

    I think it would be the handiest in 2 bar cars where you can touch on the whole graph but want something reasonable in there.

    I was going to use some trickery with linest to do the table by rows then back out and do them by columns and average the difference in the untouchable zones. The big key on what I wanted to do is only use the data I knew was good data as opposed to smoothing everything
    Tuner at PCMofnc.com
    Email tuning!!!, Mail order, Dyno tuning, Performance Parts, Electric Fan Kits, 4l80e swap harnesses, 6l80 -> 4l80e conversion harnesses, Installs

  2. #42
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    That can run into a problem if you don't have enough points in a column or row though. The higher the order of the polynomial you try fitting, the more data points you need.

    For VE tables, I typically just do them visually based on known good points. It would be awesome if HPT could add functionality to lock cells when using the smooth tool.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  3. #43
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Quote Originally Posted by DSteck View Post
    I was going to do something for the VE table, but I don't know how to do regression for a 3D surface.
    Nevermind. Figured it out. VE Smoother based in Excel coming. Forewarning, this will contain notable error given the range it will try to approximate to a second order multi variable equation.

    Again, it will use zero as a value to enter for a cell you want ignored and will operate off the assumption that the b value (in this case z intercept) is zero (zero VE at 0rpm and 0kPa).

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  4. #44
    Senior Tuner
    Join Date
    Dec 2004
    Location
    Laurel, MD
    Posts
    1,020
    to 'smooth' (it's not a correct term, but whatever) a VE table, you need a polynomial surface. where does that show up? oh, that IS the VE table in the new computers
    so might as well use that model, set it up as a combination of constant, linear, quadratic, and interactive terms, sum them up, and then the OLS the same way you did it for the math 'smoother'. the problem with surfaces like VE is that they're way to 'bumpy' to do the full surface with one set of coefficients (one smooth surface), so you end up building a piece-wise function along the zones that GM helpfully placed there for us. but then you have problems with some of these zones not well populated, so when you do the OLS on them, they end up producing crazy surfaces that dont make sense. so then you end up adjusting zone boundries to make more sensible zone-surfaces, and you end up in an iterative hell. (guess how i know?)

    another fun application of this is to use the same method to deal with VE tables that have active intake or exhaust cams. if you did it right the first time, you dont have to change any math, just feed it a matrix of more parameters (not just map and rpm, but now int or exh cams), and you get a longer vector of coefficients.

    fun!

  5. #45
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Smoother, builder, black magic spreadsheet, it's all the same. The point is to take some data we already know and build something that may be approximately representative of the actual scenario. That was the point of the first spreadsheet I posted, and is why I set it up to allow an entry of zero to nullify that cell in the calculation.

    The difference between the Excel sheet I'll make based on Alvin's idea and GM's method is that the Excel sheet won't do zones. It'll be the entire surface approximated to one set of six variables, and not 30 sets of 6 variables. It will purely be a rough extrapolation tool because of limitations of what Excel can do and because I'm not going to exert extensive effort into something born out of lack of anything better to do.

    I can do Ax²+Bx+Cy²+Dy+Exy+F, or I can do Ax³+Bx²+Cx+Dy³+Ey²+Fy+Gx²y²+Hx²y+Ixy²+Jxy+ K, or... It is just a matter of how much typing I want to do, which isn't very much!

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  6. #46
    Senior Tuner
    Join Date
    Dec 2004
    Location
    Laurel, MD
    Posts
    1,020
    excel can do zones if every frame is marked with a number of a zone it belongs to. but yes, it's a PITA and it slows things down significantly. you really oughtta do it in Matlab, it's very easy and super fast.

  7. #47
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Not everyone has that. Almost everyone has Excel.

    I could do zones, but edge matching them would be a feat in Excel.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  8. #48
    Senior Tuner
    Join Date
    Nov 2006
    Posts
    2,605
    Its not to big of a deal with h lookup and vlookup then create a new column with Max(range, 1) max(range, n...) and repeat again with Min if you wanted.

    The biggest challenge is getting linest to ignore zeros which is while I would create a spare list and count up or down with max and min.
    Tuner at PCMofnc.com
    Email tuning!!!, Mail order, Dyno tuning, Performance Parts, Electric Fan Kits, 4l80e swap harnesses, 6l80 -> 4l80e conversion harnesses, Installs

  9. #49
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Ignoring zeros is actually super easy. Make a new table with an if statement. If the cell you reference is zero, set the rpm and MAP to zero as well. Since it would force a zero constant, these points don't add weight that would skew the curve fitting.

    I'm at work right now and can't pull up a VE table into Excel. PM me and I'll give you my email and have you send me an Excel spread sheet with axis labels of a VE table perhaps from an E40.



    Thinking about it some more, since we'd only be using a small sample set of the entire table, using zones would be a bit foolish. A) not enough data and B) we'd want to use data from other zones to fill in zones that have no data.
    Last edited by DSteck; 12-21-2011 at 02:24 PM.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  10. #50
    Advanced Tuner
    Join Date
    Jun 2008
    Location
    Charleston, SC
    Posts
    969

  11. #51
    Advanced Tuner Justinjor's Avatar
    Join Date
    Jul 2011
    Location
    Tampa, FL
    Posts
    265
    In for the VE smoothing table

  12. #52
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Well shit. When doing surface regression, if you force a point where the X and Y axis values are zero, and the data is zero, it screws up the calculation. It worked fine on the MAF iteration, so I'm not sure why it doesn't work on this. It must be a limitation of the function I'm using in Excel. Welp, this is dead in the water. Lol.



    Edit: Found a work-around. It can handle a VE value of zero, but can't handle an axis label of zero. So... instead of zero, it uses a data point with an axis label of 1^-100 and 1^-100.

    If a couple people try this out and find that it gives REASONABLE values, I'll add some more formats (like the boost section for an E40 table and LS1 1, 2, and 3 bar tables). Writing everything to do this is a bitch, so don't be shocked if I'm not thrilled to add formats.
    Last edited by DSteck; 12-21-2011 at 09:55 PM.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  13. #53
    Advanced Tuner
    Join Date
    Jun 2008
    Location
    Charleston, SC
    Posts
    969
    can you add my ve table? e38 custom os

    Labels 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
    MAP Axis 15.0 22.0 30.0 37.0 45.0 52.0 60.0 67.0 75.0 82.0 90.0 97.0 105.0 112.0 120.0 127.0 135.0 142.0 150.0 157.0 165.0 172.0 180.0 187.0 195.0 202.0 210.0 217.0 225.0 232.0 240.0 247.0 255.0
    Labels 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
    RPM Axis 500.0 700.0 900.0 1,100.0 1,300.0 1,500.0 1,700.0 1,900.0 2,100.0 2,300.0 2,500.0 2,700.0 2,900.0 3,100.0 3,300.0 3,500.0 3,700.0 3,900.0 4,100.0 4,300.0 4,500.0 4,700.0 4,900.0 5,100.0 5,300.0 5,500.0 5,700.0 5,900.0 6,100.0 6,300.0 6,500.0 6,700.0 6,900.0
    -Carl
    Last edited by carlrx7; 12-21-2011 at 08:51 PM.

  14. #54
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Quote Originally Posted by carlrx7 View Post
    can you add my ve table? e38 custom os
    Post the file, turd. Or better yet, you could set up EQ_VE to do that for you. It's easy to edit the definition file to add a new format.

    Man, a 33x33 table is going to be freaking brutal to set up.
    Last edited by DSteck; 12-21-2011 at 09:59 PM.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  15. #55
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    I hate you Carl.



    The new file has updated support for the E38 custom OS with the 33x33 VE table. The Excel spreadsheet allows you to modify the axis labels as well, and it will automatically adjust and recalculate based on your axis labels. Note that due to the size of this table, the error will be pretty significant since only one set of coefficients is trying to fit to the entire table. If you want to focus on just one section, zero out the whole table first, then copy and paste just the cells you want to go over. It'll calculate out and then just copy out that same section.

    Here's the kick in the pants... the new Excel spreadsheet is half a megabyte, which exceeds what the forum will let me attach. As a result, when you download this, you need to change the file extension to .xls before using it.
    Last edited by DSteck; 12-23-2011 at 09:27 AM.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  16. #56
    Tuner 07MontRedcp's Avatar
    Join Date
    Jan 2007
    Location
    Arizona
    Posts
    187
    Quote Originally Posted by DSteck View Post
    I hate you Carl.



    The new file has updated support for the E38 custom OS with the 33x33 VE table. The Excel spreadsheet allows you to modify the axis labels as well, and it will automatically adjust and recalculate based on your axis labels. Note that due to the size of this table, the error will be pretty significant since only one set of coefficients is trying to fit to the entire table. If you want to focus on just one section, zero out the whole table first, then copy and paste just the cells you want to go over. It'll calculate out and then just copy out that same section.

    Here's the kick in the pants... the new Excel spreadsheet is half a megabyte, which exceeds what the forum will let me attach. As a result, when you download this, you need to change the file extension to .xls before using it.
    But you said you were bored....

    BJK
    APS TT 630 rwhp/654 rwtq @ 9 psi
    '07 C6, TR6060 conversion w/Z06 ratios. DTE Stage 4 Quaife Diff w/brace (3.73). Z06 Brakes, mild turbo cam & dual springs (222/222, LSA 115, .581/.581), LS9R clutch, LG GT2 Coilovers, Autometer Cobalt Fuel gauage, Progressive Alky Controller, and e-Boost2 Controller. PFADT poly bushing for control arms and Sports sway bars. MGW shifter. 3" B&B Fusion. RSI Stage 1.5 fuel system. RX dual valve catch can.

  17. #57
    Senior Tuner IDRIVEAG8GT's Avatar
    Join Date
    Aug 2010
    Location
    Roswell, N.M.
    Posts
    1,956
    Awe......







    Fuck.

    I'm lost.
    Gray Ghost- The abomination. 2007 Chevrolet Silverado CCSB. 98mm turbo, nitrous, 428LSX, Rossler 80E with a brake. Finally finished. 23 psi, no numbers, Slow as hell.

    PBM G8- Aluminum 364, twin Precision 67/66 turbos, 6L90 trans swap, CTS-V/Vaporworx fuel system, slowly making progress.

    Dads 2011 CTS-V- Stock bottom end, stock heads, LS9 cam, pullies, ported blower, ported TB, D3 goodies, and lots of nitrous.
    618/618 motor
    906/862 spray

    Caterpillar 50 Forklift- Duramax swap

  18. #58
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Quote Originally Posted by 07MontRedcp View Post
    But you said you were bored....

    BJK
    I had finished it once but somehow failed to save it, so I had to redo it. Each time took 45 minutes. No bueno.

    DSX Tuning - Authorized HP Tuners Dealer
    http://www.dsxtuning.com
    http://www.facebook.com/dsx.tuning
    Just say no to bull s***.
    IF YOU WANT HELP, POST A FILE!

  19. #59
    Advanced Tuner
    Join Date
    Jul 2008
    Location
    Louisville, KY
    Posts
    417
    Quote Originally Posted by IDRIVEAG8GT View Post
    Awe......







    Fuck.

    I'm lost.
    I'm with ya so you got company.

  20. #60
    Advanced Tuner
    Join Date
    Jun 2008
    Location
    Charleston, SC
    Posts
    969
    Quote Originally Posted by DSteck View Post
    I had finished it once but somehow failed to save it, so I had to redo it. Each time took 45 minutes. No bueno.




    thanks!