FFB: Smoothing VS change limit

Discussion in 'Wish Lists' started by Lazza, Oct 24, 2012.

  1. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    First, I know there's a smoothing option for FFB for us poor G25/27 owners. What I can't find is a simple 'max change per frame' like there is in rF1. This option would allow people to reduce the more aggressive FFB reactions while not losing any of the more subtle effects - which is what smoothing does.

    A quick example; take this raw FFB input, with some spikes representing kerb/grass reactions:

    4.0 , 8.0 , 25.0 , 5.0 , 30.0 , 12.0 , 6.0 , 8.0 , 18.0 , 8.0

    If we apply smoothing over 3 frames, and for this example I'm just using a simple average, the output looks like this (ignore the first 2 frames, since there's not enough frames for smoothing yet):

    4.0 , 8.0 , 12.3 , 12.7 , 20.0 , 15.7 , 16.0 , 8.7 , 10.7 , 11.3

    You can see the spikes (25, 30) are removed, but you also lose some fidelity on the other values.


    This is better illustrated with a series of values that might represent normal road noise:

    4.0 , 6.0 , 7.0 , 5.0 , 4.0 , 6.0 , 9.0 , 8.0 , 6.0 , 4.0

    with smoothing (3 frames) this becomes:

    4.0 , 6.0 , 5.7 , 6.0 , 5.3 , 5.0 , 6.3 , 7.7 , 7.7 , 6.0

    So smoothing does 'cure' the spiking, but it also numbs all the smaller effects. And in practice removing the rattling over kerbs and grass tends to require smoothing over maybe 5 or 8 frames, which further flattens the FFB response and introduces some potentially noticeable lag as well.


    With a 'max change per frame' option, I might decide through some quick testing that a change of 10 per frame is acceptable, but I want to remove anything higher.

    Using the same inputs as above, here's the result of a 'max change = 10' on the spiked input:

    4.0 , 8.0 , 18.0 , 8.0 , 18.0 , 12.0 , 6.0 , 8.0 , 18.0 , 8.0

    and the 'normal road noise':

    4.0 , 6.0 , 7.0 , 5.0 , 4.0 , 6.0 , 9.0 , 8.0 , 6.0 , 4.0

    The first line has removed the spikes ( > 10) with minimal effect on smaller values, while the FFB noise under my selected threshold remains unchanged.


    This is why I don't think the smoothing options cuts it, and I don't see why the better option from rF1 has been removed. (default was 15% of full FFB per frame - and I bet 95% of players never changed it or even knew it was there)

    I would happily make a simple plugin to provide this option if for some reason ISI doesn't want to implement it, but the ForceFeedback plugin interface doesn't appear able to actually change the FFB output. (can see the input, but any changes are overridden by rF2)

    If I've missed an option that addresses this, or am doing something wrong on the plugin side, let me know...
     
  2. deak1944

    deak1944 Guest

    Just checked my clean install of rfactor and the default was 100, which means disabled.
     
  3. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Appears you're right there. I definitely found 15 and had no reason to have changed it previously, but just checked my own clean install of 1255f and yep, 100 it is... maybe it changed in one of the last versions... or I changed it in my sleep...

    Anyway, still want it back in rF2 :)
     
  4. kaptainkremmen

    kaptainkremmen Registered

    Joined:
    Sep 25, 2012
    Messages:
    935
    Likes Received:
    17
    Assuming your figures are correct, the forces with Smoothing increase when the actual data decreases. I'm looking along the line of numbers and doing a higher/lower count.

    4.0, hi, hi, lo, lo, hi, hi, lo, lo, lo

    4.0, hi, lo, hi, lo, lo, hi, hi, =, lo

    It looks so wrong! I would like to try 'max change' to feel the difference. Very interesting.
     
  5. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    That's true for the example I've given, but obviously how they correlate depends on the actual data. With FFB running at 400 Hz it would be unlikely for the data to oscillate quite so quickly. But, this is why the smoothing comes with a disclaimer that it may introduce a delay - the raw FFB might have started climbing, but because the smoothing is using previous values too it might still be dropping. So real fast data might have you largely on the correct side (rising when it's rising, dropping when it's dropping) but you will lag behind a bit on the transitions.

    What I think is being overlooked here is that there are input filters available (again, the same as rF1) which makes sense for an input because you tend to hold throttle, brake, clutch, steering, ... at either a constant value or in a slow progression from one point to another - smoothing will have very little adverse effect apart from lag if you smooth over many samples.

    The whole point of FFB is having those small changes, so filtering seems a wholly incorrect approach. I just can't understand why it's being pushed as the 'solution' to horrendous rattling in some wheels, let alone how that rattling with some of the most common wheels made it through beta testing without a fix/workaround (even if it were just the smoothing for starters) being obviously available. Like in the GUI as it is now.

    The weird thing is that the max change option's been removed since rF1 and is a simpler process than smoothing. FFB setting (via the plugin interface) not working is understandable given everything's still under development, just a shame in this case since it can't be used as a workaround to the missing option... *sigh*
     
  6. Jim Par

    Jim Par Registered

    Joined:
    Jan 9, 2012
    Messages:
    11
    Likes Received:
    1
    Bringing back to life this thread as I haven't found a solution about this for rF2 yet.
    Does anyone knows anything new about this "Max change per frame" setting missing from rF2?
    Do we have another similar setting now?
    FFB smoothing is not an option for me, it stays "0".
    I just want to ease off the FFB on kerbs and off-road.
    Thank you,
    Jim.
     
  7. Skan

    Skan Registered

    Joined:
    Dec 14, 2014
    Messages:
    333
    Likes Received:
    135
    You can edit the ffb of the off-road directly in the controller.json file. I have not messed with that setting so I do not know how it might negatively affect any other parts of driving or even if it will at all.
     
  8. Jim Par

    Jim Par Registered

    Joined:
    Jan 9, 2012
    Messages:
    11
    Likes Received:
    1
    Yes I have seen this option but it doesn't change how the kerbs are feeling unfortunately.
     
  9. Spinelli

    Spinelli Banned

    Joined:
    Jan 28, 2012
    Messages:
    5,290
    Likes Received:
    32
    If the FFB is only changing certain forces, rather than a curve/multiplier that affects the entire range of forces, then I believe that will not be considered "true" FFB from the steering rack, since you are only picking specific forces to adjust rather than applying a curve/multiplier to the entire range. If I am indeed correct, then good luck with seeing this setting, there's a common misconception among many, but not all, simracers that all the matters, and all that is realistic, is to have only the pure steering rack forces and forget anything and everything else.

    If your FFB setting only affects certain forces, rather than a curve/multiplier affecting all, then that is heading in the direction of the rF1 FFB system where you can individually tune things how you want. I much, much, much prefer that system as well as it offers so much more feel based on your wheel, your settings, your simracing setup, etc. etc., and I hope ISI starts allowing some of this stuff again, at least as an option for those of us who understands just how powerful and beneficial/necessary this can be when setup correctly in terms of the "overall driving/racing experience" rather than just what's technically correct on paper with regards to the steering/wheel feedback.
     
    Last edited by a moderator: Dec 29, 2014
  10. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    Hey Spinelli,

    I suspect you've misread my post - or maybe you're just taking the concept to the extreme :)

    The wheel-jarring forces (which have been reduced since I wrote that post 2 years ago) are of greater magnitude than most/all 'normal' driving forces. A filter that you use to reduce those large spikes in the FFB also reduce all small spikes in the FFB (ie the stuff you want to feel, and purists should want to feel). The rF1 FFB option to limit the FFB per frame would indeed mask some of the FFB but would do a better job of preserving the majority (time-wise) of effects.

    So, yes: it would change the nature of the FFB somewhat; the entire range from "only when I'm on the grass at 200mph" to "I've just killed my FFB completely" depending on what you set it at. But: you could say the same for the filter most of us are using now in rF2, and I suspect a limiter would actually do a better job for this issue. Unfortunately it's been unable to be tested because plugins can't do it. Not with rF2's blessing anyway, and I can't be bothered trying to do it another way.
     
  11. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Why not simply increase or decrease the feedback in the ratio, evenly matching the capacity of the terminal/hardware. The torque and feedback charecteristic would not change only the ratio over the full bandwidth of forces. The manipulation of power ratio would become unnecessary unless there are signal peaks which should generally be limited or filtered, but if it is as you say, the filter could be seen as a kind of compensation for damping from a steering system, the many vibrations, forces reduce by geometry, gear ratio or power steering, but with the side effect of setting a total limit over the full bandwidth, what means that everything above the limit is impossible.
     
  12. Skan

    Skan Registered

    Joined:
    Dec 14, 2014
    Messages:
    333
    Likes Received:
    135
    Jim did You try the rumble strip values in the json file?
     
  13. Lazza

    Lazza Registered

    Joined:
    Oct 5, 2010
    Messages:
    12,345
    Likes Received:
    6,572
    The issue with the G27 specifically is that above a certain threshold the noise it produces increases massively, I suspect because of one of the motor shafts being pushed along its axis by the FFB. Reducing that noise by lowering all forces, or by excessive smoothing which smooths all forces, doesn't really seem the best workaround.

    Being able to set a maximum change per frame/update would allow you to reduce or even eliminate the high energy inputs that produce the excessive noise, while leaving completely untouched the 'normal' inputs you'd get while driving on the track.

    *PS: isn't it slightly ironic that the options to 'fix' excessive wheel noise like this is to reshape all FFB (via smoothing) or explicitly change the FFB strength from specific inputs (off-road, kerbs)? ;)
     
    Last edited by a moderator: Dec 30, 2014
  14. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Understand, yes the G27 motor axles moving along the longitidunal axis due to the helical gearing.

    I don't know if i understand the concept right with the change per frame/update because i do not understand the coherence to frames, but wouldn't it be more accurate to match up the ratio to the hardware, so that the software would calculate the ratio determined by the max capability.

    The filter would become needless if there is no signal producing that noise anymore. I mean where is the difference if there is a torque force applying which overcomes the max limit of the hardware and produces clipping and noise, or a vibration, jolt whatever effect signal is overdriven, when the max capability is always the limit.

    If the effects are wished, seen as usefull and realistic output of the physic, which in most cases wouldn't be there on a real steering, at least not always in similar level of impact, than it seems the choice of the developer to limit the feedback based on the peak level of effect forces, but if the choice would be real steering with nearly no such effects, the main forces would build the limit. So it seems a matter of choice and approach in the first place to me and how the feedback is transmitted to the terminal/hardware before the balance, which is out of shape.

    I'm just not a friend of randomness. :)
     
    Last edited by a moderator: Dec 31, 2014
  15. Jim Par

    Jim Par Registered

    Joined:
    Jan 9, 2012
    Messages:
    11
    Likes Received:
    1
    Yes Skan I have but not the change I was looking. It's still very hard to keep running on these kerbs, especially if you want to be competitive and consistant.
     
  16. Jim Par

    Jim Par Registered

    Joined:
    Jan 9, 2012
    Messages:
    11
    Likes Received:
    1
    I so much agree with you Lazza. I wish ISI will make that happen. It's a shame to lose all that great ffb detail on road because of that kerb, off-road extreme noise.
     

Share This Page