Page 1 of 4 1234 LastLast
Results 1 to 20 of 65

Thread: using the transient (dy/dx) filters

  1. #1
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856

    using the transient (dy/dx) filters

    Can somebody post an example of a working transient filter? I tried messing with it, but couldn't seem to get it to work the way I want it to. Also, what is the time split on the dy/dx filter? Is it just the change since the last reported frame of data?

    Basically, I'd like to set up some filters that throw out anything where the throttle blade has changed position more than 2% and MAP has changed more than 3kPa.

    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!

  2. #2
    Tuner Fast9C1's Avatar
    Join Date
    Oct 2008
    Location
    Allison Park, PA, USA, Earth
    Posts
    62
    I have nothing to contribute so am VERY interested to see what happens here.

  3. #3
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Bueller?

    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. #4
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    when did we GET transient filters?
    Steve Williams
    TunedbyFrost.com


  5. #5
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Variable transition. See the attached screen shot.

    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. #6
    Senior Tuner LSxpwrdZ's Avatar
    Join Date
    Dec 2008
    Location
    on the Dyno
    Posts
    1,825
    Interesting! I will play with that and see what I can come up with. This will work great for those customers that are throttle happy haha!
    James Short - [email protected]
    Located in Central Kentucky
    ShorTuning
    2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
    2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400

  7. #7
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Quote Originally Posted by LSxpwrdZ View Post
    This will work great for those customers that are throttle happy haha!
    I have come sooooo close to smacking some people...

    Me: "Drive smooth, no hard throttle."

    Them: "OK!" *WOT off the stop light*

    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. #8
    Senior Tuner LSxpwrdZ's Avatar
    Join Date
    Dec 2008
    Location
    on the Dyno
    Posts
    1,825
    Quote Originally Posted by DSteck View Post
    I have come sooooo close to smacking some people...

    Me: "Drive smooth, no hard throttle."

    Them: "OK!" *WOT off the stop light*
    Exactly or the ones that seem to have torrets or something cruising at 60mph and giving it half throttle and let off and continue to make abrupt throttle changes even after you tell them to make smooth easy throttle transitions. Those kinda tunes take forever and end up having alot of interpolation and extrapolation done to alot of the fuel tables.
    James Short - [email protected]
    Located in Central Kentucky
    ShorTuning
    2020 Camaro 2SS | BTR 230 | GPI CNC Heads | MSD Intake | Rotofab | 2" LT's | Flex Fuel | 638rwhp / 540rwtq
    2002 Camaro | LSX 427 | CID LS7's | Twin GT5088's | Haltech Nexus R5 | RPM TH400

  9. #9
    Senior Tuner
    Join Date
    Nov 2002
    Location
    Central Florida
    Posts
    2,503
    Does anyone know the actual "unsteady MAP" timeout
    (i.e. how long after returning to "steady" before the
    airflow returns to MAF-blend), and what the "unsteady"
    MAP or air mass delta threshold is?

    I have generally avoided using the histograms because
    of all the transitional garbage they accrue, but would
    love to come up with some generally-good filters. My
    usual method is to slide through the logs looking for
    little "regions of steadiness", picking off the error at
    that point and pushing it into VE table, then eyeball
    fitting the surface to the sparse new data set.

    I guess I'd be looking for

    steady_MAP*steady_AFR*sorta_steady_RPM

    but how to express it, I dunno.

  10. #10
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    Quote Originally Posted by jimmyblue View Post
    Does anyone know the actual "unsteady MAP" timeout
    (i.e. how long after returning to "steady" before the
    airflow returns to MAF-blend), and what the "unsteady"
    MAP or air mass delta threshold is?

    I have generally avoided using the histograms because
    of all the transitional garbage they accrue, but would
    love to come up with some generally-good filters. My
    usual method is to slide through the logs looking for
    little "regions of steadiness", picking off the error at
    that point and pushing it into VE table, then eyeball
    fitting the surface to the sparse new data set.

    I guess I'd be looking for

    steady_MAP*steady_AFR*sorta_steady_RPM

    but how to express it, I dunno.

    The histo's work great for me as long as I don't just take a whole trip's worth and try and use the data in one fell pasting swoop. I use about 6 different VE histo's with different filtering and I change the filtering and re-run them on each log examination.

    The crappiest part (and a common mistake of noobz and even some pro's) is that transition to WOT where for an instant or two, Commanded AFR falls but actual AFR is still high. I routinely see that false error copied. I'd like to be able to shift the EIO data forwards or backwards a fame or two and have the log re-compiled personally.

    The steady/unsteady state parameters seem to move very quickly.

    As for this thread, I recall DY/DX as the derivative of y with respect to x but I don't necessarily see the fit. Knocking off inner brain dust, I seem to recall the theory of derivative estimating (though Leibniz's notation seemed to be the focal point of study) where:
    dx = change in x
    dy = change in y
    x = unchanged value of x
    y = unchanged value of y

    Maybe if Marcin is strolling around he could stop by and comment cough cough
    Steve Williams
    TunedbyFrost.com


  11. #11
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    When I was in school, I could crap out derivatives and integrals left and right. I'm just hoping for somebody to chime in (possibly from HP Tuners) about correct usage in filters. My main question is how is the dy/dx parameter used... My first assumption was that it's just the change in the value from the last recorded frame, but things rarely ever work the way you imagine them to.

    Honestly, all I really care about is change in values with respect to time.

    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!

  12. #12
    Advanced Tuner umrjmac's Avatar
    Join Date
    May 2010
    Location
    Lenexa, KS
    Posts
    301
    Quote Originally Posted by DSteck View Post
    When I was in school, I could crap out derivatives and integrals left and right. I'm just hoping for somebody to chime in (possibly from HP Tuners) about correct usage in filters. My main question is how is the dy/dx parameter used... My first assumption was that it's just the change in the value from the last recorded frame, but things rarely ever work the way you imagine them to.

    Honestly, all I really care about is change in values with respect to time.
    Amen. If we confirm that is how the new filter options work, I'll have an email in to get the beta within 30 seconds of reading the post
    Kenne Bell Supercharged 2003 Corvette Z06

  13. #13
    Quote Originally Posted by DSteck View Post
    When I was in school, I could crap out derivatives and integrals left and right. I'm just hoping for somebody to chime in (possibly from HP Tuners) about correct usage in filters. My main question is how is the dy/dx parameter used... My first assumption was that it's just the change in the value from the last recorded frame, but things rarely ever work the way you imagine them to.

    Honestly, all I really care about is change in values with respect to time.
    it's the change in the value divided by the time between frames (the name should probably be dx/dt). Anyway:

    dy/dx = (y2 - y1) / (t2 - t1)
    I count sheep in hex...

  14. #14
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    Quote Originally Posted by Chris@HPTuners View Post
    it's the change in the value divided by the time between frames (the name should probably be dx/dt). Anyway:

    dy/dx = (y2 - y1) / (t2 - t1)
    Nice,

    Care to give a practical implementation example for us?

    Let's say we want to exclude AFR % err values from a VE histo when the TPS is 'moving'. I know this is relative. Are they in use as a filter or part of a user defined PID whose result is then filtered?

    Thanks Chris.
    Steve Williams
    TunedbyFrost.com


  15. #15
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    (y2-y1)/(t2-t1) .... ABS could be useful here as well to keep things in line.

    y = TPS %
    t = time

    dy/dx
    (y2-y1)/(t2-t1)
    (20% - 18%)/(15sec-16sec) gives us a 2%/sec ROC (rate of change). We could filter the function compared to a static value or range, correct? If so, can you give a string example?
    Steve Williams
    TunedbyFrost.com


  16. #16
    well i guess you could make a custom PID like:

    TPSRate = abs([PID.17.dy/dx])

    This will give you the rate of TPS change, positive or negative.

    then filter would be: TPSRate < threshold

    or you could just put it into the filter directly. If you make a custom PID at least you can see what kind of numbers the function is returning so you can set the threshold by experimenting a bit.
    I count sheep in hex...

  17. #17
    Senior Tuner DSteck's Avatar
    Join Date
    Mar 2008
    Location
    St. Louis, MO
    Posts
    4,856
    Quote Originally Posted by Chris@HPTuners View Post
    well i guess you could make a custom PID like:

    TPSRate = abs([PID.17.dy/dx])

    This will give you the rate of TPS change, positive or negative.

    then filter would be: TPSRate < threshold

    or you could just put it into the filter directly. If you make a custom PID at least you can see what kind of numbers the function is returning so you can set the threshold by experimenting a bit.


    I guess this explains why my attempt at filtering MAP changes less than 2kPa weren't working. 2kPa/0.032s is 62.5 kPa/s!!! I was trying to just type in the transient as less than 2.

    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!

  18. #18
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    Quote Originally Posted by Chris@HPTuners View Post
    well i guess you could make a custom PID like:

    TPSRate = abs([PID.17.dy/dx])

    This will give you the rate of TPS change, positive or negative.

    then filter would be: TPSRate < threshold

    or you could just put it into the filter directly. If you make a custom PID at least you can see what kind of numbers the function is returning so you can set the threshold by experimenting a bit.


    I have been playing with that for a few minutes and all I get back is TPS expressed with 2 decimal places... does d/x need to be defined or is it frame related?
    Steve Williams
    TunedbyFrost.com


  19. #19
    Senior Tuner
    Join Date
    Dec 2004
    Location
    Laurel, MD
    Posts
    1,020
    i was about to say, watch out which kind of axis you're doing it on. are you using absolute time, or frame count? remember time intervals change with different amount of bandwidth used, plus different ecu's have different bandwidth limits and scanning speeds...

  20. #20
    Senior Tuner Frost's Avatar
    Join Date
    Jan 2006
    Location
    Richmond VA
    Posts
    1,950
    I'd much rather use absolute time for portability of the expression.

    I tried about a dozen arrangements including multiple custom PIDs in an attempt to use the engine run time parameter. There is something just out of grasp right now for me.

    When I use the string that Chris posted, as I said, I get TPS value with 2 decimal places which seems to indicate that dx is a constant equaling 1 in the expression?
    Steve Williams
    TunedbyFrost.com