I'd chalk that up more to the tire model than the diff. I do preload in my model and it doesn't make that much difference really to this sort of behavior. Not nearly so much as getting the tires running correctly.

Engine torque reaction on the chassis is about as hard to do as gravity I doubt it's canned anywhere nowadays. It's much easier just to do it properly instead.

I'd be very surprised if any sim gets gravity wrong. All you do is add a vertical force to the car once each step equal to, get this, the weight of the car. Bingo. Perfect gravity model.

Skid pad values show lateral acceleration at constant velocity. I.e., "trimmed max g." Untrimmed is nearly always a bit higher. You see the same thing in my vid where it sometimes hits 1g even though it does 0.93-0.95 on the skidpad. I don't see the point you're trying to make though or how it relates to anything here.

Quote from Blowtus :

I'm pretty sure 'real life' g's are always higher than skidpad tests...

Meaning...?

Quote from Vesa :

You'll have to take into account that that road is as smooth as a babys bottom. No bumps in sight. That might account for a bit of the extra grip.

I'm not saying LFS has it right, because it doesn't. Just pointing out a detail.

What extra grip? Erh hem... That vid is running 15-20% less than the LFS super street tires on this baby smooth surface. Would you like to see what it looks like with less grip than that? Name your number

Last edited by jtw62074, Wed, 25 Jan 2006 13:35.

http://www.performancesimulations.com/files

Right click and 'save as' ToddSim13a.wmv.

Forgive the horrendous graphics, but I've got to say the lack of g-forces in LFS is probably not at all to blame for what the OP is talking about. This vid is from my sim with a 3500 lb car and 350 HP running on tires that, get this, only pull about 0.93-0.95g on the skidpad. That's about 15-20% less grip than the LFS street tires, isn't it? And there's no downforce here at all to help things out. None.

55% of the weight is on the front tires and it's got a rather stiff differential too. The throttle is shown there so you can see where there's lots of full throttle corner exits at low speed and all that's needed is a little steering correction to keep it under wraps. I drove around for about four or five hours one night with this car or variations of it and don't recall ever accidentally spinning out when just trying to put in some aggressive, fast laps. With 500HP the car is still quite controllable on those tires. On slicks it's almost boringly easy to drive until you drop the weight to 2500lb or so. And of course there's no less g-force feel on this ugly thing than anything else I've driven on a PC :P

Methinks the difference in sims is in the tire modelling Fortunately Scawen is a sharp cookie and will undoubtedly improve this over time. I personally think the multiplayer is the best out there, although at the moment GTL gets most of my play time (aside from my own ugly sim, which gets a bit more, but I may be a bit biased )

Last edited by jtw62074, Wed, 25 Jan 2006 08:16.

I wrote and sell a program similar to CarTest called Straightline Acceleration Simulator (SAS for short), but the modelling is more extensive and it includes an engine simulation:

http://www.PerformanceSimulations.com

There's a free demo there, but its use is rather largely restricted. I regularily create and tweak engines with the program for use in my car sim. Quite fun

Quote from bLaCk VaMpIrE :

i dont know if you had physics already at school, because you would have learnt, that the size of the touching area is more or less not important to the grip. rather the weight on one tyre and how good the material fixes to the ground.

That's essentially true for two hard materials pressed together, but rubber works quite differently. Rubber grip is highly dependent on contact pressure and contact area.

Left/right dominance is an issue. Why not allow it? It's there for camber/tire pressures.

Anybody else notice a decrease in longitudinal G force when the tires are locked that wasn't there before?

Thanks, Scawen, keep up the awesome work

Quote from Bob Smith :

If you remember Tristan, Todd's tyre model doesn't use any traction circle theory (or so he says) and apparently works very well (so he says). So perhaps you don't need to know what it is (although the underlying principals are important).

Just a thought.

My model doesn't explicitily use the traction circle, but it indeed does work out to follow the theory anyway. I.e., there is a maximum force vector and adding longitudinal traction/braking force will influence the side force according to friction circle theory, although the code does not directly apply the theory. Because of the approach I'm using, the theory "just happens" rather than specifically coding for it as in my previous models (like the one in VRC).

Whether or not it's nice, you don't have to take my word for it of course. Check out what Gregor said about it today

http://www.west-racing.com/for ... amp;st=25&#entry54854

Anyway, it's obvious that LFS does indeed follow friction circle theory at the limit. You can feel it and also see it happening in the (F)orces view. Watch the forward/side forces. As you get a bunch of side force and then increase longitudinal force (either traction or braking) the side force will indeed drop. If the resultant vector was drawn it would be easier to see how it works, but basically yeah, it's obviously working according to the theory too.

I'm enjoying these conversations lately too, and am a pretty big fan of LFS as well

About the weight transfer issue, it's not likely that it's done incorrectly. 3D rigid body dynamics in itself is pretty straightforward. Essentially you have a set of equations that will rotate and move a body (a car for instance) given the forces acting on it and what the angular and linear momenta/velocities were the last time you calculated everything. These equations follow Newtons laws very closely and have been used in sims since, well, probably GPL for starters

What you're typically doing when simulating a car, basically, is finding the amount of spring compression and using that to find the vertical force at each corner of the car. Once you have all those forces and add them at the appropriate locations, you then run through the rigid body dynamics stuff to rotate and move the body. On the next cycle the spring compressions have all changed (and therefore the forces that result from them). You run through the rigid body stuff again, and repeat. If you were in a steady state corner with a certain amount of throttle/braking or whatever, the car would eventually come to an equilibrium with all the forces in check and the weight transfer exactly what it should be. I.e., the vertical forces at each tire would eventually balance all the cornering/acceleration/aero drag and whatever other forces you have. So really, the weight transfer itself is very likely just fine. You don't typically calculate weight transfer in a 3D sim, it just happens as a result of how the body moves and how the springs are compressing.

So... You can't really get the weight transfer wrong if you follow the basic 3D dynamics approach. If weight transfer doesn't feel right, there could be a lot of causes. If the weight transfer is actually "wrong," that would most likely be due to a vastly goofy center of gravity height. I doubt very much that's the case with LFS though. And of course since the cars are their inventions, they can have any CG they want

Last edited by jtw62074, Sun, 23 Oct 2005 17:32.

Quote from J.B. :

This effect is often described as friction circle theory. In reality there seem to be quite a lot of different approaches to this problem, examples can be found on the Racer homepage and in Brian Beckman's PHORS series. I don't really know what the reason for all the differant approaches is, wether it's because tyres aren't understood well enough or CPU power isn't high enough, or test data simply isn't availible.

The problem typically is a lack of test data. With Pacejka's commonly used Magic Tire Model, you'll typically have a set of constants that produces the amount of lateral force given slip angle, load, and so on, but another set for longitudinal force from slip ratio. I.e., you've basically got a graph of lateral force and another one for longitudinal. However, when you're sitting at a certain slip angle and then increase slip ratio, you don't know what the combined effect is. I.e., the longitudinal force will influence the lateral force, even when you're not at the limits of the friction circle.

What developers typically do is take that lateral force and the longitudinal force separately from the two formulae, then if the vector sum goes outside the friction circle they'll either trim the resultant down to fit, or give precedence to the longitudinal force. I.e., you let the longitudinal force determine how much is left for lateral force to fit it. I tried both in Virtual RC Racing and the second method was closer to reality. However, then you don't get the variations of lateral force with slip ratio that occur inside the friction circle when you're not at the limit. This can make a sim lack quite a lot of feel and warning when approaching the limit. And even at the limit, the results aren't really right either. As soon as you hit 0.10-0.15 slip ratio or whatever, you have zero lateral force capacity. This isn't how it works in reality. Sims are harder to drive than real cars in my opinion primarily because of the commonly faulty approach used to model tires. Much more so than anything due to lack of feel in my opinion.

There are ways to deal with it. For instance, the Milliken/Radt nondimensional method described in "Race Car Vehicle Dynamics." However, this method requires even more extensive data on top of the Magic Tire Model, so you're really back to just guessing again unless you have really deep pockets.

I've figured out how to beat this though, and no, I'm not telling

Last edited by jtw62074, Fri, 21 Oct 2005 05:41.

Quote from Cue-Ball :

I'm actually really surprised that anyone who's a big fan of LFS would even like GT Legends. I played the demo and was thoroughly disappointed. The cars didn't feel the least bit lifelike to me.

That amazes me

Quote from JeffR :

An alternate definition of understeer versus oversteer, at least in terms of defining a car's setup.

You drive a car around in a circle, well below the limits. It takes a certain amount of inwards steering input to maintain the circle. Then, while holding this radius, you speed up the car until it is near the limits. The steering inputs required to maintain the raidus determine if the setup has understeer or oversteer.

Understeer - it takes more inwards steering to maintain the radius.

Neutral - it takes the same amount of inwards steering to maintian the radius.

Oversteer - it takes less inwards steering to maintain the radius.

Critical Oversteer - The steering is centered while maintaining the radius.

Critical oversteer isn't a limit, because rally cars and dirt track racing go beyond this.

That's close enough for government work and probably good enough for general discussion

If we want to be picky though this isn't really right. All that matters is what the slip angles are at the end of the day. Basically, if you're running around the circle and you're not needing to countersteer to keep the car from spinning, then the car is either in an understeer or neutral state. Whether or not you had to increase steering lock into the turn to maintain the circle doesn't really matter. You could have been running 5 degrees front slip angle and 3 rear at low speed, then kept the steering right where it was while speeding up. If you wind up at 10 front and 8 rear slip angles, the car is still understeering even though you haven't steered into the turn at all to maintain the circle. In fact, if you countersteered 1 degree to get 9 front and 8 rear, you're still in an understeer condition.

Neutral technically means the front/rear slip angles are the same.

For all practical purposes, if the car doesn't spin out it's understeering or neutral, so your description is probably close enough

As for the LX6 stuff, I'll need to reread some of the later posts to comment.

Quote from JeffR :

Assuming that the car isn't in the middle of a spin

J.B. is right. If the front SA > rear SA then you have understeer by definition. If you're in a spin, the rear SA > front and you're in an oversteer situation. The full SAE definitions are a bit more involved than that, but that's pretty much how it's defined.

Here's another similar diagram that maybe will explain what I mean:

http://www.performancesimulations.com/files/steering3.jpg

The black boxes represent tires in a top down view. The upwards direction is the direction of movement. The difference between the angle that the tire is facing and the direction it's moving (the vertical) is the slip angle, as we all know.

From top to bottom:

#1 - The tire is below the peak (it might be at 5 degree slip angle or something). The red line is the "tire lateral force," which is always measured in the tire's plane. I.e., it's always sticking out the left/right of the tire perpendicular to the direction the tire is facing. When you see a Pacejka graph or a chart of lateral force versus slip angle, you're looking at how the length of this line changes with slip angle. (The larger the force, the longer the line). The green line is the sideways component *in the car's coordinate system.* I.e., when you hit the F9 key in LFS you're looking at the lateral acceleration of the car, not the "tire lateral force." The blue line shows "induced drag." This is what causes the car to slow down more and more as you add steering. The F9 key shows lateral/longitudinal acceleration; the green and blue lines.

#2 - The tire has more slip angle and is producing more lateral force (the red line is longer.) Also, the green line in this case is longer too, so the lateral acceleration is higher than it was in #1.

#3 - The "tire lateral force" has reached its peak. I.e., we're at a slip angle where the red line is as long as it's going to get. If we assume the "tire lateral force" does not drop off at all as we increase slip angle beyond this, all we're doing is taking that same length red line and swinging it further and further towards the rear of the car as we increase slip angle further.

#4 - Here the tire is still producing the same "tire lateral force." However, notice that the lateral component in the vehicle's coordinate frame (the length of the green line) is getting shorter now. At the same time, the blue line (induced drag) is getting much larger now. As this blue line gets longer and longer we are getting more forward weight transfer. This will actually make the red line longer, but it may not increase it so much that the green line increases beyond where it was in #3. If it doesn't, then we are getting more and more understeer as we increase the front slip angle further. If it does, we'll continue to turn in harder. Whether or not this happens depends on the CG height to wheelbase ratio. The higher the CG is in relation to the wheelbase, the more front weight transfer we'll get (which increases the length of all the lines).

#5 - Here's an extreme slip angle. Notice that we've got hardly any lateral force *in the car's coordinate system*. The tire force is still the same and in fact, due to the forward weight transfer, will continue to increase. However, the green line is still very short so we have very little lateral acceleration now. The car is understeering like crazy even though the front tires have not lost any grip at all. It's just that the direction is all pointed rearwards instead of to the left of the car. I.e., the F9 readout will show a very small lateral acceleration. At some angle I'd expect it to be 30% of whatever the peak was

Now, here's the kicker. This happens regardless of whether or not the "tire lateral force" curve drops off or not. I.e., if that red line hits a maximum length and stays there as we swing it out towards 90 degrees slip angle, we'll still get understeer at some point (the green line will tend toward 0 length as 90 degrees is approached). In fact, even if the tire lateral force never peaked at all but continued to grow all the way out to 90 degrees, at some point the car will still understeer. I.e., the green line will eventually drop below it's peak.

The F9 display shows the length of the green and blue lines, not the red one.

Last edited by jtw62074, Wed, 19 Oct 2005 04:47.

Quote from JeffR :

Does anyone, other than the developers, know what the slip angle versus grip curve for the tires in LFS looks like?

Well, your four wheel drift test shows that it hits a peak at some point at about 1.15 and a bit later drops to 1.10 at whatever slip angles you were driving at.

Quote from JeffR :

Lateral grip must fall off somewhat due to slip angle, otherwise a driver couldn't induce understeer by turning the front tires inwards.

Sure you can. Turn the front wheels to 89 degrees slip angle. You won't be turning anywhere, regardless of whether the curves drop off or not. Instead, you'll get about the same effect as if you locked the front wheels. The optimum cornering steering angle is going to be somewhere between 0 and 90 then. Any less or more than that optimum and you'll be turning less hard than you otherwise would. I.e., you're probably in an understeer condition.

Quote from JeffR :

The main thing I noticed is that in real life (I've tested this with several cars going in circles at a big parking lot), the induced understeer effect is much less than it is with LFS.

This backs up my argument that real force curves don't drop off on dry pavement, and if they do it's usually very minor. Take a look at my diagram. If that red line gets shorter as you dial in more steering/slip angle, the green line will get shorter too. That could mean more understeer than you might get in reality.

Also, the center of gravity height influences this too. With a high CG you'll have more forward weight transfer, which makes the red line longer. Of course the green line would then get longer too. If the CG/wheelbase ratio was great enough you could even turn that understeer into oversteer at extreme steering angles even if the front tire curves dropped off after the peak.

Last edited by jtw62074, Wed, 19 Oct 2005 04:39.

Quote from J.B. :

Todd, when are you going to write a book? Nice explanations, as ever. :up:

Thanks. Maybe some day

Quote from JeffR :

When I do web searches for

radial bias ply racing tire slip angle

most sites reports that there grip remains about the same for bias ply slicks, but does fall off for street radial tires. Racing radials have some fall off but not as much.

Top fuel drag racing tires lose a lot of grip if they spin, but wrinkle wall, high grip tires are in a class by themselves.

The stickiest rubber compounds are found on table tennis rackets, with coefficients of friction 7 or higher.

Well, this is what Doug Milliken, author of "Race Car Vehicle Dynamics" told me, and he very frequently conducts these tests himself for various clients. He said outright that most of the books that show these swooping graphs of the tire force dropping way off after the peak are flat out wrong about this. As such, it's not surprising that most web sites are too. He'll be here next week for a meeting so I'll be sure to ask him about it again. The only tire data I've ever seen that shows a force drop off after the peak on street tires (bias or radial) is in the wet. Saw some Nascar tire data once that indeed did show a drop off after the peak, but not very much. Nothing approaching 30%. Also, during braking tests when a wheel locks there is drop off as well, but a lot of this is due to the fact that you're frying the part of the tire that's stuck in the contact patch. Overheating it will change things..

For what it's worth, I've seen coefficient of friction in a polymer of 42 in one case. Don't remember what that was though. Most rubber on glass can hit 8+ (much higher in some cases) at extremely light loads. Trivia really

Here's a crude drawing that shows how the lateral force swings sideways towards the rear of the car, causing the F9 display to show less lateral acceleration. Indeed the lateral acceleration of the car itself will begin reducing once a critical point has been passed provided the CG is not so high that the extra forward weight transfer doesn't offset it. LFS is absolutely correct in this regard.

http://www.performancesimulations.com/files/steering.jpg

Last edited by jtw62074, Wed, 19 Oct 2005 04:36.

Quote from Bob Smith :

Karts do act very strangely to normal cars so I don't think it's fair to do direct comparisons.

One thing I've never sussed out with tyres is knowing you've "got grip" and when you've "lost grip" in terms of physics. From what we've learnt about slip angles there is in fact no loss of grip, and in fact there could even be a slight increase with large slip angles. What is it that makes the car feel like it's cornering normally, and then feel like you're sliding? Is it something to do with the shape of the slip angle grip curve, or some other properties of the tyre?

As Doug Milliken explained it to me, the feeling that the car has "let go" is not because the force curves fall off after the peak (if they indeed do), but rather that the force merely stops rising. When you hit the peak you can get the sensation that the car has broken away. The shape of the curve leading up to this point will be what tells your butt whether the tire breaks away suddenly or not.

There are really two things there:

1. Where the peak occurs (i.e., at what slip angle does the force stop rising). The greater this slip angle is, the more forgiving the tires will feel. Street tires peak at very high slip angles. I've seen data showing peaks at 20 degrees. One set showed it still rising at 28 degrees!

2. The shape of the curve leading up to the peak. If the force rises very linearily and then suddenly flattens out, the tire will feel like it breaks away very suddenly. If instead it rises normally and quickly starts rolling off towards the peak (the slope decreases to 0 gently) it will feel more forgiving. Tire designers can influence this shape substantially through their choices in cord patterns/angles and so forth. Radials typically rise more quickly than bias ply tires and roll off a bit more suddenly (into a flat peak!) This makes them feel like they break away more suddenly then bias tires do and leads people to believe (falsely) that radial tires are losing a bunch of grip after the peak. T'aint so!

This is similar to people's rear ends telling them that cars actually speed up once they hit the grass sideways, which is of course nonsense. The acceleration became lower, that's all.

Last edited by jtw62074, Tue, 18 Oct 2005 23:29.

Quote from JeffR :

Some experiments I ran with LFS S2 alpha. I chose the LX6 as it has a good power to weight ratio and doesn't have downforce.

Drove in circles until cornering forces approached about 1.15g's, then continued to steer inwards until I reached the steering lock point. As I continued to steer inwards, the cornering forces dropped down to .8g's this is more than 30% loss in grip. Ran another test, initiating a spin, but then centering the steering wheel, grip remained about 1.1g's even with the car going sideways. Last test, slamed on the brakes, locked up the tires, slight loss in grip until the tires over heat. While drifting sideways or with all 4 tires locked up, there is very little loss in grip, but steering inputs can cause an extreme loss of grip. This doesn't make sense to me.

Induced understeer doesn't seem to work as well as it should, especially on the LX6 with front swaybar maxed out, and rear sway bar set to 0. With this setup, it would seem that the much stiffer front end would have significantly less lateral grip than the rear end. If the car starts spinning, pegging the fronts inwards should cause the front end to wash out, because they have a much higher slip angle, and because the suspension is set so much stiffer up front, yet induced understeer doesn't work well or not at all.

My conclusion from these experiments is that something strange is going on when one end of a car loses grip, as opposed to both ends losing grip at about the same time. When one end loses grip, the result is much less grip than when both ends lose grip.

I don't see anything wrong with that at all. As Skiingman said, the lateral force vector is rotated further and further rearward as steering angle is increased. If you steered the front wheels to 90 degrees you aren't going to corner at all, but you haven't "lost grip" by any sense. So this "two wheel grip loss" at the front is not necessarily tire grip loss at all, you might be losing 30% lateral acceleration because you've changed the direction of the force vector so severely.

In addition, LFS likes 0 dynamic camber at all times. Any deviation from this at all reduces grip. So as someone else pointed out, the front "wash out" will be compounded by the severe camber induced by steering the front wheels (caster/kingpin inclination causes this to occur). However, the pushing is mostly due to that lateral force vector pointing rearwards. Remember that the lateral acceleration shown with F9 appears to be in the car plane, so the only time you're actually seeing the real grip level is when all four wheels are pointed straight ahead and you're sliding along sideways somewhat. Not suprisingly, you'll see the highest lateral acceleration then when the car has been pitched into the turn and the front wheels are more or less straight ahead.

This can be confirmed with "mass moment method" (MMM) measurements/calcuations done on the Chaparal racing car in Milliken's "Race Car Vehicle Dynamics." Peak lateral acceleration is not obtained in a trimmed attitude. If it were, the car would have no stability at all in a steady state corner at max lateral acceleration (by definition).

The only issue I'd take with Skiingman's post is the statement that a sliding tire produces 30% less grip than a non sliding one. Of course, this really means that somewhere past the peak force slip angle the force drops 30%. This simply is not true. Street tires on dry pavement peak and are flat, flat, flat, even the radials JeffR mentioned Ok, maybe as you get out to 60 degrees at high speed you may see some drop, but it's not anything approaching 30%..

But anyway, Jeff's test that showed the grip dropping from 1.15 to 1.10 when pitched sideways with the wheels straight ahead indicates that the curves in LFS with those tires peak at 1.15 and then drop very slightly to 1.10. Quite realistic for some speeds and tires

Last edited by jtw62074, Wed, 19 Oct 2005 00:27.

If you could do the ride height reduction setting assymetrically, you could put wedge into the car by adjusting one wheel independently of the others.