I don't know.
Got me again. The one shown in the graph in my post was just under 250 ft-lb. I don't know how typical that is though.
I don't know how it works mechanically. To me it's just a magic black box
Yes, that's probably right. However, in order for the road reaction torques to exit at all, there will usually be engine torque of course. Mathematically you're already taking that into account by dealing with it only with the reaction torques coming in from the wheels. If LH = 300 and RH = 500 then there's probably already an engine torque there.
"Limits" is correct. If the car is in a very gentle turn the diff is locked and the slip ratios and vertical loads may be nearly the same. The outside tire force is not necessarily fixed to the ratio. It can be less or identical to the inside tire such as when running straight.
Hi guys,
Diffs are pretty odd devices and can be difficult to understand. It's frequently a source of great confusion for developers too, ranking right up there with tire modelling, so don't feel bad if they go over your head. It took me a long time to get my head around this subject too.
Here's a map showing how a limited slip differential operates, taken from Milliken's "Race Car Vehicle Dynamics."
http://performancesimulations.com/files/diff2.jpg
Much of the confusion comes about due to the desire to visualize the engine input torque creating the locking torque in the diff, then wanting to know how that torque is then fed to the wheels. In reality, at least mathematically speaking, the engine torque input is not really a variable to be looking at here.
Looking at this graph, instead of thinking about torque going outward to the wheels from the engine, flip it around and imagine the reaction torque from the road coming into the tires. (Huh? :razz The fact of the matter is you won't know how much torque is going to each tire from the engine while the car is cornering unless you know the slip ratios and weight transfer.
The engine is trying to turn the wheels forward (positive torque). At the same time, however, the tire rubber being stretched is trying to slow the tire back down by twisting in the opposite direction (negative torque). Let's just call this negative torque the "road reaction torque." This is the tire and road interaction that is fighting the engine.
Whether or not the differential remains locked depends on what the two road reaction torques are coming in from the left and right tires. With a given percentage locking factor, there is a constant ratio that can't be exceeded between the left and right sides (except when operating within the preload area. More on that later.) In this particular graph the ratio is 2.90:1. I don't recall what percentage locking factor that is, but it's not important.
We'll accelerate hard in a left hand turn. We have a healthy amount of weight transfer to the right side tire so the forward traction force is greater on the right than it is on the left. This also means that our road reaction torque (the negative torque reaction) is higher on the right than the left.
Looking at point A on the chart, we have 500 lb-ft torque on the right side and 250 lb-ft on the left (really they should be negative values, but this map is symmetrical so I'll just stick with positive numbers). That ratio is 500/250 or 2:1. This is lower than our 2.90:1 torque bias ratio, so the diff remains locked. This can be verified on the graph by seeing that point A is inside a shaded area. Any time you're in the shaded area the diff is either locked or will become locked soon. For now just consider it locked so we can ignore transient phases that don't last very long anyway.
If we suddenly increased weight transfer to the outside tire, the forward force at the outside (RH) tire will rise and the forward force at the inside (LH) tire will drop. We might find ourselves at point C, with RH=750 ft-lb and LH = 200, where we are outside of the shaded area.
Here's where the differential magic happens. We're outside of the shaded area so our diff begins slipping and the wheels begin rotating at different speeds. This means that the slip ratios at the tires change. It turns out that the slip ratios will adjust themselves in such a way as to move us down to point B. The diff is not locked, but the outside tire will not produce any more than 2.90 times the force that the inside tire is producing.
Mathematically we started with RH = 750 and LH = 200, a ratio of 750/200 = 3.75:1. Our differential only allows the outside tire to produce 2.90 times whatever torque the inside tire produces. The differential slips and the outside tire slows down just enough to arrive at a new slip ratio that produces 2.90 times whatever the inside tire was doing (LH = 200).
The RH torque becomes 2.90 * 200 = 580 lb-ft.
Ok, next chart:
http://performancesimulations.com/files/diff4.jpg
Here we see the forward forces at the tires as we accelerate in a left hand turn. An open differential is similar to our LSD except it has a torque bias ratio of 1:1 instead of 2.90:1. What this means is that the outside tire can produce no more than the inside tire can. The forward forces remain the same. If you increase weight transfer and cause the inside tire to reduce force, the outside tire force will drop right along with it. This is because the differential action changes the slip ratios at the tires "just right" to maintain this force ratio of 1:1.
The second diagram is a limited slip diff with a torque bias ratio of 2:1. The wheels remain locked together or the differential slips in a way that makes sure that the outside tire can produce no more than 2 times the force that the inside tire produces. If the torque bias ratio is 5:1, it can make 5 times the force, etc.. The locking percentage maps to this torque bias ratio.
Ok, so what about preload?
Let's go back to the first diagram. The differential map. Without that preload area, at very low traction forces (low throttle), the ratio of forces can still only be 2.90:1. Preload allows the torque bias ratio to become much greater in low throttle or light coasting/braking situations, such as when you are approaching the apex of a corner and maintaining more or less a constant speed.
Point D shows RH = 750 and LH = 50 (or so). With a torque bias ratio of 2.90 the differential would slip and produce RH = 50 * 2.9 = 145 lb-ft while LH remains unchanged as always. However, with that preload area there we can have a constant torque difference exist without the torque bias ratio sticking its head into our business. The preload here is just under 250 lb-ft. This is where the preload crosses each axis (just a bit to the left of point E). In this particular case we move to Point E right at the edge of the preload shaded area where our outside tire produces 300 lb-ft of torque. Note that the inside tire is only producing 50 (6:1 ratio), but the preload allows us to have this big torque on the outside tire anyway. The preload is 250, and indeed in this case the outside tire produces 50 + 250 = 300 lb-ft of torque.
So long as the torque difference between the left/right tires is less than 250 lb-ft the axle will remain locked, regardless of the ratio of the forces. I.e., LH = 10 and RH = 200 gives a torque *difference* of 200-10=190 lb-ft, with a torque *ratio* of 20:1. The preload wins since our preload torque is 250 lb-ft and our axle remains locked
What's important here is that preload is not effecting the differential operation at all except in these situations where there isn't very much throttle or braking being used. Anything beyond that and the locking torque from torque bias ratio will win. The triangular shapes devour the preload area underneath it. They are not cumulative. You have a torque ratio between the tires as well as a torque difference. Locking percentage is equated with torque ratio while preload is linked to torque difference.
Back to the other diagram showing the axles: The third picture shows a light throttle situation without any preload. The outside force can grow no larger than 2 times whatever the inside tire force is. However, if we add in some preload, the difference can be 250 lb-ft (or whatever our preload setting is), regardless of the ratio that would result from that. The fourth picture shows the outside tire force growing considerably, well in excess of the 2:1 ratio allowed by the torque bias ratio.
As we feed in more power and the torque *difference* between the tires grows larger than 250 lb-ft (our preload setting), then the 2:1 ratio kicks in and the outside force will not go over 2 times whatever the inside tire force is. So when you're on the throttle or brake really hard while cornering, the preload is not having any effect at all.
What I suggest you guys do is try the forces view looking down on the car from the top, then drive in circles at the car park to see the forward forces and how they change with locking percentage and preload. With no locking percentage (torque bias ratio of 1:1), and 250 lb-ft preload, the outside tire will not produce any more than 250 lb-ft of whatever the inside tire is producing. I don't know if there is a preload setting on the open diff, but essentially it's just a constant torque trying to speed up one wheel and slow down the other. Once the forces get large the torque bias ratio effect overpowers that. It's one or the other.
Diffs are pretty odd devices and can be difficult to understand. It's frequently a source of great confusion for developers too, ranking right up there with tire modelling, so don't feel bad if they go over your head. It took me a long time to get my head around this subject too.
Here's a map showing how a limited slip differential operates, taken from Milliken's "Race Car Vehicle Dynamics."
http://performancesimulations.com/files/diff2.jpg
Much of the confusion comes about due to the desire to visualize the engine input torque creating the locking torque in the diff, then wanting to know how that torque is then fed to the wheels. In reality, at least mathematically speaking, the engine torque input is not really a variable to be looking at here.
Looking at this graph, instead of thinking about torque going outward to the wheels from the engine, flip it around and imagine the reaction torque from the road coming into the tires. (Huh? :razz
The fact of the matter is you won't know how much torque is going to each tire from the engine while the car is cornering unless you know the slip ratios and weight transfer.The engine is trying to turn the wheels forward (positive torque). At the same time, however, the tire rubber being stretched is trying to slow the tire back down by twisting in the opposite direction (negative torque). Let's just call this negative torque the "road reaction torque." This is the tire and road interaction that is fighting the engine.
Whether or not the differential remains locked depends on what the two road reaction torques are coming in from the left and right tires. With a given percentage locking factor, there is a constant ratio that can't be exceeded between the left and right sides (except when operating within the preload area. More on that later.) In this particular graph the ratio is 2.90:1. I don't recall what percentage locking factor that is, but it's not important.
We'll accelerate hard in a left hand turn. We have a healthy amount of weight transfer to the right side tire so the forward traction force is greater on the right than it is on the left. This also means that our road reaction torque (the negative torque reaction) is higher on the right than the left.
Looking at point A on the chart, we have 500 lb-ft torque on the right side and 250 lb-ft on the left (really they should be negative values, but this map is symmetrical so I'll just stick with positive numbers). That ratio is 500/250 or 2:1. This is lower than our 2.90:1 torque bias ratio, so the diff remains locked. This can be verified on the graph by seeing that point A is inside a shaded area. Any time you're in the shaded area the diff is either locked or will become locked soon. For now just consider it locked so we can ignore transient phases that don't last very long anyway.
If we suddenly increased weight transfer to the outside tire, the forward force at the outside (RH) tire will rise and the forward force at the inside (LH) tire will drop. We might find ourselves at point C, with RH=750 ft-lb and LH = 200, where we are outside of the shaded area.
Here's where the differential magic happens. We're outside of the shaded area so our diff begins slipping and the wheels begin rotating at different speeds. This means that the slip ratios at the tires change. It turns out that the slip ratios will adjust themselves in such a way as to move us down to point B. The diff is not locked, but the outside tire will not produce any more than 2.90 times the force that the inside tire is producing.
Mathematically we started with RH = 750 and LH = 200, a ratio of 750/200 = 3.75:1. Our differential only allows the outside tire to produce 2.90 times whatever torque the inside tire produces. The differential slips and the outside tire slows down just enough to arrive at a new slip ratio that produces 2.90 times whatever the inside tire was doing (LH = 200).
The RH torque becomes 2.90 * 200 = 580 lb-ft.
Ok, next chart:
http://performancesimulations.com/files/diff4.jpg
Here we see the forward forces at the tires as we accelerate in a left hand turn. An open differential is similar to our LSD except it has a torque bias ratio of 1:1 instead of 2.90:1. What this means is that the outside tire can produce no more than the inside tire can. The forward forces remain the same. If you increase weight transfer and cause the inside tire to reduce force, the outside tire force will drop right along with it. This is because the differential action changes the slip ratios at the tires "just right" to maintain this force ratio of 1:1.
The second diagram is a limited slip diff with a torque bias ratio of 2:1. The wheels remain locked together or the differential slips in a way that makes sure that the outside tire can produce no more than 2 times the force that the inside tire produces. If the torque bias ratio is 5:1, it can make 5 times the force, etc.. The locking percentage maps to this torque bias ratio.
Ok, so what about preload?
Let's go back to the first diagram. The differential map. Without that preload area, at very low traction forces (low throttle), the ratio of forces can still only be 2.90:1. Preload allows the torque bias ratio to become much greater in low throttle or light coasting/braking situations, such as when you are approaching the apex of a corner and maintaining more or less a constant speed.
Point D shows RH = 750 and LH = 50 (or so). With a torque bias ratio of 2.90 the differential would slip and produce RH = 50 * 2.9 = 145 lb-ft while LH remains unchanged as always. However, with that preload area there we can have a constant torque difference exist without the torque bias ratio sticking its head into our business. The preload here is just under 250 lb-ft. This is where the preload crosses each axis (just a bit to the left of point E). In this particular case we move to Point E right at the edge of the preload shaded area where our outside tire produces 300 lb-ft of torque. Note that the inside tire is only producing 50 (6:1 ratio), but the preload allows us to have this big torque on the outside tire anyway. The preload is 250, and indeed in this case the outside tire produces 50 + 250 = 300 lb-ft of torque.
So long as the torque difference between the left/right tires is less than 250 lb-ft the axle will remain locked, regardless of the ratio of the forces. I.e., LH = 10 and RH = 200 gives a torque *difference* of 200-10=190 lb-ft, with a torque *ratio* of 20:1. The preload wins since our preload torque is 250 lb-ft and our axle remains locked
What's important here is that preload is not effecting the differential operation at all except in these situations where there isn't very much throttle or braking being used. Anything beyond that and the locking torque from torque bias ratio will win. The triangular shapes devour the preload area underneath it. They are not cumulative. You have a torque ratio between the tires as well as a torque difference. Locking percentage is equated with torque ratio while preload is linked to torque difference.
Back to the other diagram showing the axles: The third picture shows a light throttle situation without any preload. The outside force can grow no larger than 2 times whatever the inside tire force is. However, if we add in some preload, the difference can be 250 lb-ft (or whatever our preload setting is), regardless of the ratio that would result from that. The fourth picture shows the outside tire force growing considerably, well in excess of the 2:1 ratio allowed by the torque bias ratio.
As we feed in more power and the torque *difference* between the tires grows larger than 250 lb-ft (our preload setting), then the 2:1 ratio kicks in and the outside force will not go over 2 times whatever the inside tire force is. So when you're on the throttle or brake really hard while cornering, the preload is not having any effect at all.
What I suggest you guys do is try the forces view looking down on the car from the top, then drive in circles at the car park to see the forward forces and how they change with locking percentage and preload. With no locking percentage (torque bias ratio of 1:1), and 250 lb-ft preload, the outside tire will not produce any more than 250 lb-ft of whatever the inside tire is producing. I don't know if there is a preload setting on the open diff, but essentially it's just a constant torque trying to speed up one wheel and slow down the other. Once the forces get large the torque bias ratio effect overpowers that. It's one or the other.
Last edited by jtw62074, .
I didn't know this one
I've always used them interchangeably.If you've got a wheel or joystick with enough buttons on it, it's a good idea to map the speed limiter to one of them to keep you from fumbling for the keyboard. You can do this in the options screen.
Have fun with LFS! It rocks!
Have fun with LFS! It rocks!
Yep, me too.
My job a few years back was to sell used electronics equipment on eBay. I did probably twenty to fourth auctions every day for a year and a half or so. Thousands and thousands of them (our department did probably 40,000 in that time). PayPal was 99+% fool proof. I can recall maybe two or three problems in the whole department that occurred, and they were a result of people trying to scam us. Not the other way around or PayPal making a mistake.
Granted, the OP isn't talking about PayPal, but the point is online stuff is much more secure than many people seem to believe.
Just won this yesterday. Can't wait to get it Need to get an NVidia card though, but oh well. I've been itchin' for an 8800GTX
http://cgi.ebay.com/ws/eBayISA ... spagename=ADME:L:RTQ:US:1
Need to get an NVidia card though, but oh well. I've been itchin' for an 8800GTX http://cgi.ebay.com/ws/eBayISA ... spagename=ADME:L:RTQ:US:1
Whoops, I stand corrected. Turns out this is a bit more towards oversteer than most production cars, but by racing standards it is indeed quite understeery, especially in comparison to some of what you guys like to run
Understeer gradient at the limit was about 3 deg/g, which is about the same as the most oversteery production sports cars.
Perhaps. I should probably try a stiff rear ARB like the LFS cars use. The default set ups have the same tendency, however. Locked or very stiff front diff increases oversteer with throttle rather than reducing it. I'll have to give that a go and see if I can get it to happen at all in mine.
Weight should have no effect at all. Power is irrelevant too as during the testing as I used all throttle positions. Even just a very, very slight increase in throttle caused understeer.
Combined slip and front wheel drive
Ok, here's my little dissertation on why combined slip might be responsible for what's happening with the front wheel drive cars.
http://performancesimulations.com/files/combined.JPG
Here we have three brilliantly illustrated pictures of a car travelling forward toward the top of the page. The car is in a left hand turn and we're looking at the forces and components thereof at the right front tire.
Picture A is cornering with some throttle. The black line is the traction force due to slip ratio which is pointing in the same direction the tire is pointing. Steered a bit to the left. The green line is the lateral force that's pointed in the same direction as the spin axis of the tire. The red line is the sum of both forces. I.e., the red force is identical to the black and green lines acting together as separate forces. (The grey lines at the tip of the red line are supposed to be parallel to the black and green lines, but I'm not coordinated enough to do that perfectly by eye
Coders often make lousy artists )
Now, what we'll do is increase the throttle (traction force) while keeping the steering the same. Two things can happen depending on how the tire model handles combined slip. These are represented by pictures B and C which are supposed to have the same traction force and direction (black line)
In picture B the lateral force does not change when you increase the throttle. This is the most frequent screw up that new sim developers do and is often the main reason why when you get a new guy on the sim dev block there is just something wrong with the handling that you can't quite put your finger on. With front wheel drive it becomes much more obvious, but with rear wheel drive it's a bit more subtle.
With rear wheel drive you end up with handling such that subtle changes in throttle don't change the attitude of the car in the way you'd expect, if at all. Increasing throttle increases rear weight transfer and therefore lateral force at the same time, which causes a reduction in rear slip angles (the car tends to understeer when you're under the limit). This is because the lateral force is not being reduced as the traction force is increasing as it does in a real tire.
With this type of approach, the lateral force doesn't begin dropping until, and this is very important, the combined force (the red line) hits the edge of the friction circle. Then, one or both forces are modified in order to fit friction circle theory, which says that the red line can only be so long. So if you calculate the length of the black line by using slip ratio, then calculate the length of the green line by using slip angle, the resultant (the red line, the "combined force") can be too long.
There are at least three main ways to deal with this in a Pacejka type or other strictly empirical model (which I think LFS does not use, correct?). One of them is fairly complex and I'm not entirely sure how it works, but is quite good and gives proper results. The other two are the more obvious things to try and kept me up thinking many nights in the early days
One way is to let the longitudinal force stay right where it was and then scale back the lateral force in order to bring the combined force back inside the traction circle. There's a very serious drawback to this method though and many of the sims you've driven over the years worked precisely this way. If your slip ratio hits 0.1 or so, which might be the peak of the traction force, the lateral force completely disappears no matter what the slip angle is. Yikes..
Ok, let's rewind just a little bit to illustrate what this does to the handling of the car. Imagine you're in the turn (still RWD, I've gone off on a tangent here). You're nowhere near the limit of traction. You're well inside the friction circle. Now, you start increasing throttle slowly towards full. If your tire model is calculating the length of the green line (lateral force) strictly as a function of slip angle, it will not change length at all. The lateral force at the rear of the car stays the same in the absence of any rear weight transfer. With the weight transfer the lateral force increases and the slip angle reduces. I.e., as you begin feeding in the throttle you get more and more understeer.
At some point the combined force (the red line again) finally hits the traction circle and you rather suddenly begin scaling back the lateral force. Suddenly, you are getting oversteer and even a little bit more throttle might put the traction force right over the limit. The lateral force plummets to near 0 very quickly.
Sudden, uncontrollable, snap oversteer with no warning whatsoever. Sound familiar?
This is precisely to me how LFS felt until last April's patch (Q?) and is what I was going on about with the combined forces being way, way wrong. After that it was *drastically* improved and as a result the cars felt much easier to drive. And of course, many people cried "arcade" when in actuality LFS had taken a massive leap forward in the realism department.
On to the second of the three main ways to deal with combined forces in an empirical type of model: Instead of leaving the longitudinal force right where it was and scaling back the combined force (red line) to fit the traction circle, you just scale the combined force back. The direction of the force doesn't change then, but the size of it does. The traction and lateral force (black and green) simply scale back and remain proportional to each other.
EDIT: The above paragraph should read:
On to the second of the three main ways to deal with combined forces in an empirical type of model: Instead of leaving the longitudinal force right where it was and scaling back the lateral force (green line) to make the combined force (red line) fit the traction circle, you just scale the combined force back. The direction of the force doesn't change then, but the size of it does. The traction and lateral force (black and green) simply scale back and remain proportional to each other.
This produces much nicer results without the sudden snap oversteer following understeer with slowly increasing throttle in a RWD car. This is precisely how Virtual RC Racing's combined slip model works (the public version that is, the new version in development is totally different). This is wrong too, however. Again, you have a situation where the lateral force is not changing as you feed in more traction force at all until you hit the friction circle. Then, it suddenly begins scaling back, but it does it more subtly and in combination with the traction force.
Both approaches are wrong at the friction circle limit and even more so when operating below the limit, as increasing traction force in that area does not change the lateral force at all. You get increased rear weight transfer which reduces the rear slip angles (increasing understeer), then suddenly at the limit the tires let go as the lateral force suddenly begins plummeting. You can't really steer with the throttle properly by coaxing the rear end out when under the limit like you can in reality in many cars.
When it's done properly, the lateral force will drop with increasing traction regardless of the slip angle. Even when you're way under the limit. As you approach the limit it winds up very smoothly transitioning into limit behavior. As such, you can steer with the throttle quite nicely. It's all very predictable and drifting actually becomes quite a lot easier. Suddenly you find yourself steering as much or more with the throttle than the steering wheel, even when you're below the traction limit. I've yet to see that anywhere except in Gregor Veble's model for Racing Legends and my own.
Ok, back to the FWD (that was a long tangent):
Starting at figure A, we are cornering pretty hard, a bit under the limit of the front tires, with some throttle. We then increase throttle. If our combined slip model works as either method described above, we can easily find ourselves in situation B. The lateral force stays the same and the combined force (red line) gets bigger and changes direction.
The blue line: This is the yaw component on the car that is trying to twist it to the left. If that gets longer we tend to get oversteer. See picture B? We increase throttle, the blue line gets longer, and we go away from understeer or might even get oversteer if it's long enough.
With a proper combined slip model, we should wind up with something more like figure C. We increase forward tractive force, the lateral force (green) drops, and the combined (red) force grows and changes direction. My illustration shows the combined force being shorter, but really I meant it to grow instead. However, the result is that the blue line gets shorter, meaning less lateral force or yaw torque in the car's coordinate system, and we get understeer instead of oversteer.
The combined slip model in LFS seems quite good to me. At least as good if not better than any other sim I've tried. Before patch Q it was, umm.., not so good, but now it's quite super. However, there is probably room for improvement here and what we see with the front wheel drives is likely proof that something is just a little bit amiss there still.
It's still my favorite sim to drive though, aside from my own of course The new sounds are just fantastic and now that I've got a G25 wheel I can finally use FFB without getting all the rocking of the wheel around the center. The FFB is very good in LFS. FFB has kind of ruined some of the other sims for me in comparison, quite frankly
http://performancesimulations.com/files/combined.JPG
Here we have three brilliantly illustrated pictures of a car travelling forward toward the top of the page. The car is in a left hand turn and we're looking at the forces and components thereof at the right front tire.
Picture A is cornering with some throttle. The black line is the traction force due to slip ratio which is pointing in the same direction the tire is pointing. Steered a bit to the left. The green line is the lateral force that's pointed in the same direction as the spin axis of the tire. The red line is the sum of both forces. I.e., the red force is identical to the black and green lines acting together as separate forces. (The grey lines at the tip of the red line are supposed to be parallel to the black and green lines, but I'm not coordinated enough to do that perfectly by eye
Coders often make lousy artists )Now, what we'll do is increase the throttle (traction force) while keeping the steering the same. Two things can happen depending on how the tire model handles combined slip. These are represented by pictures B and C which are supposed to have the same traction force and direction (black line)
In picture B the lateral force does not change when you increase the throttle. This is the most frequent screw up that new sim developers do and is often the main reason why when you get a new guy on the sim dev block there is just something wrong with the handling that you can't quite put your finger on. With front wheel drive it becomes much more obvious, but with rear wheel drive it's a bit more subtle.
With rear wheel drive you end up with handling such that subtle changes in throttle don't change the attitude of the car in the way you'd expect, if at all. Increasing throttle increases rear weight transfer and therefore lateral force at the same time, which causes a reduction in rear slip angles (the car tends to understeer when you're under the limit). This is because the lateral force is not being reduced as the traction force is increasing as it does in a real tire.
With this type of approach, the lateral force doesn't begin dropping until, and this is very important, the combined force (the red line) hits the edge of the friction circle. Then, one or both forces are modified in order to fit friction circle theory, which says that the red line can only be so long. So if you calculate the length of the black line by using slip ratio, then calculate the length of the green line by using slip angle, the resultant (the red line, the "combined force") can be too long.
There are at least three main ways to deal with this in a Pacejka type or other strictly empirical model (which I think LFS does not use, correct?). One of them is fairly complex and I'm not entirely sure how it works, but is quite good and gives proper results. The other two are the more obvious things to try and kept me up thinking many nights in the early days
One way is to let the longitudinal force stay right where it was and then scale back the lateral force in order to bring the combined force back inside the traction circle. There's a very serious drawback to this method though and many of the sims you've driven over the years worked precisely this way. If your slip ratio hits 0.1 or so, which might be the peak of the traction force, the lateral force completely disappears no matter what the slip angle is. Yikes..
Ok, let's rewind just a little bit to illustrate what this does to the handling of the car. Imagine you're in the turn (still RWD, I've gone off on a tangent here). You're nowhere near the limit of traction. You're well inside the friction circle. Now, you start increasing throttle slowly towards full. If your tire model is calculating the length of the green line (lateral force) strictly as a function of slip angle, it will not change length at all. The lateral force at the rear of the car stays the same in the absence of any rear weight transfer. With the weight transfer the lateral force increases and the slip angle reduces. I.e., as you begin feeding in the throttle you get more and more understeer.
At some point the combined force (the red line again) finally hits the traction circle and you rather suddenly begin scaling back the lateral force. Suddenly, you are getting oversteer and even a little bit more throttle might put the traction force right over the limit. The lateral force plummets to near 0 very quickly.
Sudden, uncontrollable, snap oversteer with no warning whatsoever. Sound familiar?
This is precisely to me how LFS felt until last April's patch (Q?) and is what I was going on about with the combined forces being way, way wrong. After that it was *drastically* improved and as a result the cars felt much easier to drive. And of course, many people cried "arcade" when in actuality LFS had taken a massive leap forward in the realism department.
On to the second of the three main ways to deal with combined forces in an empirical type of model: Instead of leaving the longitudinal force right where it was and scaling back the combined force (red line) to fit the traction circle, you just scale the combined force back. The direction of the force doesn't change then, but the size of it does. The traction and lateral force (black and green) simply scale back and remain proportional to each other.
EDIT: The above paragraph should read:
On to the second of the three main ways to deal with combined forces in an empirical type of model: Instead of leaving the longitudinal force right where it was and scaling back the lateral force (green line) to make the combined force (red line) fit the traction circle, you just scale the combined force back. The direction of the force doesn't change then, but the size of it does. The traction and lateral force (black and green) simply scale back and remain proportional to each other.
This produces much nicer results without the sudden snap oversteer following understeer with slowly increasing throttle in a RWD car. This is precisely how Virtual RC Racing's combined slip model works (the public version that is, the new version in development is totally different). This is wrong too, however. Again, you have a situation where the lateral force is not changing as you feed in more traction force at all until you hit the friction circle. Then, it suddenly begins scaling back, but it does it more subtly and in combination with the traction force.
Both approaches are wrong at the friction circle limit and even more so when operating below the limit, as increasing traction force in that area does not change the lateral force at all. You get increased rear weight transfer which reduces the rear slip angles (increasing understeer), then suddenly at the limit the tires let go as the lateral force suddenly begins plummeting. You can't really steer with the throttle properly by coaxing the rear end out when under the limit like you can in reality in many cars.
When it's done properly, the lateral force will drop with increasing traction regardless of the slip angle. Even when you're way under the limit. As you approach the limit it winds up very smoothly transitioning into limit behavior. As such, you can steer with the throttle quite nicely. It's all very predictable and drifting actually becomes quite a lot easier. Suddenly you find yourself steering as much or more with the throttle than the steering wheel, even when you're below the traction limit. I've yet to see that anywhere except in Gregor Veble's model for Racing Legends and my own.
Ok, back to the FWD (that was a long tangent):
Starting at figure A, we are cornering pretty hard, a bit under the limit of the front tires, with some throttle. We then increase throttle. If our combined slip model works as either method described above, we can easily find ourselves in situation B. The lateral force stays the same and the combined force (red line) gets bigger and changes direction.
The blue line: This is the yaw component on the car that is trying to twist it to the left. If that gets longer we tend to get oversteer. See picture B? We increase throttle, the blue line gets longer, and we go away from understeer or might even get oversteer if it's long enough.
With a proper combined slip model, we should wind up with something more like figure C. We increase forward tractive force, the lateral force (green) drops, and the combined (red) force grows and changes direction. My illustration shows the combined force being shorter, but really I meant it to grow instead. However, the result is that the blue line gets shorter, meaning less lateral force or yaw torque in the car's coordinate system, and we get understeer instead of oversteer.
The combined slip model in LFS seems quite good to me. At least as good if not better than any other sim I've tried. Before patch Q it was, umm.., not so good, but now it's quite super. However, there is probably room for improvement here and what we see with the front wheel drives is likely proof that something is just a little bit amiss there still.
It's still my favorite sim to drive though, aside from my own of course
The new sounds are just fantastic and now that I've got a G25 wheel I can finally use FFB without getting all the rocking of the wheel around the center. The FFB is very good in LFS. FFB has kind of ruined some of the other sims for me in comparison, quite frankly
Last edited by jtw62074, .
[Takes podium]
I haven't run the FWD cars in LFS very much until quite recently thanks to STCC forcing me to in the quest for virtual stardom. (Becky and Tristan, you guys rock!)
I definitely see what you're talking about now with on-throttle oversteer increasing as you increase differential locking, and getting really quite extreme with the locked diff. With the curiousity piqued I decided to try an FWD in my system to see how it reacted.
The car is not the same as anything in LFS of course and I removed camber effects in order to eliminate camber changes as a variable. The test car was 3280 lb with a bit under 450HP at the wheels. It is pretty stiffly sprung and slung very low with a CG height of only about 15 inches, so weight transfer effects are lower than in most race cars, probably. The weight distribution is 51.9% front and 48.1% rear. Again, this may be very far off from the LFS cars. I don't recall off hand what they are like, but it seemed they were much more front heavy than this.
My anti-rollbars were 300lb/inch front and 100lb/inch rear. The LFS cars IIRC have the split biased toward the rear. I'm not sure if that effected the results much. I wouldn't expect it to, but without testing that I can't say definitively. Front spring rates were a bit higher than the rear too.
I tried two sets of tires, both with rather high cornering stiffness. The first set had grip comparable to a performance street tire at about 1g, but with quite a bit higher cornering stiffness (peak force slip angle was around 6-7 degrees). Something along the lines of some really low profile tires. The second set was a nice set of racing slicks that pull about 1.4-1.5g and peak at around 8-9 degrees. These are pretty typical of road racing slicks in many series.
The basic results were largely the same with both sets. With an open diff plus 150 lb-ft of preload (so not totally open) laying in to the throttle even a little bit in a corner just spins up the inside wheel with both sets of tires, not surprisingly. Cranking the preload up to 1500 lb-ft improved turn in on light throttle a fair amount and acceleration out of the corners considerably. Completely locking the diff (15000 lb-ft preload) was similar to this (1500 lb-ft was nearly locked anyway).
Right. On to the infinite flat expanse of green background color for some skidpad-like testing
Here the car was driven at a constant speed of around 40-60mph or so and steered just to the point where the front tires had peaked (7-9 degrees depending on the tire). The understeer gradient, a numerical description of the SAE definition of the amount of understeer or oversteer, is shown on screen. Because of this the results were not subjective (based on my feel or impression) at all.
With both sets of tires, any amount of throttle increase at all caused the understeer gradient to climb. Even just a tiny increase in throttle resulted in this 100% of the time. As throttle increased further the understeer climbed right along with it. Flooring the throttle of course caused it to skyrocket as the front wheels spun.
The test was repeated by steering the front tires way past the peaks up to 20 degrees slip angle. Perhaps the tires pulling "forward" towards the inside of the turn might cause it to oversteer, right?
Nope. The results were the same. Any increase in throttle at all increased understeer.
It should be noted that as the CG is raised the effect would be even more pronounced. I suspect rather strongly that the front/rear weight distribution would not change this tendency at all.
In LFS, my guess is this is most likely a combined slip thing causing it. Regardless of the slip angle, any traction (throttle) increase should reduce the lateral force which would wind up with greater front slip angles, whether the diff is locked or not. The forward traction vector's lateral component does not make up the difference and cause oversteer or any decrease in understeer at all, at least not out to 20 degrees slip angle at the fronts which is twice as much steering as needed to maintain the circle.
Regarding the Formula cars at high slip angles being too easy to catch: I'm a data person of course so would prefer to see numbers, but if something is wrong here I strongly suspect it's due to a lack of variation in downforce with yaw rather than any slight problem with combined forces in the tire model (as this is improved, drifting has always become easier rather than harder in my experience). I.e., as you increase yaw angle the downforce can increase a little bit out to a few degrees slip angle (that's right, it can increase; That's one of the things those little vertical fins on the front wing do. I'm not sure about the rears though.) Going beyond some point likely causes a decrease in downforce which is probably missing, or not pronounced enough.
[Steps down from podium]
Last edited by jtw62074, .
All of the above
I like the new sounds! Great improvement! I think this is the first time I've really cranked up the sounds since the original F08. So far I've only tried three of the cars, but am looking forward to trying the rest too!
Great job!
Great job!
This is some interesting reading:
"A Global Intelligence Briefing for CEOs"
http://www.performancesimulations.com/files/GIB.rtf
"A Global Intelligence Briefing for CEOs" by Herbert Meyer
(About Herb Meyer: served during the Reagan administration as special assistant to the Director of Central Intelligence and Vice Chairman of the CIA’s National Intelligence Council. In these positions, he managed production of the U.S.National Intelligence Estimates and other top-secret projections for the President and his national security advisers. Meyer is widely credited with being the first senior U.S. Government official to forecast the Soviet Union’s collapse, for which he later was awarded the U.S. National Intelligence Distinguished Service Medal, the intelligence community's highest honor. Formerly an associate editor of FORTUNE, he is also the author of several books. Herbert Meyer P.O. Box 2089 Friday Harbor, WA 98250.)
"A Global Intelligence Briefing for CEOs"
http://www.performancesimulations.com/files/GIB.rtf
"A Global Intelligence Briefing for CEOs" by Herbert Meyer
(About Herb Meyer: served during the Reagan administration as special assistant to the Director of Central Intelligence and Vice Chairman of the CIA’s National Intelligence Council. In these positions, he managed production of the U.S.National Intelligence Estimates and other top-secret projections for the President and his national security advisers. Meyer is widely credited with being the first senior U.S. Government official to forecast the Soviet Union’s collapse, for which he later was awarded the U.S. National Intelligence Distinguished Service Medal, the intelligence community's highest honor. Formerly an associate editor of FORTUNE, he is also the author of several books. Herbert Meyer P.O. Box 2089 Friday Harbor, WA 98250.)
I do diff preload in my sim and quite frankly there's not a whole lot of difference between having it and not having it. There are some subtle things like stability under braking being improved a little bit, but it's not very noticeable. The scenarios bought up on this subject are under full cornering. In these situations the torque bias ratio is overpowering the preload anyway which often completely nullifies its effect so long as you don't have really tremendous weight transfer across the axle.
An LSD is applying torque to each wheel in an attempt to lock them together like a clutch. Preload is very easy to model. All you do is calculate the locking torque between the wheels on that axle given their input torques and the locking percentage or torque bias ratio, which seems to me is already done in LFS properly and is fairly straightforward. Preload just says that there's a certain minimum amount of torque that will always be applied. If the locking torque from the torque bias ratio/locking percentage ends up being smaller than the preload torque, you use the preload torque instead. Outside of the preload torque envelope (when locking torque is greater than the preload) the preload has 0 effect, such as in hard cornering. You use one torque or the other, not both at the same time.
A locked axle is not tough to model. All you pretty much have to do is make sure both tires are turning the same speed. That'd be pretty hard to get wrong, especially if you treat the entire thing as a single rotating object.
I haven't tried a FWD in my sim in years, certainly not with the current model, so couldn't comment on whether it behaves the same way. If there's something a bit off though it's probably in the tires indeed. Mine feels somehow a bit different to drive than LFS, likely because of the tire model. Combined slip modelling probably has a lot to do with it and becomes rather important to get right with a locked axle. I can run around with a locked axle in my deal all day long and it doesn't suddenly turn into snap oversteer under throttle everywhere (RWD), even with generous power, no downforce, and so-so tires. It's just not as critical in my model for some reason.
If I were to look for a problem here it'd be in the combined slip workings of the tire model first. It's very good now and vastly improved since a year ago, but perhaps not quite 100% on just yet. I could be wrong though
An LSD is applying torque to each wheel in an attempt to lock them together like a clutch. Preload is very easy to model. All you do is calculate the locking torque between the wheels on that axle given their input torques and the locking percentage or torque bias ratio, which seems to me is already done in LFS properly and is fairly straightforward. Preload just says that there's a certain minimum amount of torque that will always be applied. If the locking torque from the torque bias ratio/locking percentage ends up being smaller than the preload torque, you use the preload torque instead. Outside of the preload torque envelope (when locking torque is greater than the preload) the preload has 0 effect, such as in hard cornering. You use one torque or the other, not both at the same time.
A locked axle is not tough to model. All you pretty much have to do is make sure both tires are turning the same speed. That'd be pretty hard to get wrong, especially if you treat the entire thing as a single rotating object.
I haven't tried a FWD in my sim in years, certainly not with the current model, so couldn't comment on whether it behaves the same way. If there's something a bit off though it's probably in the tires indeed. Mine feels somehow a bit different to drive than LFS, likely because of the tire model. Combined slip modelling probably has a lot to do with it and becomes rather important to get right with a locked axle. I can run around with a locked axle in my deal all day long and it doesn't suddenly turn into snap oversteer under throttle everywhere (RWD), even with generous power, no downforce, and so-so tires. It's just not as critical in my model for some reason.
If I were to look for a problem here it'd be in the combined slip workings of the tire model first. It's very good now and vastly improved since a year ago, but perhaps not quite 100% on just yet. I could be wrong though
Last edited by jtw62074, .
I'm curious. Did you run the same inflation pressures on the street as you did at the track? Were the pressures lower than the "normal street tires" you'd otherwise have probably used?
Thanks.
No, I had no idea. You sure? I'm not too worried about it quite frankly, but maybe need to have the guys take a look at it.
Ah, yes, it's just bad timing I think. The new guys are in Slovakia and we so far have not all gotten together in the Netherlands to work (Trencin, SK, instead a couple of times per year). The last event I went to I think was the 1/8 Euros in Luxembourg. This was quite some time ago though. Probably before we released V1, come to think of it.
I'd be happy to pay for additional content too. Whatever they put out and want to sell I'll most likely buy
In the mean time though, there is so much content already that I still haven't even driven all the tracks, much less in reverse, nor have I had serious seat time in all the cars. So I've got plenty of unexplored areas in LFS for quite some time yet. On the other hand, additional content is always welcome.According to my engine simulation transient effects like this don't last long at all. The pressure traces in the cylinders, intake and exhaust manifolds reach a stable point in two or three engine cycles (four to six revolutions) usually at the most, and this is slamming the engine from idle/closed throttle @ 9000 rpm to full throttle @ 1-2000 rpm in an instant, so it's a much more severe case than simply opening the throttle.
4 revolutions at 2000 rpm = 0.12 seconds
4 revolutions at 6000 rpm = 0.04 seconds
This is an instantaneous full closed to full open throttle change though (vice versa is about the same). In reality though, even "booting it" quickly will take some fraction of a second. I'd be surprised to see any increase in torque at all during this period. Depending on the precise instant you changed the throttle, some cylinders will probably get a slight increase in volumetric efficiency (torque), while others get a slight decrease. Here you're probably talking more about instantaneous torque changes over the course of a cycle rather than the average torque you feel in your butt or measure on the dyno over several cycles.
My guess is that it would be immeasurable on a dyno and not something you'd feel in your behind. The momentum effects in any single intake track might increase, but they are lost the moment the intake valve closes which causes the intake gases to come to a screeching halt at the valve. You'll have a little different reflection dynamic the next time the intake valve opens and closes, but again, this would be different from one cylinder to the next. Some will most likely lose while others gain, and all a very small amount I suspect...
Ahha.. Good threads. Thanks
FGED GREDG RDFGDR GSFDG