1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

FFB Wheel Analysis using WheelCheck from David Tucker

Discussion in 'Sim Racing Hardware' started by magicfr, Mar 13, 2013.

  1. magicfr


    Wheel Check Data analysis
    WheelCheck presentation

    Wheel check is a tool made by David Tucker from iRacing.com.
    This tool is ouputing position of wheel when applying a constant force.

    Here a graph of the brute datas.

    The thing is , position of the rim is not very intuitive to compare two FFB Wheel.

    What is important on a FFB Wheel for best feeling and immersion

    1st important thing is the max holded torque of the FFB Wheel, this is the most important thing. A high torque wheel will be capable to give high force on your hand,and to change the speed of rotation quickly

    2nd thing, is the max speed, when you release the rim for counter steering or when drifting, it’s important that the wheel is capable of turning fast, from several source, it seems that 200rpm is a good value for max speed.

    Good news, it’s quite easy to compute speed from position data.
    Bad news, the position data is quite noisy or low frequency for some wheels.

    For example the CSR-E has noisy input, and the T500RS is only 100hz in this firmware ( if somebody have more recent data with lastest T500RS firmware please post an answer ).

    Interpolating and smoothing data.

    1st pass of the algo is to interpolating data to 1000hz.
    Then I smooth the data with a convolution, I did a 5 pass of convolution with a size of 4 datas.

    Here the smoothed graph

    Computing Speed

    From the smooth data it is easy to compute speed.
    Each data is poll with 1ms time between them.
    So Speed is speed = pos(t+1) - pos(t) / ( timeStep).
    With timeStep in seconde = 1/1000
    This give use a speed in DirectXValue / seconds.

    The wheel are set to 900°, that means that from -10000 to +10000 we have 900° so a directX value
    of 1 is equal of 900°/20000 = 0.045

    So Speed in degree/s = (( pos(t+1) - pos(t)) * 0.045) / (1/1000) = 1000 * ( pos(t+1) - pos(t)) * 0.045
    We want Speed in RPM, Rotation Per Minute, 1 rotation = 360degrees
    Speed in RPSecond = 1000/360 * ( pos(t+1) - pos(t)) * 0.045
    Speed in RPM = 60 * 1000 / 360 * ( pos(t+1) - pos(t)) * 0.045

    And here the resulting graph:


    With this graph it’s way easier to compare wheels.
    The highest the value, the highest the max speed of the wheel.
    The more vertical the slope of the graph the more the wheel has torque ( acceleration ).

    Logically my High torque ( 12N.m ) wheel has the more vertical slope when acceleration, when the curve is going away from 0.

    We can see that other wheel have strong decceleration when deccelerating ( slope is going back to 0 ).This indicate that they have more mechanical friction.

    You can see that, at end of graph, when WheelCheck don’t give any force. Most wheel are going back to 0 speed very quickly except the G25-E ( Arc Team) and my Wheel, it’s because those wheel have low friction, and so they continue to turn on there momentum.

    You can also see that around 1429ms, those both wheel have speed that reach 0, if you compare with the position graph, it’s because they reach mechanical stop of the wheel ( on my wheel it’s a FFB stop ).

    Last thing, my wheel is accelerating so quick, that it reach it’s max speed very fast.

    Why no Acceleration Graph

    I tried to compute the acceleration graph, but the more you integrate value, the more the result is noisy, so the graph was not usefull.
    It’s better to compute acceleration from the speed graph.
    Best way is to compute acceleration by calculating how much time the wheel take to reach max speed.

    For example for my wheel 42V 5A, 0 rpm to 183rpm in 130ms, that is 183/(0.130) ~= 1400 RPM/s

    with 48V 6A, 0 rpm to 207rpm in 100ms, that's 207/(0.1) ~= 2070 RPM/s

    For the CSW F1 it’s 0 rpm to 180rpm in 300ms , that is 180/0.300 ~=~ 600 RPM/s acceleration.
    • Like Like x 1
  2. magicfr


    [Reserved for updated graph datas]
  3. Pax7

    LifeOn2 Development


    Yep, as shown by MagicFr, WheelCheck can be useful to test and measure various aspects of wheel performance. David first released it in May 2011 and there is a wealth of discussion and practical test results in its official thread, here:


    One word of caution however: It has been shown that some of the published test result data for wheels is faulty due to erroneous wheel settings and/or data interpretation. I have personally shown one data set to be erroneous, in autumn 2011 I think it was. So, if you need to be sure to have correct reference data, you should perform the tests yourself, or let somebody you know can do it correctly perform them.

    I btw used WheelCheck and a data analysis SW to make a performance evaluation of the Fanatec CSR Elite in Dec 2011, here


    EDIT: No luck with writing fast posts with links today it seems... ;)
  4. I have modified my CSW with new motors and I am running a peak of 350 RPM when checked with wheelcheck. The graph looks similar to the Magic wheel but with ~350 instead of ~200 RPM. I will probably modify the drive ratio to get the RPM down a bit, is this too fast in your opinion? It is able to hold 13Nm of torque. If I drop the ratio it will get even stronger. Obviously I didn't bother with the smoothing on my graph, but you get the idea. I did interpolate the 500Hz data to 1000Hz though. My wheel does 0-350 RPM in 107ms, or 3271 RPM/sec. acceleration.

    P.S. for the folks who aren't real good at analyzing graphs, you'll notice flat spots on the RPM curve, these are not anomalies, they are where the wheel reached the called for position and stopped, so there was 0 RPM.

    I would be happy to provide the test data if anyone wants it. I attempted to upload it but apparently .csv files are not allowed.


    Here is a short vid of the wheel running wheelcheck @36VDC:
  5. magicfr


    I think 200rpm, is way enough, so you can reduce tension ( V ) and keeping intensity ( A ) at it current value, so you will get less speed with same torque.
    Congrat with the modification :)
  6. Thanks, it is definitely like a whole different wheel, it feels amazing. I don't think I can just reduce the voltage without dropping the amperage, since they are related. I think maybe a slightly bigger pulley at the left end of the base would be better way to go to maintain the torque. This is easy for me to make as long as I have room. It should be OK, it doesn't take much of an increase in diameter to get a good increase in circumference.
  7. 200t/mn is the goal and yes 350t/mn is too fast.

    You can see professional FFB Wheels by Sensodrive have 200 t/mn...


    Yes you can reduce voltage without dropping the ampérage. Take a 24V power supply with the same intensity (A) like your actual power supply and you will have the same torque...

    What is the motors used in your combo ?
  8. magicfr


    I agree with Stef, you should be able to reduce voltage with same intensity.
    Electric motor torque is dependant to intensity
    Electric motor speed is dependant to voltage
    intensity*voltage=watt = powe

  9. Pax7

    LifeOn2 Development

    Hello eKretz,

    very nice to see this kind of mod to the "new" Fanatec line of wheels. It would be interesting to have more details on what changes you have made.
    I see you are altering the motor power supply input voltage. Are you simply using another power supply for the entire wheel system (potentially hazardous), or are you feeding higher voltage just to the motor power electronics?

    Obviously, if a mod like this is done properly, it could well be a big hit in the community. If you put together a little kit and DIY instructions, you will probably have a nice little extra income :)

    And also

    I am not sure I would trust Sensodrive specs alone to be the reference for "correct" peak rotational velocity.

    It would be prudent to point out that it is not necessarily that simple. Electrical circuit theory laws are in effect in this case too ;)

    But sure, depending on the power supplies, the motors and how the motors are connected, it may work like you describe in this case.
  10. I have too an experience with my homemade wheel. I tested a 60V power supply with the wheel to get 281 t/mn. Sure it is too fast for vintages cars with more than 900° (when the car drift the wheel turns too much quickly)... ;)

    We have also made the Niels Donuts test to see if the wheel returns correctly to center after donuts with more than 900°, and with 200 t/mn it's ok...
  11. I am using a bench power supply with much better motors, and the circuitry (Drive chips, etc.) in the wheel is rated for 40V/50A max. so it is fine as long as it doesn't overheat. I have 70V/20A available. The board in the wheel regulates input voltage for all the components that need it. This is not a mod that can be done without extra cooling though, as the chips get hot quick at 40V not to mention the motors themselves. I am running a water cooling setup, it is still in progress at the moment:






    Yes, the reason I say that I can't drop the voltage without dropping the amperage is that when the voltage is lowered, the motors draw less amperage, this is all interrelated. However, I have experimented with several different voltages etc. and even at the stock 24V it is a very nice improvement. at stall at 24V IIRC it draws ~8.2A and at 40V about 13.5A. At 24V the stock power supply (rated 5A) is incapable of functioning as the voltage drop is too much and the wheel shuts off. The wheel's circuitry pretty much gives a straight amperage pass-through from the power supply to the motors, which lets them draw whatever they are capable of via their resistance.

    All that being said I am still experimenting to find the combination that is best.

    BTW, the website link doesn't work for me. It says website not found in German (which I speak semi-fluently). I also tried just plain sensodrive.de and it seems to be down.

    Oh, just re-read and forgot to mention that the motors are Buhler
  12. Ohhh, those are wicked sweet!
  13. I'm sure they are wicked expensive as well. Definitely interesting project, guys. Will follow.
  14. I had a quote of 15.000€ for the SENSO-HT :O_o:
  15. magicfr


    I tested last wheelCheck 1.62, and it's bugged as hell.
    When my wheel is centered, it read -20000 for position, I check with DIView and Assetto corsa, directInput value is really 0.
    So I can't test my wheel anymore.
    That sad, the value on those graph are with 42V 5A, and now my wheel is with 48V 6A, I wanted to compare but the exported CSV are wrong.
  16. magicfr


    I fond a 1.6 version which working perfectly.
    Updating 1st post with my new graphs :)
  17. Nice analysis! I too have modded my Fanatec, in my case using four stock motors most of the time along with some drag reduction and heat tweaks.

    An interesting result is the behavior of the T500 in your RPM graph. It had a heavyweight punch in step 1 and then faded to a lower weight class for steps 2 and 3.

    I'd want to see that confirmed and examined. If it is occurring I wonder what is causing it and if it can be rectified. It's not belt slip since that device uses 5mm pitch cogged belts which don't slip. The motor would still be cool. So I am wondering about protection or circuit board or PSU shortcomings.
  18. magicfr


    Step 1 is a bit special because it's start when the wheel is at rest at 0 RPM.
    All other steps, the wheel is turning and so the inertia of the wheel and mechanichal friction are more important.