Sim Racing servo ffb systems : OSW & Bodnar

Discussion in 'General Discussion' started by Adrianstealth, Jun 1, 2015.

  1. Adrianstealth

    Adrianstealth Registered

    Joined:
    May 28, 2012
    Messages:
    4,490
    Likes Received:
    1,004
    )-: queued for 30mins at the Royal Mail collections for my friends servo system, but it was just screen protectors for my tablet

    darn he's going to be disappointed as he's back off hols today and wanted to get it going this weekend

    sent an email to Ollie as would have expected delivery by now


    EDIT : someone in the Iracing forum just received theirs ( his was order 70 ) , my pals was 73 so I'm guessing next week now ( + put my mind at rest as thought it was lost or something ) phew
    ( my pals still going to be disappointed lol ....ill let him have a few hours on my rig to keep him going ) , some big releases from Iracing next week too
     
    Last edited by a moderator: Dec 4, 2015
  2. Euskotracks

    Euskotracks Registered

    Joined:
    Nov 29, 2013
    Messages:
    1,093
    Likes Received:
    190
    I think I didn't explain myself correctly.

    Since the wheel receives FFB input, it needs to be working in "torque control mode". I don't mean with this that the torque needs to be measured in order to close the loop. I just want to make the difference between "position/speed control mode which is very well controlled thanks to the encoder providing accurate feedback.

    It is clear for what you have said that the servo system uses current measurement to determine how much torque is providing. This can be tested and tuned accordingly to those test results. I am perfectly fine with this.

    However, I don't know how much time it takes from when the FFB input is provided to when the driver actually provides the calculated intensity. For sure it is not instantaneous. Depending on if this delay is negligible or not in terms of input lag, a PID control trying to anticipate FFB could be required.

    I would like to know about it since I am thinking in building my own system depending on the complexity of the task.

    Enviado desde mi GT-I9505 mediante Tapatalk
     
  3. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    I'm not sure if I'm understanding you correctly. Are you're saying servos can be used to determine how much torque it is outputting?

    Edit: Never mind. Think I realised what you mean but don't think it would work.
     
    Last edited by a moderator: Dec 4, 2015
  4. Euskotracks

    Euskotracks Registered

    Joined:
    Nov 29, 2013
    Messages:
    1,093
    Likes Received:
    190
    Servo systems are not used for determining anything but for providing motion/torque.
    In the case of a FFB device, the system is built to provide the torque that the sim sends to the FFB device.
    So in some way, the system needs to "know" how much torque it is providing. Up to here I think everything is quite clear.

    The way to know how much torque is being delivered is by controlling the delivered current.
    Apparently controlling current is not straightforward and is being done in closed loop control by accurately measuring it.
    I am basing these facts in Led's post:

    About the PID controller:
    The IONI/Argon controller in not closed loop on torque output but it IS closed loop with current output (current output is very finely measured in the IONI/Argon driver)
    So, the PID could be easily tuned to your likings in order to attain a situation like above with super reactivity, little oveshoot (and if you like no overshoot at all) and optimal stability


    I am not sure the reasons why this current wants to be controlled in closed loop which for sure adds complexity to the system.
    However I presume that the reasons could be among the following:

    - The electrical resistance is not a constant value and hence the voltage needs to be adapted.
    - The delay of the response in the permanent transitory state of the system needs to be somehow compensated to reduce input lag.

    As the system is working in closed loop, there needs to be some PID control trying to make that current match a determined set point.
    The PID has three constants that need to be fine tuned to match the desired result. In many real cases not all the three componnets are used.
    It will always depend on the dynamics of the system, on the input rate and on how fast do you want the response of the control.

    So my whole point is that any closed loop system requires its PID fine tuning in order to best provide the desired output.
    Since Ioni control for OSW seems to be working in closed loop, I wonder if those PID values have been tuned for rF2 FFB or for other racing titles.
    I would presume for example that different FFB rates would lead to different PID constants.

    So when I read people changing different settings as STS, dampenings, ... to improve FFB I think it is not the right way.
    Not at least if you want to simulate the real FFB of a car. My POV is the following.

    There are only two reasons why FFB could be unrealistic:
    1. Because rF2 calculates a wrong value for it. This could be due to a problem with the tire model for example.
    2. Because the FFB device cannot accurately reproduce the commanded FFB.

    For 1 there is no solution since it is highly unlikely that users can fix a physics problem by filtering and transforming the FFB input.
    For 2, if using a DD wheel, the way to procede would be to optimize the PID so that the device output best matches the input.

    In order to know if the wheel could be the cause of a poor FFB, its ability to reproduce a determined input should be tested.
    Apparently this cannot be done in a dynamic way but at least it can be done in a let's say static test.
    For that test the OSW performs very good according to the graphs posted some posts earlier.
    However I wonder if in other more complex dynamic circumstances the output is as good.
    If the only test available is the commented one, it is likely that the cloed loop PID variables have been tuned in order to optimize the results for that test.
    However that would not prove that an OSW perfectly provides the required FFB.

    I am just trying to understand how can be checked if a FFB device accurately provides the FFB commanded by the sim. Apparently is not possible.
     
    Last edited by a moderator: Dec 5, 2015
  5. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Ok, i'm on the same page as you now.
     
    Last edited by a moderator: Dec 5, 2015
  6. Led566

    Led566 Registered

    Joined:
    Jan 25, 2012
    Messages:
    358
    Likes Received:
    24
    Please refer to this page: http://granitedevices.com/wiki/Control_modes

    To put it VERY simple, in a motor like the MIGE (or Lenze or Kollmorgen used in Bodnar) there is a linear relationship between current delivered to motor and torque produced by the motor, for example MIGE declares 2.2 Nm/A for the ST10010 and it is valid with a fairly good approssimation at ANY rotational speed, i.e. the motor is rotating at 100 rpm and absorbs 1A then it is producing 2.2 Nm of torque at shaft, then you apply a braking force of 2.2 Nm and it stops but still absorbs 1A.
    And so on.
    This means if you measure current absorbed you are measuring torque delivered at shaft.
    This is also why controlling current in closed loop is like controlling torque in closed loop (with a fairly good approximation).

    Here we are dealing with AC (Alternate Current, not Assetto Corsa...:cool:), so we need to consider Impedance.
    Resistance doesn't vary with motor speed, Reactance does, this is why you need to measure and control current.


    When I wrote "PID could be finely tuned", it was for the sake of simplicity.
    What the end user could easily do with the IONI/Argon software is to tune TWO parameters that are in direct correlation with the parameters of the PID, in the software those are called MRU and MLU.
    How to do it is explained in details in the Martin Ascher DIY guide to OSW:http://ascher-racing.com/wp-content/uploads/OpenSimwheel-Tutorial.pdf
    Go to page 24 to see what is the step response of a correctly tuned system with an Argon and a MIGE motor.
    We are talking about 25 ms to go from zero to max torque.
    No overshoot.
    Typical lag I see (time that pass between setpoint change and the instant where motor starts moving) is below 5 ms.
    NO OTHER COMMERCIAL FFB SYSTEM (except of course Bodnar) COULD PERFORM ON PAR.
    In fact much of them are "lagging" WAAAAY behind.

    Now you are talking about tuning a completely different loop: you are adding the simulator, the computer, the videocard, the monitor and finally the pilot in-the-loop you want to tune.
    This is NOT possible NOR correct to do it in the IONI/Argon PID controller.
    The PID controller of the IONI/Argon deals only with the transfer function of the Driver+Motor system.

    All the rest should be (and in fact is) demandated to the parameters the user can tune in the sim/computer/pilot.


    Provided the IONI/MIGE system is correctly setup as specified above, that IS the correct way.

    You may have a point here.
    Basically you are saying that the step response of the FFB system may not be representative of the behaviour of the FFB wheel in ANY OTHER situation.
    This is perfectly possible.
    Based on my experience in design and control of very complex systems, and other considerations too long to be reported here, if the IONI/motor system is correctly tuned as stated above (i.e. the step response of the system is correct), I'm pretty confident that is not the case.
    On the other way I also already stated that is possible to record from granity software the torque setup + current output during an rF2 session, and make all the possible analysis you want.
     
  7. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Thanks led, enjoyable and insightful read.
     
  8. Euskotracks

    Euskotracks Registered

    Joined:
    Nov 29, 2013
    Messages:
    1,093
    Likes Received:
    190
    Thanks for your comments. I really appreciate your time replying.

    However I still cannot understand why the original FFB signal needs to be transformed.
    As I said, if the FFB is correctly calculated and the wheel does a good job reproducing it, why alter it with filters and transformations? I simply don't understand it.

    Enviado desde mi SM-T530 mediante Tapatalk
     
  9. Joe

    Joe Registered

    Joined:
    Nov 14, 2013
    Messages:
    372
    Likes Received:
    21
    My take on this is (learn from some reading):

    Servo motor could be very linear in terms of input signal vs output torque for given a constant speed (without any close-loop-feedback control).
    For those expensive servo motors such linearity holds even with rotational speed varying, extend up to pretty high speeds.

    For our sim steering wheel application, those DD servo motors probably are good enough without worry any feedback loop control on torque.
    For stepper motor, however, the linearity may hold up to 500 rpm only. this could be a problem for fast cornering. To model such problem, the way I can think is to use a torque sensor to test statically (with different constant speeds) and model the outcome. Then program in SW/firmware (predict) to fit the linearly. Hence apply to the steering wheel application. The true FID feedback loop for torque just will not work on sim racing application. Just not possible. (we have no "objective" torque to compare in real time and such close loop takes too long). If take 5ms, then we down by 50% performance in the FFB. Just not acceptable. Most FID control loop I saw are way over >>10ms.
     
  10. Beano

    Beano Registered

    Joined:
    Jun 9, 2015
    Messages:
    29
    Likes Received:
    0
    Hi,
    I have not read through the whole thread to see if you have fixed this, but this statement is not correct.

    The current IONI and Argon drives used in DIY DD FFB wheels drive permanent magnet servo motors, NOT induction motors. Also, there is a significant difference in quality between 'real' servo motors, as used in CNC systems, vs stepper motors, which of course can work as a servo motor with the addition of a position encoder....but steppers, even those with encoders are typically used in applications where the primamry concern is cost.

    So in Barry's review, the type of servomotor used, will indeed be the major difference between the Bodnar, AF and OSW. If you could replace the stepper on the AF, it will certainy be the best solution, as the SW on this is far superior to what 'OSW' or Bodnar is using.....if we do indeed need/want that SW, is another topic for another thread, but imho, the best DD wheel out there would be SimXperience SW, IONI/Argon controller and AKM servomotor.

    In these applications, steppers may indeed work ok, but it will need a significant effort on sw side, and at the end, it doesn't matter how much you invest in sw, it will not keep up with a high quality servo motor.

    Anyway, there' s a product to fill every gap in the market now, and all DD wheels are so far above the rest, it doesn't really matter which one you get...the step-up will be enormous. If you are lucky enough to own one, whatever brand/model, thank the sim-racing gods.

    Cheers,
    Beano
     
  11. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Yes, your right. I quite wrongly thought they were induction motors at the time. I don't know/remember if i clarified that i had made a mistake in calling them induction motors in a later post.

    Sure.

    So if we looked at it from the type of motor (i.e. servo vs stepper) perspective only, holding all other variables (e.g. software, motor driver/controller, motor peak torque and power, etc) constant, do you think there will always be a noticeable difference in ffb performance because of the motor type?

    I ask because Led566 wrote a few posts back of how the IONI/Argon drivers have very tightly current controlled due to its closed loop current controller (and presumably the bodnar as well) which may explain why they have such incredibly fast torque responses and consistency as demonstrated in some of the RFR Wheel test response graphs. In contrast, the AF Pro's RFR Wheel response test does not look anywhere near as good which seems to confirm the observations made by people (for example those in Barry's DD comparison video) comparing the AF Pro to the OSW and bodnar wheels. They all described the same problem with the AF Pro vs the OSW/bodnar, that it felt less responsive/immediate in producing ffb details.

    Also, since maximum torque production begins to drop quite drastically for steppers past a relatively low speed (compared some permanent magnet motors that can sustain rated torques up to ridiculously high speeds) and i'm curious if that is partly the cause for the observed difference in ffb responsiveness of the AF Pro vs bodnar/OSW wheels.
     
    Last edited by a moderator: Dec 7, 2015
  12. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    To add, i found it interesting that the RFR Wheel test results of the AF Pro looked just like the T500's whilst the G25's was incredible (given what it costs and what it's made of) The G25's result is not as good as the OSW's ofc but it is so much closer to the OSW's than it is to either the AF Pro's or T500, by a very considerable margin. What i find interesting about this is the fact that both T500 and G27 are permanent magnet motors (albeit brushed) and yet the difference in responsiveness between these two wheels in the RFR Wheel test looks as big as the difference is between the AF Pro and OSW/bodnar wheels.

    Which makes me wonder if the significant difference in responsiveness performance (based on RFR Wheel test results) could be due to something more than just the type of motor employed. For example, could it simply be down to the tight current control in the driver (or perhaps lack-there-of in the case of AF Pro)?

    Just throwing ideas out. Looking forward to hearing what you think.
     
    Last edited by a moderator: Dec 7, 2015
  13. Beano

    Beano Registered

    Joined:
    Jun 9, 2015
    Messages:
    29
    Likes Received:
    0
    My honest impression is that it all boils down to ones own expectations, vs your wants, needs and spending power.

    We are talking rather small differences here, when comparing DD to non-DD wheels - the purist and DIY'er will most likely always go for servomotor (at least how we define it) as opposed to a Stepper, simply because you will need a rather complex control software to make the best of the hardware limitations. The AF is a good example of just how good you can have it with a stepper, it is testament to the SW and how well that have been programmed.

    For any business wanting to maximise on their ROI, this is the way to go, but, there is also much to be said for how little time was spent on the DIY side, and for it to be actually very competitive in this environment. Again, we are splitting hairs here, comparing a Lambo vs Maserati vs Ferrari, so it will all depend on the above: Wants, vs Needs vs Cashflow....

    I am a DIY'er and will always go for the DIY option, merely because I enjoy playing with things, and how to make them work......for many, the AF will be the most affordable and easily obtainable solution, whilst for a select few, the Bodnar will be the best option.

    One note though: There will be a "Best" option out there for everyone, it doesn't generally mean it is THE best option, but what matters though is that it is YOUR best option.

    Go and enjoy, I am really happy DIY DD FFB is driving improvements both over at Bodnar and AF side, with much more to come from the DIY stable :)

    ps: To answer your question though: With all else being equal, there always will be a gap in ultimate performance of a Servo vs a Stepper (as we define it) - there is only so much you can do to overcome the inherent limitations of a Stepper design - but, even in Servo world, the 10-pole AKM servo, as used by Bodnar, is significantly better than the 8-pole designs used in the Lenze and MiGe, as well as other servos. If you have money to burn, buy the best....otherwise, we are privileged to have many other great options, from DIY to AF to non-DD wheels....

    Cheers,
    Beano
     
  14. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Sure. However i'm curious if that inherent limitation should even be observable at the sort of speeds and performance envelope demanded of a motor by sim racing ffb. From the way you answer, i take it you believe it should be observable and is in fact the very reason for the performance difference.
     
  15. Adrianstealth

    Adrianstealth Registered

    Joined:
    May 28, 2012
    Messages:
    4,490
    Likes Received:
    1,004
    seems most of us have a different opinion lol

    anyone else with another version? ( don't forget the graphs )

    if your interested in writing a book on it -it's the people that work in industrial robotics are the ones to listen too trust me (unless one has other intentions of doing something else with all this type of information ......)

    my pals OSW due tomorrow ( had reply from Ollie ) hopefully have it up & running for him quickly & in times for iracings new build this week, servo happyness coming his way!
     
  16. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Here's a look at the AF Pro vs OSW (with mige) response from RFR Wheel tests Beano.

    AF Pro:

    [​IMG]

    vs

    OSW (with Mige):

    [​IMG]


    and then here's the T500 vs the G27 for comparison (both using brushed motors)...


    T500:

    [​IMG]

    vs

    G27:

    [​IMG]


    Ofc, the difference between the t500 and g27's result could simply be due to construction. But i thought at the very least, it might serve as reason to suspect that the AF's difference in responsiveness vs the OSW/bodnar's is due to more than just the motor type and/or construction. I'm perfectly happy to accept it's simply due to the motor type if so, just want to be as sure as can be leaving no stone unturned.

    Sure but some of us would like something more than just opinion, like objective conclusion(s) based on facts if, when and where possible. Hence some prefer a deeper discussion into the subject area in question.

    I also agree (generally speaking) but when you stated that the bodnar wheel is a servo system (i.e. operating as a servomechanism) based on the information you recalled from some friends who work in the robotics field, it caused ones eyebrow to rise given that ffb in racing sims only send torque demands to a motor and never position/velocity demands. The mode of operation of ffb between race sim and ffb motor is the same for all ffb wheels as far as i can tell/know.

    If a motor responding to torque demands makes a "servo system", then wouldn't that make every ffb wheel by definition a "servo system" too?
     
    Last edited by a moderator: Dec 7, 2015
  17. Beano

    Beano Registered

    Joined:
    Jun 9, 2015
    Messages:
    29
    Likes Received:
    0
    If you have the different DD wheels side-side, and jump from the one to the other, there is a clear difference in performance, not only from the torque perspective, but also speed-of-response and FFB precision, resolution, or, fidelity, however one wants call it. But, if you had to test one wheel today, another a few hours later, the last the day after, the difference is not so clear.....

    I will always pick the DIY version over anything else, as by nature, I am a tinkerer...for others though, a different selection would be best suited......so I am very happy with the different incarnations of DIY DD wheels I have, I have a Lenze with 16k PPR encoder, both small and Large Mige, one with 2.5K PPR and the other with the newer 5K PPR encoder, as well as a newly arrived AKM53 with 10k PPR encoder...I am currently assembling an IONI-based drive controller, sporting the very first IONI Pro HC to be available.....this will drive the AKM, and I have a 2nd Pro HC, which will drive my Lenze...

    I have enough spares and parts for another 8 units, tbh, so I am all for DIY, trust me.....but, I also realize, the observable differences between the best and worst DD FFB wheel is rather small, compared to coming from non-DD to DD FFB...

    Hope this makes sense...

    edit:

    In actual fact, yes, as a loose definition, it would.....although, strictly speaking, and as you have mentioned, we are sending PWM/DIR signals to the DIY and Bodnar wheels and steps/micro-steps to the AF......so none are in essence close-loop systems, as we control the current to these motors to determine the force-feedback amplitude, and a simple DIR signal to change the direction - I guess one could always introduce a torque-sensor, but it would be an expensive solution to address a non-existent problem...

    I can see the benefits of having velocity and position sensors for those applications, but a torque-sensor for our application would be well over-the-top.....and won't improve anything tbh...

    Cheers,
    Beano
     
    Last edited by a moderator: Dec 7, 2015
  18. Joe

    Joe Registered

    Joined:
    Nov 14, 2013
    Messages:
    372
    Likes Received:
    21
    thanks DrR1pper for posting those charts. They tell the story all.
    I will sell my AF wheel......
     
  19. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Thanks for the insightful reply Beano.

    Quick question please...do you notice a difference between the 16k, 10k, 5k and 2.5k encoders?

    Understandable. OSW next?
     
    Last edited by a moderator: Dec 7, 2015
  20. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,583
    Likes Received:
    101
    If we talk about PPR, then Lenze has 4096PPR encoder. Older "small MiGe" 2500PPR and the newest has 5000PPR. Kollmorgen might have from 1024 up to 10000PPR, depending on the model version. You can multiply PPR by 4 to get an effective resolution of the encoder. I'm curious, Beano do you feel any changes when going from 2500PPR to 5000PPR for the small MiGe?
     

Share This Page