The online racing simulator
netKar level feedback
(67 posts, started )
Quote from Ball Bearing Turbo :Indeed, before anyone says anything, yes the "regular" physics run at 100Hz, but the internal tire physics do run at 2000Hz.

Android, I think I misunderstood your diagram initially but now I get what you mean. Wouldn't this increase rolling resitance even more than the other way? Do we have any data to back up the thought that they're modelled the way you describe?

Well, the "data" I have is just observation of what happens. Just try to create a severe flatspot and then drive slooooooowly to and over the spot. Maybe make a custom camera view to watch it from the outside, so you can better see what exactly goes on. The tyre will reach the worn down section and then literally fall down an ingame centimetre or so only to bump back up once the end of the worn section is reached.

The rolling resistance is pretty low with this method, as you just have to drive over a tiny ledge. If you play around with the clutch and drive at a really minimal speed over it, then you might be able to see a very small resistance when approaching the ledge, but it's really really tiny. If instead the flatspot was modelled as a real flat area, then the rolling resistance would be much higher at that spot and would give a much more noticeable bump in force feedback (at least if left and right tyres aren't flatspotted at the exact same spot).
Quote from Ball Bearing Turbo :Indeed, before anyone says anything, yes the "regular" physics run at 100Hz, but the internal tire physics do run at 2000Hz.

Android, I think I misunderstood your diagram initially but now I get what you mean. Wouldn't this increase rolling resitance even more than the other way? Do we have any data to back up the thought that they're modelled the way you describe?

Are you sure its 2000hz? that seems a bit high that would mean that LFS can calculate all the tires on the car in half .5 milliseconds. I woud say maybe you meant 200hz, but I can't imagine how you could even run them and different speeds
Yes, it really runs at 100Hz with 20 subsamples for the tyre physics, so they essentially run at 2000Hz.
So says Mr Roberts, and he should know.
Quote from Test Driver :An update of the FF effects in LFS would be great. The FF feels so slow, mostly because the general steering resistance is too high, making very fast movements (and especially countersteering) with the wheel impossible.

which wheel do you use ?
sounds a lot like youre using a dfp which has quite a few flaws int hat respect which arent lfs' fault

Quote :Another strange thing with the FF in LFS is that while the car is standing, you can move the wheel without effort, but once the car starts rolling the steering resistance raises, but this should be the complete opposite way

is the effect correct when you brake while turning the wheel with your car standing still ?
chances are its not down to the ffb itself but rather hidden in the rolling loss calculation which doesnt come into effect while the car is standing still (ie the bearings and everything else dont keep the tyre from rolling unless its rolling at speed)

Quote from Ball Bearing Turbo :Android, I think I misunderstood your diagram initially but now I get what you mean. Wouldn't this increase rolling resitance even more than the other way? Do we have any data to back up the thought that they're modelled the way you describe?

i think the mistake your making here is assuming that androids gif implies that there is a sharp ledge if you model the tyre the way id guess lfs does

if it did the scene where the flatspot is exactly perpendicular to the road surface the tyre would be standing on the ledges on both sides of the flatspot

so discounting tyre flex (ie a solid tyre model) the centre of the wheel would be always exacly r + thread_thickness[i] above ground and thats it

what this means is that there is nothing that models the way in which the tyre goes from one thread thickness to the other which would cause the rolling resistance change
instead the tyre will just hop up and down stressing the suspension

the important bit here is that there is nothing modeling the change in thickness ie nothing to the effect of
rolling_restance = f(thread_thickness[i-1]; thread_thickness[i]: thread_thickness[i+1])

Quote from lalathegreat :Are you sure its 2000hz? that seems a bit high that would mean that LFS can calculate all the tires on the car in half .5 milliseconds. I woud say maybe you meant 200hz, but I can't imagine how you could even run them and different speeds

at least according to scawen lfs has an inner physics loop which afaik does tyres and suspension at 2khz and an outer one which handles collisions damage and a few other things at 100hz
Quote :at least according to scawen lfs has an inner physics loop which afaik does tyres and suspension at 2khz and an outer one which handles collisions damage and a few other things at 100hz

ok that makes sense. i got the impression that only the tires were done at 2khz
Quote from Shotglass :could someone please clear one thing up for me ? from my understanding sat need a certain degree of long slip otherwise there wont be an arm to act on yet those curves never refer to any long slip

Here's a quick and dirty diagram showing the relationships between lateral force, pneumatic trail, and aligning torque in the case of 0 caster and kingpin inclination. This is pretty much what results in the aligning torque diagrams you see in graphs measured on tire testers:

http://www.performancesimulations.com/files/sat1.JPG

The pneumatic trail is just a torque arm. The lateral force varies all throughout the contact patch since you have different amounts of lateral distortion. At the front there's very little and as you proceed towards the rear it increases, then decreases. As a result, the center of force (lateral force centroid) moves forward/rearward in the patch (up and down in the diagram) as the slip angle, vertical load, and resulting lateral force change.

As slip angle starts at 0 and then increases, the force centroid starts pretty much at the center of the patch, then moves rearward toward where the lateral force arrow is drawn in the diagram, then returns toward the center at around the lateral force peak. Sometimes it can probably wind up slightly forward of center at even higher slip angles which would give a reversal in the aligning torque. This is fairly common from data I've seen to date.

Note that there is no longitudinal force here at all. Longitudinal force in the case pictured would project straight through the center of the patch (upwards/downwards in the diagram) and not produce any aligning torque at all since a torque arm would not exist, even if the force centroid was forward or rearward of the patch center. The situation changes if the force centroid is left or right of this, however. More on this in a minute.

Mechanical trail is added here:

http://www.performancesimulations.com/files/sat2.JPG

The steering axis due to caster angle intersects the ground somewhere. At 0 caster the intersection would be smack in the middle of the contact patch. As you increase caster, you move the intersection point forward. The distance from that intersection point to the patch center is the mechanical trail. The total torque arm then becomes the mechanical trail plus the pneumatic trail (not quite, but that's extremely close; there'd be 2% cosine error at 11 degrees caster). In this case, unless you had very near 0 caster, it's unlikely the total aligning torque would ever drop to 0 or reverse direction.

Tire data does not include this mechanical trail. It's measured purely in the tire plane with 0 caster as in the first diagram. NetKar appears to work this way to me. It seems to ignore suspension geometry effects in this aspect, at least.

LFS seems to me to have more or less a constant total trail (pneumatic + mechanical). The effect of that would be the same as multiplying lateral force by some value to get the aligning torque. I don't feel any drop off in aligning torque at high slip angles at all. However, with caster and the resulting mechanical trail in the picture, this should be more natural feeling except that you do miss what would become a subtle drop off in aligning torque as you approach the limit. So it makes sense that some people feel the front tire limit better in NetKar than in LFS, while others might feel that the loss that (to me at least) is missing in LFS is exaggerated in NetKar to the point of it feeling not as good.

To each his own. Some would prefer to feel the front tires approach the limit even if it means the steering wheel getting extremely light or losing all torque (perhaps even reversing it) all of a sudden. Others (myself included) would prefer to have the aligning torque build up and just hit a maximum somewhere without going completely limp. You can still feel the limit with that approach in some sense in that at some slip angle the aligning torque stops increasing. NetKar's approach is better in one way, LFS's is better in another. Overall, if you compared steering wheel torque in each sim to reality, I'd bet LFS's is a lot closer, even if it's missing some magnitude of the drop near the peak that you get in NetKar.

Ideally you'd combine the two approaches. If NetKar is indeed missing the mechanical trail as I strongly suspect, it could be added through a bit of wrangling. The pneumatic trail would have to be calculated in real time from the aligning torque, then the mechanical trail added to it, and finally, the lateral force multiplied by this total trail to get the aligning torque. That would be a huge improvement for Netkar's FFB.

LFS seems to work a bit differently, but similar principles could be used. The point is to add in the pneumatic trail (if my impression of it being either missing or not actually varying with slip angle is correct, of course) in addition to the mechanical trail, and use the combined two for the final aligning torque. If LFS is using a brush or similar tire model, the pneumatic trail can be calculated directly from it. Problem solved. This works quite beautifully, actually.

In both sims, with this approach you get the best of both (I've driven a model that does this). You have the heavy handed feeling you get in LFS, but with a much more subtle version of the aligning torque drop off you get in NetKar near or after the peak, so you feel the limit coming at the front tires better. In addition, you wind up tuning the feel at the steering wheel by adjusting caster, something that race engineers due in real life. A racing engineer once told me that he sometimes does this to trick his driver into using more steering than he otherwise is inclined to, even if it means giving up a tiny bit of lateral force capability. After all, he wasn't using it anyway.

Tire pressure greatly effects pneumatic trail, and therefore aligning torque, as well. This is why in reality you can feel a tire losing air pressure through the steering wheel mid-corner. Drivers can probably feel the tires heating up and building pressure over the first couple of laps as well. If so, that would largely explain it.

Back to longitudinal (traction/braking) force: This can indeed cause an aligning torque at the steering wheel too.

http://www.performancesimulations.com/files/sat3.JPG

Case # 1 shows the steering axis intersection point with the ground like in the previous diagram. Here we have some caster which puts this point forward of the contact center and lateral force centroid. In this case, as mentioned briefly earlier, there is no torque arm present so no aligning torque is generated when traction or braking force exists.

Kingpin angle is the equivalent of caster, but in the left/right direction rather than forward/rearward. In a nut shull, the steering axis can be positioned pretty much anywhere and slanted in any direction, so it can (and usually does) intersect the ground somewhere to the left or right of the patch center in addition to forward/rearward. With a left or right offset present there is indeed a torque arm that longitudinal force will act on, which will generate an aligning torque when traction force is present, and an opposite aligning torque when braking force is present. The steering axis is changing with suspension travel too. (This isn't all as hard as it sounds, by the way. The calculations are pretty short and straightforward)

Granted, in a perfectly straight line you wouldn't notice it if the suspension is symmetrical because the left/right tires would cancel each other's aligning torque out at the steering wheel. In other situations though, there would usually be some effect.

The force centroid may not naturally be in the exact center of the contact patch. Due to asymmetries in the design, the presence of camber, tire pressure, and so forth, the force centroid is probably slightly to the left or right of center. However, the point is that there can be a torque arm in the lateral direction and one in the longitudinal for both lateral and longitudinal forces to act on. In aligning torque data this may partially explain why aligning torque sometimes reverses at some high slip angle.

Kingpin versus caster angle :
http://www.performancesimulations.com/files/steeringaxis.JPG

Scawen (or Stefano, for that matter), in case you're reading: To get all these effects, you just need to make sure the force centroid is free to move about in the contact patch (or something equivalent), then take the portion of the resulting tire force vector acting perpendicular to the steer axis. Voila: Steering torque that varies with everything under the sun including suspension travel and so forth. It can be taken further than this, but it's a good approach.
yes i rekcon it is a great sever but most of the time i go on there always wreckers :S
Quote from jtw62074 :Back to longitudinal (traction/braking) force: This can indeed cause an aligning torque at the steering wheel too.

what i actually meant was that the long force would pull the force centroid out in the long direction to create the arm any lateral force can generate a torque on
apparently my embarassingly basic understanding of how the contact patch acts unter purely 1 dimensional loads hindered me from getting what actually happens
I hope this will help clarify the difference in aligning torque due to pneumatic trail alone versus pneumatic and mechanical trail acting together. This is relevant to the NetKar versus LFS FFB discussion.

First, here's one of BuddhaBing's links so we can remember what aligning torque curves are and how they look:

http://i1.tinypic.com/4y7l0ys.jpg

The curves rise, hit a peak, then drop back down and sort of tend to level off somewhere either above or below the 0 aligning torque line. Data like this is measured on a tire machine with no suspension. These curves result from the lateral force and the pneumatic trail, but without any mechanical trail.

Now here's a fancy new diagram from me

http://www.performancesimulations.com/files/sat4.JPG

Aligning torque is the blue line times the red line. It's just a force acting at some distance from a fulcrum, like a wrench on a nut. Only this wrench is trying to twist the tire and the steering wheel that's connected to it around the steering axis of the suspension.

The left column of pictures show what's happening in the first link (the aligning torque curve graph). Note that the slip angles and magnitudes of the force and trail here do not match up with what's shown in the aligning torque graphs. My picture is completely fictional and is just for illustration.

Starting at the top left picture, at some tiny slip angle there is a little lateral force (red arrow). The force is acting very close to the center of the contact patch. The aligning torque is this distance multiplied by the force. A small force times a very small distance gives a very small torque, as shown in BuddhaBing's link where the aligning torque is very near or equal to 0 at 0 slip angle.

As lateral force increases, the force centroid moves rearward in the contact patch as we go to the fictional 3 degrees. Both the torque arm (pneumatic trail) and the lateral force get larger. The aligning torque is increasing. In my example pictures, at 5 degrees the aligning torque and lateral force are both quite large. Here we're right around the peak of the aligning torque curve.

As we continue to increase slip angle, the force centroid moves back towards the center and may even move past it so it's forward of the center of the contact patch. The aligning torque drops and may become negative.

If Netkar is using the aligning torque curves from the Avon data directly, then it should be working just like the left column.

The right column shows the same thing, but the torque arm includes both pneumatic trail and the mechanical trail on top of it. For this fictional tire, at 7 and 10 degrees slip angle the lateral force is the same. The pneumatic trail decreases (torque arm, blue line, gets shorter), so you'd feel a little bit of lightening in the steering wheel as you get near and past the peak of the lateral force curve, but nowhere near what you get in the left column where pneumatic trail is assumed to be the only influence. There, if the pneumatic trail is cut in half and the lateral force held constant, the aligning torque is cut in half too. In the right column, the pneumatic trail is only part of the total trail, so cutting it in half (or doubling it) changes the aligning torque much less. The aligning torque is more dependent on lateral force than it is in the left column.

One could imagine a third column there with the trail held constant. I.e., the lateral force doesn't move forward or rearward in the contact pach, but sits still somewhere. This wouldn't necessarily have to be the center, so you could have pneumatic trail included that is constant and doesn't vary with slip angle as one type of model. The aligning torque then varies only with the lateral force. This is how LFS feels to me and is likely the big difference in FFB between NetKar and LFS. If that is correct, then the second column would sort of blend the two together and better reflect reality, for both LFS and NetKar
[Chanting]
Todd!
Todd!
Todd!
[/Chanting]

You really know how to explain stuff to the layman. You a teacher, by chance? Anyways, your reflections seems logical and they support what I feel whan I drive my car. Let's hope Scawen reads these threads (I suspect he does).
Really really great explanations, Todd. :clapclap:
#63 - w126
Quote from atledreier :Let's hope Scawen reads these threads

Do you think he's been getting complete lines of code from Yoda in his sleep?
One interesting thing:
Take the FZR (since it has big rear traction), give it very high brake force.
Move the bias to 95% front.
Attempt a burn out, but you won't be able to since there's too much traction - the fronts will stay locked while the rears push the car foward.

Note that the wheel goes 100% limp, and can be spun freely....

Or just get up some speed, lock the fronts, and note the same thing.
Quote from Ball Bearing Turbo :One interesting thing:
Take the FZR (since it has big rear traction), give it very high brake force.
Move the bias to 95% front.
Attempt a burn out, but you won't be able to since there's too much traction - the fronts will stay locked while the rears push the car foward.

Note that the wheel goes 100% limp, and can be spun freely....

Or just get up some speed, lock the fronts, and note the same thing.

with locked fronts there shouldnt be much of any lateral force and thus no wheel torque
while the lfs behaviour might be a bit too idealized i dont think it is completely off per se
Todd, how does pneumatic trail vary with the longitudinal forces acting on the tyre? None of the definitions of pneumatic trail I have come across mention longitudinal forces yet surely they must have some effect? I've come across an informal and very brief mention that pneumatic trail decreases with vehicle speed but haven't seen any formal definitions that include (directly) a longitudinal force component. Any pointers to papers etc would be very much appreciated.
I'm not sure and haven't seen any specific mention of how pneumatic trail varies with it.

There are graphs floating around out there, however, that show aligning torque as a function of longitudinal force at several different slip angles. In one of them, even at 0 slip angle there was some non-0 aligning torque (about 10 N-M) which increased towards about 20 N-M with a bit more than 5000N longitudinal force (compared to 160 N-M at 6 deg slip angle, the highest the data went to). It was a little different under braking than acceleration, but there was this small change even at 0 slip angle.

Up to about 4 degrees the aligning torque rises a little bit under very slight braking, then drops off toward "0." This doesn't specifically say what the pneumatic trail is of course. If there was a combined operation lat/long force diagram to go along with it maybe the pneumatic trail could be figured from that. I haven't looked for this at all, but will keep my eyes open for it.

netKar level feedback
(67 posts, started )
FGED GREDG RDFGDR GSFDG