• Tyres... The Big One. (Part III) - Realtime Model Batch Tests

    As you will no doubt be aware, a lot has happened in recent weeks. I would like to assure you all that despite the change in management, the blog will continue into the future.

    Next up, to an answer to a question from the floor. I mentioned last time, that we basically optimize the tyre for their state that they see in corners. This is because the majority of lap time is gained through corners. You spend a significant portion of time cornering, corner apexes are also the slowest speeds you see on the lap, so you spend a good portion of time there. Higher apex speeds also mean you spend less time braking, even though prioritizing braking would allow slightly later braking. This is all a generalization, but cornering is predominantly the most important way to gain lap time.

    With that out of the way, this is the last major part covering the tyres for the Brabham itself. Although we'll have one more post covering the finishing touches. Specific aspects of tyre design will also be covered in the future too.

    Now, continuing from last time. We have our basic tyre created, its basic geometry is done, a lookup table has been generated. How do we tune, and improve the tyre further? How do we know whether the tyre correlates well to a set of data? Data that we either have in our hands, or sometimes as an abstraction in our minds (if we're unfortunate to lack hard data). It's time to get plotting. To help with this, we have a new spreadsheet to generate realtime batch tests to be used for tTool available here.

    Realtime Batch Tests
    To begin with, realtime batch tests provide a way for us to analyze a tyre over a given range of conditions and measure the outputs. Essentially, via these means, we're able to create slip curve plots, rolling resistance plots, aligning moment plots, deflection tests or just about anything else you might desire. The spreadsheet is simply a way to help generate these test cases. If you used the QSA Batch Test spreadsheet, then the realtime batch tester should feel somewhat familiar. I have tried to be somewhat descriptive within the spreadsheet, so be sure to check out the comments. To help get started, on the first sheet '1 General', you'll see another 'Information Desk:' describing some of the basics. Still staying with the '1 General' sheet, you should specify some basic data on the left side. The most important are fields are highlighted in orange 'Approx. Radius' & 'Max Load', among others, which are used to define the starting point of the tyre dimensions. They also define test boundaries to make the tests run quicker and keep them more relevant.

    Next is to tackle the 'initial tests', named as "1 Test-Initial" in the spreadsheet. In similar fashion to the QSA Batch Tests, we need to collect some data from the tyres before we can generate specific tests optimized for them. Of course, realtime batch tests are performed MUCH faster than are QSA tests. A single QSA test can take between a 30 seconds and an hour (or more in certain extreme cases), realtime tests typically take a only a couple seconds each. So this should be done rather promptly. It will look similar to the following:


    A quick look inside the spreadsheet.

    So what do all these things actually do?
    Start - this is just defines a starting value, or base value to work from.
    Step - Essentially, this is the linear value increment as for every step.
    Stepē - This is an increment based on the square of the Nth step value.
    # of Steps - The number of total steps before resetting itself and starting from base again.
    Step Count - How many times before lines we go before we actually consider the step to be an increment.
    Disable Before - Disables the value BEFORE this Nth step. If you have a 'Fall-back' value defined, it will use that instead of leaving this blank.
    Disable After - Disables the value AFTER this Nth step. If you have a 'Fall-back' value defined, it will use that instead of leaving this blank.
    'Fall-back' - A fallback value to use if the step doesn't meet any criteria or results in an error.

    For the initial tests, the pre-loaded values are set to recommended specifications, as such, if you filled the '1_General' sheet correctly, it should not require any alterations. However, the spreadsheet remains open to edit as you wish if you feel you'd like to do things differently.


    Vertical Deflections
    You should have already adjusted the tyre correctly to make the vertical deflections match some set of expectations via the geometry model (QSM, quasi-static model). Nonetheless, you may consider testing them (your tyre) within the realtime model as a verification or to aid in setting up the .TBC tyres to correlate. You can also alter the 'LoadVsDeflectionMultiplier' variable in the realtime section of the TGM file, which when increased, artificially makes the tyre stiffer vertically, while values less than 1.0 will soften the vertical stiffness. If you need to use a value significantly divergent from 1.0, you should consider an alteration to the design your tyre construction properties. It is clearly easier to do this while you're early in the design phase. In other words, it would be wise to ensure your tyre delivers accurate results in the quasi-static analysis section of ttool, prior to building entire lookup tables. Lookup tables may take 12-36 hours to generate, depending on your hardware, the detail of your tyre itself, and the number of test permutations of your lookup table.


    Rolling Resistance
    If you're satisfied with the vertical stiffness, the next thing to move onto is rolling resistance. All tyres produce some rolling resistance. This is primarily due to energy dissipated by the tyre tread through a process known as hysteresis. Essentially, kinetic energy is lost through resistance to deflection, converted to the form of heat energy. Therefore the amount of deflection plays a role as well as visco-elastic effects (essentially damping). Generally speaking, tyres with greater hysteresis also have higher grip. This benefit will be greatest on surfaces with a rougher road texture, but also wet surfaces where the opportunity for chemical/molecular adhesion is vastly reduced. In the end, rolling resistance is one of the first things to adjust. It is also relatively simple to alter. The fact that it has a direct bearing on other tyre attributes also gives it precedence. When creating the tyres, you will have run with some placeholder values. You will then need to alter them based on the results from such tests.

    Damping is controlled by the following variables:
    BaseDamperPerUnitArea=(<lateral>, <vertical>, <longitudinal>)
    HystereticVerticalDamperPerUnitArea=<vertical>
    VerticalDampingMultiplier=(<bristle1>, <bristle2>, <bristle3>, <bristle4>, <bristle5>, <bristle6>, <bristle7>)

    The vertical damping multiplier refers to individual bristles in the cross section. From left to right, we have 7 of these 'columns'. The reasons the individual columns are adjustable is mainly to do with the fact that some distortion is partially ignored in the edges of the tread. The calculation is based on how much the tyre has been deflected, while this is fine in principal, it ignores some secondary distortion and flex, so this is a compensatory mechanism.

    If your tyre's overall rolling resistance is about half what it should be, you'd roughly need to double the BaseDamperPerUnitArea (middle value) and HystereticVerticalDamperPerUnitArea. If it's too high, you'll need to reduce it by roughly the proportion that it's too high by. If you need more at lower speed, and less at high speed, you would increase HystereticVerticalDamperPerUnitArea and decrease BaseDamperPerUnitArea. As I touched on previously, you will need to consider that at higher deflection, the bristle is moving though a greater range and this will increase rolling resistance.

    For the Brabham BT44, the expected rolling resistance coefficient is approximately 0.015 for the front, and 0.018 on the rear tyre (using the TBC values as the basis). From the previous blogs, we also know that bias ply tyres have a predisposition to a higher rolling resistance than do radial tyres. Furthermore, racing tyres place a greater emphasis on grip, than they do rolling resistance. As such, these tyres generally use a rather high hysteresis compound. So these particular tyres will have just about the highest rolling resistance of those used for any car. For example, we might expect around 0.007-0.01 for a modern radial street tyre, and 0.009-0.014 for a modern racing radial. It should be noted that these numbers are certainly not a one size fits all, but the vast majority of tyres will fall within this realm, at least under normal operating conditions (and on a hard surface). Rolling resistance also varies with load, speed, and temperature. As you increase load, deflection rises, and with it, rolling resistance does too. As such, it's a good idea to test under multiple relevant conditions.

    One final thing to note about rolling resistance, is that the latest TGM variables, "StaticDiffusiveAdhesion=(,,) & SlidingDiffusiveAdhesion=(, , )" also slightly contribute to the overall rolling resistance. These variables effectively 'bond' the tyre to the asphalt, which in turn requires a force to separate from the asphalt, a phenomenon which also slightly extends the contact patch.


    An example of the output of rolling resistance at various loads.


    Lateral Sweep Tests
    At this point you'll probably be eager to see where the peak slip angle is, and the overall shape of the slip curve. These can be performed with the tests provided by 4_FullSweep or 5_LatPeaks. There are also preset longitudinal tests provided @ 6_LongTest. Finally, you can try the '7_Combined' tests to generate a sort of 3dimensional table.

    What do lateral tests look like? My own data for the BT44 tyres looks something like these:


    Bias-ply tyres typically exhibit a wider peak slip angle, while more modern radials peak slightly earlier with a slightly larger falloff. Performing these tests will measure the data necessary to capture these characteristics in your own tyres. Of course, the lateral and longitudinal stiffness data for the tyres are ideally determined by the QSA batch testing spreadsheet I introduced last time. Those tests provide a good representation of the high detail model into the realtime model. Thus, if you find that the tyre ends up having too tight of a slip angle or too loose, you would ideally reconstruct your tyre with stiffer or softer belts, or bias-plies or whatever you feel might be the best way forward. The discrepancy between what you intended and the actual results essentially form the basis of how much you will need to alter the construction by. All else being equal, the peak slip angle is mostly dictated by the stiffness of the tyre in the tread area. Hence bias-ply's exhibit a wider peak slip angle than do radial tyres. Rubber stiffness and properties also have some effect and obviously temperature effects are not to be forgotten here. Something to note that might not be intuitive, is that at higher temperatures rubber will grip more optimally at lower sliding speeds, than at low temperature, which will find grip at higher sliding speeds. The flip-side is that the rubber can actually distort more at high temperatures as the compound softens. So the actual sliding speed at the surface may actually reduce as the temperature increases. For these reasons, there's no better substitute than testing the realtime model with the batch tester to establish these properties.

    The sky is basically the limit when it comes to realtime batch tests, so feel free to test whatever you need, in whichever way you feel is best. For example, there are additional pre-configured tests (6 & 7) covering longitudinal slip or combined curves. They operate in a similar manner and are just basic examples of what is possible. The only difference is that for the longitudinal tests to be appropriate, we need to run an additional longitudinal slip offset value. Essentially, the purpose of this is to correct the 'longitudinal slip' to provide 0 longitudinal force at "0% slip".
    This article was originally published in blog: Tyres... The Big One. (Part III) - Realtime Model Batch Tests started by Michael Borda
    Comments 10 Comments
    1. MaD_King's Avatar
      MaD_King -
      Thank you.
    1. Slamfunk3's Avatar
      Slamfunk3 -
      Thanks Michael
      TK
    1. Eddy's Avatar
      Eddy -
      Again no driving tonight but reading 😉 Thanks Michael
    1. Helio Zuazua's Avatar
      Helio Zuazua -
      Superb
    1. TheGame316's Avatar
      TheGame316 -
      Thank you Michael. a couple of questions at the moment though.
      1. Following the Info Desk on the spreadsheet. Each time you Copy and paste test values to custom_realtime.ini. Do you overwrite the previous tests. eg Perform initial test, then when copying def test values do you overwrite the initial test values.
      2. Info Desk 2C when you say relevant parts, I am not 100% sure what you mean but have a feeling it may relate to my first question, could you please elaborate.
      3. Info Desk 3A you state this should be adjusted prior to running other tests. Does this mean if changes are made here, that I should go back and redo the initial and deflection tests before continuing onto the 4-Test sweep.
      4. Could you elaborate on the static and sliding diffusive adhesion values please.
      5. In relation to multipliers, you state a warning regarding the use of values significantly divergent from 1.0. What is considered significantly divergent?
      6. With quite a difference in the information provided in the tgm quickstart guide and what is in a current tgm file, is there a plan to update the guide.


      Do you think it would be possible for you to make a video showing how to do the basic tests as outlined in the spreadsheet. Seeing something being done, rather than reading can help get the information across clearly. Also considering that the vast majority of today's modders do not have access to complete tyre data, some form of troubleshooting guide would also be helpful. eg. You are wanting to change a certain characteristic, look at x values but be mindful that changes here will affect y values.
    1. Michael Borda's Avatar
      Michael Borda -
      1. Yes, you overwrite this file, but this file is really just a batch command to tTool telling it what to do. You are 100% free to rename this file as you wish, e.g. "custom_realtime1.ini",...2,...3 etc, if you prefer this method. You'll just have to enter the correct name in ttool.

      2. Well, the thing is, new tests from ttool are appended to the existing .csv file, until you restart ttool or change the filename of the .tgm. So in certain cases, you'll need to be careful about what you paste into the realtime tTool batch test generator. For example, after completing the first 4 'initial' tests, the first 4 lines (5 if you include the headings) in the "CustomRealtimeTable.csv" file will be those initial tests, which have to be ignored for the other tests that you will have done.

      3. I simply meant that it should be done prior to lateral / longitudinal tyre tests, etc. They are actually in the correct order. You shouldn't need to rerun any previous tests if you change the damping values in the tyre.

      4. I covered static and sliding diffusive adhesion in a previous blog entry. With it, in theory, we could simulate a gecko sticking to the ceiling (or if you put some really crazy values, a whole car). If for example, you had a contact area of 0.01mē and you had the maximum adhesion force (2nd value), of say 10000, under optimum conditions it would bond to the ground with an assumed force of 0.01*10000 = 100N.

      5. Significant in this case would be more than about 15-20% in one direction or the other. So a reasonable range for this value would be 0.8-1.2, but I wouldn't take this as an absolute. Certainly the closer to 1.0, the better. In any case, if you have a working tyre, and need to use 0.6 or something, then you're still better to keep that value at 0.6 until you have a new working version of the tyre with a new lookup table completed.

      6. Indeed the old guide is a bit lacking in areas. It's something I'd like to get around to doing. My intention is to release some minor fixes and improvements to the spreadsheets that were released previously, first. So yes, but it might not even be ready this year.

      The video idea is something I've been wanting to do for a while. I need to look at my options but it's certainly something I would like to have a go in the short term.
    1. Feralarri's Avatar
      Feralarri -
      Hi Michael Is there poss an error in 3 Test-Crr or am I doing something wrong here?

      When I paste results from this, the LongVel value is the same for all tests whereas your results sheet shows an increasing velocity.

      When you look at the 3 Test-Crr tab it shows the LONGVel at 13.89 for all tests .
    1. Feralarri's Avatar
      Feralarri -
      -
    1. Michael Borda's Avatar
      Michael Borda -
      Sorry, the vBulletin software on the site is kind of iffy in some ways. I didn't get notified. I'm not seeing any attachment to your post, but I think I know the issue of which you speak.
      There's a kind of logic bug in the spreadsheet.

      To work-around it, simply delete the 0 from cell B2. I'll try to get an update done soon.
    1. Feralarri's Avatar
      Feralarri -
      Thanks Michael

      I have worked it out I believe. B2 should be zero, B3 I changed from 0 to 1 and I changed B5 (# of steps ) from 510 to 6. This then caused the longvel to repeat its pattern every 6 steps. I have attached a before Attachment 20975

      and after
      Attachment 20976

      shots showing the cells changed in pink
  • Recent Forum Posts

    Naiba

    Design a ship

    An update on the bridge for people not keeping an eye on discord. The old bridge didn't make sense under constant-g thrust, so the technician stations

    Naiba Today, 01:43 PM Go to last post
    Lazza

    rfactor 1

    Ok, yep, that's the full one. When you first load up the game and it asks for your name etc, once you've selected a series (and a car?), you can go straight

    Lazza Today, 04:50 AM Go to last post
    alexis miaritis

    rfactor 1

    The size of the file is 1.15 GB

    alexis miaritis Today, 02:13 AM Go to last post
    Lazza

    rfactor 1

    Did you download and install the full version of rFactor? Or the lite version?

    For me the full version lets me choose a season (fresh profile).

    Lazza Yesterday, 06:49 PM Go to last post
    Neo

    Use more than 2 GB RAM

    You were right. With larger grid of cars the game started using above 2 gb. So the patch is wofking fine)

    Neo Yesterday, 05:33 PM Go to last post
Join us