The online racing simulator
Searching in All forums
(635 results)
jtw62074
S2 licensed
This topic has never come up when discussing vehicle dynamics modelling with other people that also do this professionally for a living. And nope, I'm not familiar with frequency domain solutions so I may indeed be missing your line of thinking. I've just got a vague idea of what you're talking about.

Anyway, I'll continue to wait for a specific vehicle modelling calculation as an illustration of your point to show us all what we're missing.
jtw62074
S2 licensed
That's still way too vague of a description/idea to come to any conclusions about sampling rates. What are the outputs of the function? What exactly does it do? What if it's a closed form system that is not time dependent? You can't really get all into a discussion at that level without breaking out the math and getting into the nitty gritty ins and outs of the whole system. As I said before, this is just a bunch of handwaving really and not meaningful at this level at all...
jtw62074
S2 licensed
How can you reach that conclusion while having no idea what the fractal model entails or what specifically it would be used for?
jtw62074
S2 licensed
Maybe so, Loopingz. I'd like to see whatever you can dig up. However, I think those calculation errors are very small indeed compared to the tire model errors in probably every case ever done to date.

Anyway, for sure you can't tell one bit of difference in the handling whether the tires in VRC run at 600Hz or 30000Hz, other than the frame rate drops a bunch. Which is really more accurate? I couldn't tell ya and am not concerned with it. I've changed the frequency of VRC wildly during testing, not told anyone about the changes, and nobody ever noticed a difference, so I'm not too worried about it although it's probably a valid point, however minor in this case.
jtw62074
S2 licensed
Quote from Shotglass :
the damping you were talking about ... a piece of rubber cant possibly jump from being deformed fully in one direction to being deformed fully in the other one its characteristics will always create a smooth transition with an upper frequency depending on the type of rubber ... and to model that you need a sampling frequency depeneding on the upper frequency of your rubber

These are all pretty vague generalizations without mention of anything very specific though. I.e., can you give an example of a specific calculation you have in mind?

If you're talking about calculating pressure for instance on either side of an asperity, and you're looking at a damping/hysteresis effect, well... Damping force is function of velocity. It's an instantaneous, analytical calculation. And again, if you're talking about moving a block of rubber step by step, the higher the sampling rate you use (smaller time step, more calcs per second), the closer you'll get to reality. I guess I still don't understand where this "threshold frequency" is that your referring to, at least not with any specific example. I've read nothing about this anywhere in any papers/books on tires or vehicle dynamics at all.
jtw62074
S2 licensed
Quote from Woz :Slicks are less forgiving because of the shape of the curves. Its all in how the transition takes place. In the same way low profile road tires are less forgiving that tires with a bigger sidewall. The more gradual the transition the easier it is to feel the change in state and so the easier it is to catch.

Right. Another thing too is that slicks generally, although not always, peak at lower slip angles than street tires do. It's quite a bit easier to toss a car around if the peaks come in at larger angles.
jtw62074
S2 licensed
Quote from Shotglass :well if you restrict your calculations to a certein update frequency what youre doing is sampling in the time domain
now rubber with its low pass characteristics cant deform at any speed which means that there will be a frequency above of which the error wont decrease much more

What low pass characteristics are you talking about here? This seems like just a bunch of hand waving to me at this point.
jtw62074
S2 licensed
Quote from axus :If I may just go back to the tyre 3D modelling using displacement maps thing, AFAIK part of the reason why road tyres are more forgiving and regain grips quicker than slicks is because twisting several small patches of tyre is easier than twisting one big tyre patch, not just the fact that you have a smaller contact patch for the same total area of tyre. Is this correct? And how would a tyre model, such as yours, go about working around that if so?

"Regain grip quicker?"

http://www.performancesimulations.com/files/tire1.JPG

Where do these tires (except on the wet) ever "lose grip" in the first place? They don't. This is a myth that just won't die. They just rise up to a point and stay there. And slicks look pretty much the same. The curves for every tire are just a bit different. Some rise faster to their peaks and roll off a bit more suddenly than others do, but this appears to have a lot more to do with the cord angles and construction of the tire itself than the tread. The presence of tread pretty much just lowers the cornering stiffness (so the peak comes in later), but you can compensate for that with the construction to put the peak pretty much wherever you want it. I.e., if you saw more graphs like this you'd have no way of knowing whether you were looking at slicks or treaded tires, other than with a slick the peak might be much higher.

A tire feels "forgiving" when the force curve levels off gradually over a large slip angle range like the bias example in the link above. Your street tires work this way because that's how the tire engineers designed the construction. Ease of control is one reason. This is influenced much more by cord angles/layout than anything else.

When people feel their tires "lose grip," they aren't really losing any grip at all. It's just that the rate at which the force is increasing with slip angle is leveling off. It's a bit like people thinking their cars actually speed up once they hit the grass. Nope. They've lost acceleration, but the acceleration hasn't reversed.

When you're looking at a side force/slip angle situation, the tread's not so much twisting as it is being pulled to one side.
Last edited by jtw62074, .
jtw62074
S2 licensed
Quote from Shotglass :maybe its just because nobody ever cared much about the frequency other than trying to reach a high one
but the model shouldnt really have any influence on this

You might want to research tire modelling online a bit. I think you might find this is sort of irrelevant.
jtw62074
S2 licensed
Quote from Shotglass :hmm i wonder how much you could gain on that by using an fpga programmed to do as many elements a possible parallely

No idea. Still, you'd have to do it 28,800 times faster than that to run all four tires
jtw62074
S2 licensed
Not sure. I suppose that might depend on the model, but I've never read any mention of this in any tire literature.
jtw62074
S2 licensed
Can you think of a specific example or case? I'm not quite sure what you mean there. Do you mean something like a dynamic FEM type of situation? If so, I don't know. Haven't seen anything on it or given it much thought really.
jtw62074
S2 licensed
Quote from Shotglass :btw concerning this ... is there any reserch on the frequency response of a tyre to find the minimum frequency for a tyre model to be correct ?

How correct? Nothing's perfect generally. Although you could say that if you used any model at all and it output the same forces at any given condition that a real tire did, it's correct regardless of the frequency you're running it at. I.e., 5 deg slip angle for tire X at 1000N load makes 600N. As long as your model outputs that same number, it's right, even if you calculated it only one time.

Generally the more you crank up the frequency the closer you get to the "right answer," so it's more a matter of how close is close enough.
jtw62074
S2 licensed
Quote from axus :Well... that's the type of simulation I was thinking of using interpolated lookup tables for rather. I'm no fan of the ones available at the moment but I think that as long as your data is accurate and consistent (which seems to be most of them's problem - the lateral behaviour doesn't match the longitudinal behaviour, or it feels like that anyway... you can almost tell the engine is canned if you know what to look for) it might work. Get sim racers to use their spare CPU cycles to create the lookup table and send it back via the internet a la SETI@home and voila! You get a few percent closer to modelling tyre physics.

Oh, sure, that's always a possibility. I've always thought it would be cool to run an FEM model, then just use Pacejka's Magic Formula to match it up as though it was a real tire test. Then you could essentially run the steady state FEM model in real time in the sense you're outputting the same forces the FEM model produced in the first place.

One of the tire models I wrote a few years back modelled each contact patch with about 50 2-D springs, so it was a sort of super-simple wannabe FEM model, I suppose (and didn't run particularily fast either). That worked pretty well actually, but with the calculations I was doing it dawned on me that the whole thing could be done with a little calculus much faster and you'd effectively have an infinite number of springs.
jtw62074
S2 licensed
Quote from axus :Any plans to model this?

Not at the moment. If you're going that detailed you might as well be using FEM. That's not quite possible yet though. Just saw one online that takes two hours to run a single second's worth of simulation for one tire on a 3.4Ghz workstation, so that sort of thing is a ways off yet. You could probably do some cheaper workaround that sort of emulates the behavior though, I suppose, but for the most part I think with tread you're mainly reducing contact area. If you want to model all the tread blocks squirming around and so on accurately you'd need FEM. And with the unknowns in that you might not be any more accurate than if you skipped it entirely and just tweaked the cornering stiffnesses and so on to match up with the basic tire type you're doing anyway. I.e., the exact rubber properties and so on aren't going to be known anyway. Garbage in, garbage out.
jtw62074
S2 licensed
Quote from Shotglass :wouldnt the rubber also bend around the fractal pattern creating spots of upwards forces where the rubber sliding across the fractal deforms when it slides towards a peak ?

Yes. The hysteretic friction component is primarily this that's occurring. If you slide rubber across a rough surface (imagine just one protrusion), on one side of the protusion you have higher side pressure than on the other because the rubber acts a bit like a damper (since it's viscoelastic) and doesn't immediately "snap back" into position and apply full pressure on the back side of the protrusion. It's a bit like a car moving through air or a boat through water where the pressure is greater at the front than the rear and causes an opposing force. In that analogy the air is the rubber and the car is a surface protrusion.

That's only the hysteretic (damping loss) part of the friction though, which is relatively minor compared to the adhesive, "true surface area" part of it. Something like 5-15% depending on the rubber compound.

But yes, you'd have vertical pressure variations on a microscopic level throughout the whole nominal contact area.
jtw62074
S2 licensed
Quote from axus :Good point also. Todd, about your model, does it use a tyre displacement map or something of the sort to grooved tyres correctly?

Nope.
jtw62074
S2 licensed
Quote from axus :Go on an autocross or skidpan and swing the car into understeer. You feel heavy vibrations through the wheel and you can hear knocking sounds (not the most pleasant sound you'll hear your car make). If you look at the tyre it is bouncing (or maybe loosing grip and regaining it within instants, but I can't explain that one with anything apart from it bouncing). Considering this is a flat surface, I'm struggling to explain this bounce otherwise... although come to think of it, perhaps the limitations of having an anti-roll bar and hence not entirely independent suspension :doh:.

Oh, ok, I understand. I don't usually use FFB except for the spring centering effect so never noticed it. Indeed, the tire is essentially a spring, so if you're getting some bouncing that's probably why. Wouldn't really be because of surface roughness so much I think.
jtw62074
S2 licensed
Quote from axus :Hmmm. I never actually knew what causes those bounces of the tyres. I really wasn't sure how significant the effect I was talking about is. I couldn't come up with any other explaination for the bouncing though. As for the two curves, its just a vertical force exerted by the tyre (hmmm, come to think of it, as much as... 30N at most with extreme slip, if even?) due to high slip and the effects of road graining. I suppose some investigation and testing would be appropriate. I doubt I can get a tyre to slip enough to measure this with my bare hands though.

The vertical force at the tire is for the most part going to equal whatever the load is on it. I.e., whatever portion of the car's weight it's supporting.

I'm not sure what this bouncing you're referring to is though. Keep in mind the entire tire is a flexible, springy type of thing, so maybe that is causing whatever you're picturing to some extent?
jtw62074
S2 licensed
Quote from axus :Um... true.
Note to self: Think before you post... and word your sentences carefuly.

What I'm talking about is vertical forces caused by small unevenness of the road surface - road texture. You were talking about using fractals to represent the road surface accurately. What I'm talking about is if you look at each of those tiny little bumps and imagine the small bits of tyre pushing against them. This would cause a vertical force, making the tyre bounce at high slip - as seen with the front tyres in understeer. I guess this could quite easily be mapped by two curves - one for vertical force against lateral slip and one for vertical force against longitudinal slip (combined by Pythagoras' Theorem I guess).

EDIT: Changed the first post a bit to make more sense.

The little grains and roughness in the road don't cause the tire to bounce, do they? I haven't seen it happen and wouldn't expect them too. They're downright tiny. All you're really getting there are localized high and low pressure areas at the road/tire interface and hysteric effects from the vibration in the rubber as it moves across the pattern. The surface roughness/pattern influence the real area of contact too, which is an important part of the picture.

As for the mapping of the two curves, I'm having a tough time visualizing what exactly that would accomplish.
jtw62074
S2 licensed
I'm not really sure what you mean there. Are you talking about bumps? In reality, even on a very smooth surface at a constant slip angle (there is no such thing as understeer to a single tire, there is only slip angle/slip ratio) the lateral/longitudinal forces are rather noisy indeed. The graphs you see with the nice, smooth line are averaged values, not the pure data. You don't lose anything by using that though as if the force jumps from one time step to the next like this: 100, 110, 100, 110,100, 110, you'll get the same results as if you were using 105 all along, so I wouldn't get too hung up on the high frequency/resolution stuff.
jtw62074
S2 licensed
Quote from axus :Do you also have some form of surface displacement map also? That you can vary using things like surface roughness and density of uneven parts (just making up my own terms here but what I mean is how many protrusions there are per area of road and how much they protrude)? I guess that to simulate tyre behaviour accurately, you'd need to consider the road also.

Not yet, but I plan on it. Road surfaces and the resulting rubber friction can be modelled pretty accurately on small scales with self affline fractals, so you'd be able to change the texture of the road and it would effect the grip, temperature, and so on.
jtw62074
S2 licensed
The "steering while stopped" behavior is due to torsional stiffness. Here's data from a couple of tires measured where the tire is stationary with some load applied, then twisted through different angles:

http://www.performancesimulations.com/files/tiretorsion1.JPG

This sort of works a bit hand in hand with aligning torque in reality, but is not typically very important or noticeable at anything approaching racing speeds since it really has more to do with how quickly you steer the wheel than what the slip angle is. If you were doing a driving simulator for very low speed driving it'd be a nice touch to include, but the aligning torque when the wheel is rolling is much more important.

In LFS and most other sims I just use spring centering FFB, so don't have any LFS specific comments on this.
jtw62074
S2 licensed
Quote from axus :Very impressive indeed. I presume its in that Virtual RC simulation? Any plans for putting it in anything else?

Yes, that's in Virtual RC Racing. One or another variation of that same approach has been there since probably 2001. I use the same basic model (but beefed up a bit for better camber and other effects) in my full sized car sim:

http://www.performancesimulations.com/files/ToddSim13a.wmv
http://www.performancesimulations.com/files/ToddSim10a.wmv
http://www.performancesimulations.com/files/ToddSim11a.wmv
http://www.performancesimulations.com/files/ToddSim7a.wmv <--Drag car. This was using a much simpler model.
jtw62074
S2 licensed
Quote from loopingz :Hi magic, remember that raising Hz has limit because with higher Hz you multiply the computation (and model) errors. There is a limit where accurracy is coming down with the hertz growing. Anyway I don't know where is this limit, it is hard to calculate and to test but possible.

Not sure what you mean there, Loopingz.. Higher frequency sampling is a way to increase accuracy rather than decrease it. One of my earlier tire models for VRC was running at 30,000 Hz at one point. That was way overkill though but it did run ok in real time. The current VRC tire model runs at about 600Hz IIRC.
FGED GREDG RDFGDR GSFDG