Page 3 of 3 FirstFirst 123
Results 41 to 50 of 50

Thread: Ford Speed Density Calculator

  1. #41
    Advanced Tuner CCS86's Avatar
    Join Date
    Nov 2017
    Location
    Austin
    Posts
    913
    Quote Originally Posted by Puggyberra View Post
    There may be a more 'correct' way to do this, but my method is weighting the correction factor the same as the mapped point. For instance if your load correction is lets say 0.2 and you are using mapped points 3 and 4, with 3 being 40% and 4 being 60% i would multiply that .2 * .4 and .2 * .6, then apply the weighted correction into each mapped point. Generally I do a flat cut off somewhere around 20% where I stop weighting (ie: if its 82% and 18% I'd do the entire correction to the mapped point at 82%).

    Theoretically this method has inaccuracy in it, after all your entire error could be coming from mapped point 3 while mapped point 4 is dead on... but that's going to be pretty uncommon because in that case mapped point 3 would have to be off by a larger margin. This isn't necessarily a hard and fast rule, and you can play around with the numbers as you see fit. For instance if your values are always dead on until mapped point 3 is activated you could do the same weighted correction, but only apply it to mapped point 3 assuming the entire error is in that mapped point. If you wanted to attempt a 'one shot' fix then you'd have to multiply your correction factor by the inverse so that when it is only weighted 40% it applies the appropriate correction, but doing such a drastic change could cause you to overshoot and chase around a bouncing number.

    Hopefully this helps some and good luck getting it nailed down, thanks for posting your progress so far for others to learn off of.



    Thanks for the reply.

    It's basically trying to solve a 2, 3, or 4+ variable function without enough data. You can guess and solve it in one circumstance, only to have it be off in another.





    Quote Originally Posted by murfie View Post
    Correctly calibrating Each mapped point would require you to keep the cams locked at defined angles. Then go through all tables and calibrate them. This gives you a point with defined maps aka a mapped point. Once all mapped points are calibrated then the ecu will blend between them and generally the conditions will match as thats what mapping things out does for you. There's no way to get around that it's time consuming. The more MPs you have calibrated the better cam control you get. You can simplify and only use a few mapped points. You can also do what was mentioned above and filter data by mapped point weight and try to work with what was already mapped.


    Murfie, that makes sense. I was thinking about it somewhat differently. By locking into specific MPs and logging, you may get into an operating condition (load/rpm) that you normally wouldn't see in that MP alone. But that's okay. The MP is most importantly defining cam angles, which probably have the biggest affect on the rpm/load/MAP relationship you are trying to tune.

  2. #42
    Advanced Tuner CCS86's Avatar
    Join Date
    Nov 2017
    Location
    Austin
    Posts
    913
    Hey guys, I'm running into some difficulty with the SD calculator.

    I locked out my tune into single MPs and generated a good amount of data to fill in the calculator table (RPM, air load, actual MAP). I filled it in, extrapolated, interpolated and smoothed everything into this:


    SDC-OP-Before1.png
    SDC-OP-Before2.png





    After clicking 'Calculate Coefficients', this is what I get:


    SDC-OP-After1.png
    SDC-OP-After2.png



    Look at how big some of the % differences are for individual cells.


    SDC-OP-Percent.png


    I know that this table is only an input, to populate the quadratic coefficients, and the solver is doing it's best to fit the data. But these changes just seem to big to give accurate results.

    Is there some sort of technique in prepping this table so that it has an easier job fitting, with smaller changes? When I was only modifying smaller sections and blending with the original, I saw much smaller changes to this table. Now that I am trying to start from scratch, they are much larger.

  3. #43
    Tuner
    Join Date
    Jun 2017
    Location
    South East Georgia
    Posts
    127
    This probably won't be super helpful... but I believe that's a limitation of the equation when the mapped point winds up being a fair distance outside the range it would actually be referenced in. Do you happen to have a more stock Ford vehicle to try a similar methodology on that you could then compare your results to the stock results to confirm if that's that case?

  4. #44
    Advanced Tuner CCS86's Avatar
    Join Date
    Nov 2017
    Location
    Austin
    Posts
    913
    Quote Originally Posted by Puggyberra View Post
    This probably won't be super helpful... but I believe that's a limitation of the equation when the mapped point winds up being a fair distance outside the range it would actually be referenced in. Do you happen to have a more stock Ford vehicle to try a similar methodology on that you could then compare your results to the stock results to confirm if that's that case?


    Nope, all I have is my 2012 GT with M122 blower. I realize that this will have a very different MAP curve to stock.

    I'm kind of shocked that the Roush575 tune, for example, has stock SD tables. I'm finding them to be used heavily all the time in fueling.

  5. #45
    Tuner
    Join Date
    Jun 2017
    Location
    South East Georgia
    Posts
    127
    I don’t drop into the non-ecoboost forum often enough it seems heh. Have you seen a log of the Roush575 tune? I’m wondering if they just let the fuel trims and wideband take care of fueling?

  6. #46
    Advanced Tuner CCS86's Avatar
    Join Date
    Nov 2017
    Location
    Austin
    Posts
    913
    Quote Originally Posted by Puggyberra View Post
    I don’t drop into the non-ecoboost forum often enough it seems heh. Have you seen a log of the Roush575 tune? I’m wondering if they just let the fuel trims and wideband take care of fueling?


    When I was using Lund, I believe he started with that tune as a base. It was terrible in all regards: stalling, throttle response, transient fueling.

  7. #47
    Tuner
    Join Date
    Jun 2017
    Location
    South East Georgia
    Posts
    127
    All that would make sense if the SD tables were off. I’m surprised as well they’d ship that as a ‘tune’ though.

  8. #48
    Senior Tuner
    Join Date
    Jan 2013
    Location
    Hawaii
    Posts
    1,474
    So I have been plotting data and getting regression lines in excel. After entering the coefficients into the cells, I thought i'd see what the calculator comes up with for new load values. I found it interesting as the calculator seems to like it.

    one line HPT SD calc number 3.PNG

    one line HPT SD calc number 2.PNG

    NEW data regression.PNG

    Another problem with the calculator I see is you have to use your MAP sensor data in one of the AXIS values. Then you populate the table with load values. Usually the load value from the OBD port has a lot less data points coming out of it relative to an external sensor. I think many people are trying to populate a graph and paste it into the calculator and it is just not happy with the data. Then people try to manipulate it just to get it to work and get a major headache from it.
    I like to set the load/ air mass as the row axis, and collect the sensor data to plot data points from the sensor. I find you can get many more valid points for your plot. Also using excel I'm not limited to the number of data points or get points that were just an average of many other points(not good for accurate regression). The scanner doesnt like pulling out graphs with 100+cells but it will do it.
    "We can never be right, we can only be sure that we are wrong"- feynman

  9. #49
    Tuner in Training
    Join Date
    Dec 2008
    Posts
    47
    Quote Originally Posted by 4wheelinls1 View Post
    So I had a couple of hours to play but have run out of time on this car. I just played in Mapped point 0 0,0 cams. Using an external MAP sensor through MVP1 I did some recalculating of load as I understand it. I did notice some improvement to calculated MAP however is was still drifting as the pressure rose, possibly the limitation of the formula? I also introduced some blowthrough but that only seemed to worsen the problem. I did have all sorts of issues with Air mass throttle correction closing the blade and stalling the car, I gather I need to look back to my torque/inverse tables, Its quite odd the car seemed to go miles backwards in throttle surge to the point that it would force a throttle shut down. By turning off all the desired air mass the throttle was stable and did not shut down. I'm thinking that turning off the desired air mass may be important prior to logging so it's not chasing its tail then to re-enable it once the torque/inverse tables are corrected. I have another car tomorrow but again will be pushed for time so will probably leave the old tune as it was as that works. One step forward two steps back.

    There are a number of logs attached that show the throttle closing and throttle corrections as well as the error in absolute vs Calculated MAP.

    Cheers
    Where do you turn off the desired airmass tables?

    Cheers

  10. #50
    Advanced Tuner
    Join Date
    Jul 2017
    Posts
    392
    Hi guys,
    Back on trying out this calculator, I have a MAP sensor hooked up, 2014 GT500 car & ecu, letting the MAP sensor populate the load in the cells, then copying & pasting these values into the calculator. Obviously some cells can not be populated so some cells are not changed, so I wind up smoothing, or interpolating before using the calculator, if I don't, data is all over the place, and car runs like bad data, if I smooth it then calculate, way better.
    Who else is using this calculator with any success? How are you using it?