Why would the results from a 120 degree slip angle be any different from a 60 degree one?
I don't have the files handy at the moment so can't double check, but it seemed to me they dropped very much more rapidly than this. I could be mistaken though.
Most likely, LatPeak shifts the curve to the right as load is increased. However, at any given load there is a peak and some point after that where the force drops to 67% of that peak. LoadSens is load sensitivity, another modifier that varies friction coefficient with load. With this too, however, you have a peak and some point after that where the force drops to 67% of the value. DryLatLong is another multiplier. Same thing there though. There's 100% at one point and 67% at another. I'm not sure what DropoffFunction does, but I very much doubt it fixes anything up.
I don't think there's any magic stuff going on there. When driving it I can immediately tell that the force curves drop off quite a lot after the peak in all of ISI's sims that I've tried, with the exception of the Mustang and perhaps a couple of other cars in GTL. It's easy to tell. Kick the car a bit sideways going into a corner and you'll slide right off the outside of the curve where if you were running at the peaks you would have rather easily made it. Try countersteering a little bit (not tons of lock) and you'll see the cars spin faster than if you left the steering alone or steered into the spin instead. If steering into the spin = easy spin correction then the curves definitely are dropping quite a bit.
Doug wrote back and said the tire type and data on pressures must have been proprietary since they weren't included in the paper, so that can't be made public. Sorry!
I didn't see any mention of the pressures in the paper either, so I just sent an email to Doug about it. Will see what he says and get back to you.
Subjectively speaking about grip loss after the peak, I suspect we get as much or more of our feeling about what the car is doing from not just the acceleration, but also "jerk," the derivitive of acceleration. "Jerk" is to acceleration as acceleration is to velocity, or velocity is to position. It's how quickly acceleration is changing. E.g., at time 0 acceleration is 0g, at time 0.25 it's 0.5g, at time 0.5 it's 1g. The acceleration is changing 2g/second. That's jerk.
When cornering you feel the 0.5g or 1g, but you also feel how quickly the acceleration is rising. In the case of a set of car tires that peak at 1g, you have positive "jerk" as you're on your way up to that 1g, but before you hit it and the curve starts rolling off into the peak, the "jerk" changes and eventually becomes 0 as you finally hit that 1g. To our rear ends this feels very much like there's been a sudden loss in traction, but there doesn't need to be any loss at all in order to get this impression. It's a bit like sliding sideways into the grass. The acceleration drops considerably extremely quicly (very high negative jerk). Many people swear their cars actually speed up when they hit the grass, which of course is not the case, but our "jerk meter" seems wired into all of this and produces that sensation. That's my theory anyway
A jerk curve would essentially be the slope of the lateral force vs. slip angle curve. Drivers' descriptions of how one tire feels very forgiving while another one feels like it loses grip abruptly seem to match this concept. If the force curves rise and then very gingerly roll off into the peak at 12-15 degrees slip angle, the jerk curve is low and the driver feels that the tire is very forgiving and easy to drive. Switch to an F1 tire that peaks at 4 degrees slip, where the force rises very rapidly and then very quickly rolls off into the peak, and the driver will say that the tire is very snappy, tricky to drive, and more importantly, he might say that the tires lose grip very quickly, even if the force is actually totally flat after the peak. The jerk is considerably larger (not the jerk driver, the jerk = derivitive of acceleration ).
When cornering you feel the 0.5g or 1g, but you also feel how quickly the acceleration is rising. In the case of a set of car tires that peak at 1g, you have positive "jerk" as you're on your way up to that 1g, but before you hit it and the curve starts rolling off into the peak, the "jerk" changes and eventually becomes 0 as you finally hit that 1g. To our rear ends this feels very much like there's been a sudden loss in traction, but there doesn't need to be any loss at all in order to get this impression. It's a bit like sliding sideways into the grass. The acceleration drops considerably extremely quicly (very high negative jerk). Many people swear their cars actually speed up when they hit the grass, which of course is not the case, but our "jerk meter" seems wired into all of this and produces that sensation. That's my theory anyway
A jerk curve would essentially be the slope of the lateral force vs. slip angle curve. Drivers' descriptions of how one tire feels very forgiving while another one feels like it loses grip abruptly seem to match this concept. If the force curves rise and then very gingerly roll off into the peak at 12-15 degrees slip angle, the jerk curve is low and the driver feels that the tire is very forgiving and easy to drive. Switch to an F1 tire that peaks at 4 degrees slip, where the force rises very rapidly and then very quickly rolls off into the peak, and the driver will say that the tire is very snappy, tricky to drive, and more importantly, he might say that the tires lose grip very quickly, even if the force is actually totally flat after the peak. The jerk is considerably larger (not the jerk driver, the jerk = derivitive of acceleration
).
Last edited by jtw62074, .
And at the highest slip angles it's 67%. I'd call that a cliff.
The only tire data I've seen that drops off significantly after the peak in slip angle on dry surfaces are big truck tires. They tend to have very large tread blocks that can do all sorts of nasty things at high slip angles which probably explains it to some extent, but in car tires this just isn't the case. Inadequate torsional stiffness in the belts might have something to do with it too, although that's pure speculation on my part and not based on any data.
In the wet there is usually quite a large drop off in both car and truck tires, but even a truck tire in the rain doesn't drop off as much as GTR/GTR2's racing slicks do in the dry. Longitudinal grip (acceleration/braking) usually does drop off at high slip ratios, but again, not generally clear down to 67% of the peak value.
Here's some data that is an exception to the rule, but again, these are giant bus tires (see page 24):
http://www.edccorp.com/pdfs/WP2005-1.pdf
Even these big guys don't really show much of a drop off in lateral grip with slip angle (page 23). In some situations, even in the wet, the grip just keeps climbing slowly "after the peak" rather than falling off. Anyway, these are more along the lines of what you're driving in GTR/GTR2/rFactor, at least in terms of longitudinal slip. They're more like big bus tires in the rain with extra traction than any car tire I've ever heard of.
This is precisely why when you countersteer just a little bit after hanging the back out a touch, the car, instead of straightening up or slowing down its rotation, just spins faster. Your real car would do that too if that's how the curves looked. It doesn't because the real curves don't look like that. That should be a dead giveaway but most developers miss it 
A friend of mine and I did indeed encounter this in a light pickup truck in the rain once. We went a little too quick through a corner, the tail came out a little, he countersteered nice and smoothly, and the truck just accelerated around more quickly as he steered and wound up pointing backwards. Thankfully we slowed down quite a bit before hitting the little tree, sending the back of my head through the window behind us which exploded into a million pieces
But anyway, that's a pickup truck tire in the rain, not a car tire or racing slick in the dry. Big difference!Here's a paper written by a friend of mine (Milliken, author of "Race Car Vehicle Dynamics):
http://does.eng.buffalo.edu/pu ... ons/Kasprzak.SAE.2006.pdf
Dr. Kasprzak, Doug Milliken, and I attended an SAE conference together last October and had quite an interesting talk or two about tires. They are involved in a lot of tire testing (they have their own library of tire data that I'd kill to get my hands on) and definitely know their stuff. These are the guys that teach racing engineers in Nascar and so on what they didn't know they didn't know about vehicle dynamics and tires. Their paper there shows some real tire data that goes quite a bit past the peak force slip angle. See figures 6 and 8. Even the longitudinal curves didn't drop on these tires, at least out to whatever slip ratio the measurements were run to. This is somewhat uncommon from my understanding, but has been known to happen with some tires provided they aren't smoking hot from wheel spin/lock.
For those curious, figure 12 shows tire pressure effects :eclipseeh
Last edited by jtw62074, .
The only "concrete" info (in quotes because I don't recall the numbers) I can recall on wings and how downforce is effected by yaw angle was from an SAE conference I went to last October. One of the presentations I sat in on was a CFD (computational fluid dynamics) deal where they were showing the latest aerodynamic modelling on a formula car. Very impressive stuff to be sure!
Anyway, you know those little vertical fins that you see somewhat outside of the center on the front wing of a lot of formula cars? The thing I remember most was a CFD comparison before and after adding a pair of those little fins. What tended to happen when yaw angle was increased to 4 degrees, which is right at the peak of the lateral force curve (tire slip angles would be about 4 degrees then too, pretty much), was that a set of low pressure "bubbles" would come off the outside half of the wing and stream into/around the outside pod. It was neat seeing it in CFD. When he showed the fins the bubbles coalesced and more or less disappeared, restoring pressure in that area.
The result was a slight increase in downforce. However, I don't know if there was an increase from 0 degrees yaw or if the fins just increased downforce at 4 degrees when compared to 4 degrees and no fins. I want to say there was indeed an increase from 0 degrees yaw, but don't really remember for sure. The impression I got was that the tiny, extra downforce (14 or 18 lb comes to mind, but I could be remembering wrong) wasn't generated at the wing, per se, but as a result of how the airstream went around the side pod afterwards, slightly forward of the car's center of gravity and thereby increased front downforce. A slight change in pressure on any top or bottom surface anywhere on the car changes downforce, so changing the front wing can actually effect more than just that part. It changes the airflow at the rear too. I don't know how much though. I've got quite a lot of aero data on one formula car and they don't specify a change in rear downforce when the front wing flaps are adjusted, so perhaps it's not much or enough to worry about.
In conclusion, I don't know how much yaw angle effects downforce on wings. Without end plates I'd tend to think the downforce would decrease with any yaw angle at all, but don't really know for sure. Remember that the airflow around the rest of the car is also influencing downforce, and that is changing with yaw too. With end plates I wouldn't be surprised to see it go either way over a small yaw angle range. Indycars on super speedways tend to spin in a real hurry when you get the slip angle much past the peak, so it looks to me like downforce plummets pretty quickly after a point. That's at the rear though where the airflow has gone all crazy over the body compared to the front wing which still has clean air. So it's a body/wing interaction. There could very well be a small region from 0-5 degrees or so where there's a slight increase, but I don't know for sure. If there is, I doubt it's much though (1-2% or less if I had to take a wild guess). Measured data has gone contrary to my intuition before, so take that with a grain of salt
BTW., the CFD simulation covered probably just a couple of seconds of simulated time. IIRC, the computation took 120,000 CPU hours, which would take over 13.5 years on one PC!
They ran it on a massive CPU cluster and it took a week straight, 24 hours a day. All for a few seconds of simulated time. The result was sure pretty to look at, although I don't know how practical that really is. EDIT: If each CPU did 1.3 billion calculations per second (my P4 3.6Ghz does that), 120,000 CPU hours would be 561,600,000,000,000,000 calculations (561.6 quadrillion!)
Last edited by jtw62074, .
Manufacturers have the option of quoting power figures as SAE standard. When they do this, a neutral SAE official attends and oversees the test to make sure there's no monkey business going on. As a result, the figures manufacturers are quoting ought to be quite accurate, at least as far as what the engine is doing on a test stand.
I've had a lot of people use rolling road dyno data in my drag racing simulation and report back very accurate results, so I'm a little surprised to see that these dyno shops overquote the power so frequently. They'd need to alter the software that comes with the dyno in order to do it, I'd think (maybe there's an ini file or something that allows it).
Still, the reasoning makes sense for overquoting, so perhaps it's not all that surprising or infrequent.
I've had a lot of people use rolling road dyno data in my drag racing simulation and report back very accurate results, so I'm a little surprised to see that these dyno shops overquote the power so frequently. They'd need to alter the software that comes with the dyno in order to do it, I'd think (maybe there's an ini file or something that allows it).
Still, the reasoning makes sense for overquoting, so perhaps it's not all that surprising or infrequent.
I think it'd be annoying. I want to get to the race, not spend 2-3 minutes doing a slow lap on a track I've been running race after race all evening anyway, plus running for countless laps begging for a restart after a race ends, etc.. So -1 from me in general, but for the leaguers that want it, why not? It might be fun now and then on a big race.
One thing that drives me nuts about other sims is that racing is actually rather rare on public servers. Far too many times in rFactor and GTR I've run practice, qualifying, warmup, and whatever else for an hour only to do a measely 8 minute race at the end of it. Sit at the PC for 4 hours and spend 20-30 minutes actually racing. Wash, rinse, repeat. I don't need to practice for another hour again on this track tonight in preparation for another 5 lapper. Ugh.
LFS for me shines largely because the racing is much more immediate and common. Adding a parade lap to me just pushes things away from that, albeit by only one lap, so I wouldn't cry too much about it, I suppose
Might as well have the option to do it though if most people want it.
One thing that drives me nuts about other sims is that racing is actually rather rare on public servers. Far too many times in rFactor and GTR I've run practice, qualifying, warmup, and whatever else for an hour only to do a measely 8 minute race at the end of it. Sit at the PC for 4 hours and spend 20-30 minutes actually racing. Wash, rinse, repeat. I don't need to practice for another hour again on this track tonight in preparation for another 5 lapper. Ugh.
LFS for me shines largely because the racing is much more immediate and common. Adding a parade lap to me just pushes things away from that, albeit by only one lap, so I wouldn't cry too much about it, I suppose
Might as well have the option to do it though if most people want it.
Yes, I think it's right. If it were too low the top speed of the cars would also be too high.
Been gone for a few weeks with work. Sorry for the late reply.
Indeed, in the case of an open diff with no preload they should be the same just as you thought. The rolling resistance and whatever other mysterious factors someone would interject into the picture that might be at play should effect the differential "just so" to produce equal forces at the tires as you expected (once the car is in a nice, steady state without blipping the throttle/brakes or shifting weight around with the steering wheel of course; those are transients and alter the picture a bit as discussed elsewhere briefly). The slip ratios would adjust themselves to equalize the forces. Good detective work
This is a fairly minor detail though and doesn't throw a big wrench into LFS physics by any means, so I'd hate to see folks run with this and say "the diff modelling isn't right and this is why this and that, etc., is wrong" ad infinitum. So what if the open diff might not be truly, totally open? Who really wants an open diff, anyway?
In a racing sim, you want a good LSD model, and LFS has that
Last edited by jtw62074, .
:iagree:
As others pointed out, the main reason for heel-toeing is that when you push in the clutch the engine rpm drops (or remains pretty much unchanged for an instant). When you downshift and release the clutch again, the clutch plates speed up the engine a bunch in a very short period of time until the clutch plates are locked together again. Action/reaction: If the engine speeds up very quickly over a fraction of a second as a result of the clutch plates, the driven wheels will also slow down very quickly, acting just like you stabbed the brakes momentarily, but only at the rear wheels. If you're already at the limit of braking then you'll go over it on the rears, and if you're turning at the same time you'll upset whatever balance you had going into the corner. It's a lot like you flipped the brake bias momentarily toward the rear for a fraction of a second until the engine rpm rose to where it should be in the new, lower gear.
This may or may not upset the car/tire combination. Some won't care much while others might. If the engine has extremely low inertia it can change speeds quickly while hardly hitting the rear wheels. In such a car you might not need to bother with it, especially if the gear ratios are close together.
In Atari's Hard Drivin' way back in the day, which was really the first real sim, I didn't know about heel toeing and it hadn't occurred to me. In that sim I found that letting the clutch out slowly during the downshifts was enough to keep things stable. I do this in LFS too rather than trying to operate both the brake and gas with one foot. In the arcade F355 challenge I did the heel toe thing because the pedals were set up a bit better for my feet than the G25 ones are. In LFS now I generally shift directly to the gear I want and then very slowly let out the clutch so the engine rpm rises gingerly to the new, higher rpm. In reality this is probably a bad idea as it causes a lot of unnecessary clutch wear. However, I don't have to replace the clutch in LFS so I'll keep right on doing this for a bit
This may or may not upset the car/tire combination. Some won't care much while others might. If the engine has extremely low inertia it can change speeds quickly while hardly hitting the rear wheels. In such a car you might not need to bother with it, especially if the gear ratios are close together.
In Atari's Hard Drivin' way back in the day, which was really the first real sim, I didn't know about heel toeing and it hadn't occurred to me. In that sim I found that letting the clutch out slowly during the downshifts was enough to keep things stable. I do this in LFS too rather than trying to operate both the brake and gas with one foot. In the arcade F355 challenge I did the heel toe thing because the pedals were set up a bit better for my feet than the G25 ones are. In LFS now I generally shift directly to the gear I want and then very slowly let out the clutch so the engine rpm rises gingerly to the new, higher rpm. In reality this is probably a bad idea as it causes a lot of unnecessary clutch wear. However, I don't have to replace the clutch in LFS so I'll keep right on doing this for a bit
Last edited by jtw62074, .
A 2000 lb car with a frontal area of 20 ft^2, a 0.36 drag coefficient on a 20 degree downgrade would reach terminal velocity in neutral at 194 mph (318kph) (ignoring rolling resistance losses, which are pretty miniscule at this speed). The hill you're talking about can't possibly be that steep
(The steepest street in San Francisco is 17 degrees according to a very quick search.)http://en.wikipedia.org/wiki/Drag_equation
Others can check my math here. I did this quickly so might have made a mistake:
F = 0.5 * air_density * drag_coef * frontal_area * vel^2
air_density = 0.00233 slug/ft^3
In our case, we need F = sin(20) * 2000lb = 684 lb drag to balance the car and keep it at a constant speed.
Setting F = 684:
684 = 0.5 * 0.00233 * v^2 * 7.2
v^2 = 684/0.5/0.00233/7.2
v= (684/0.5/0.00233/7.2)^0.5 (square root)
285 ft/sec = 194 mph
Seems reasonable as if you take the car in free fall with F = 2000lb then you get 323 mph (530 kph).
Anyway, mistake or no mistake, I don't think those slopes are anywhere near 20 degrees. The point is, there is of course some speed at some slope where the car is balanced at a constant speed due to aero drag and rolling resistance. If you are going faster than this when you shift to neutral, the car will slow down. If you're going slower, it will speed up. If you're going just the right speed, it will stay right at that speed, which is probably what you were doing at 145kph in LFS.
Nothing is probably wrong here at all
I doubt it. Air resistance is modelled very accurately using the drag equation in the above link. All you need is air density (easy to calculate), a drag coefficient (make one up for your fictional car), and the speed it's moving. Presto. 99.99% perfect aero drag model
That's dreadfully slow for that power to weight ratio. Running 260hp with an 876kg car through my drag racing simulation, Straightline Acceleration Simulator, reveals wildly varying times depending on tire traction, wheelbase, and center of gravity height.
With driven tires having a constant friction coefficient of 1, cg height of 22 inches, 106 inch wheelbase, 260hp @ 5000rpm and 340ft-lb @ 2940 rpm, Mucie M21 4 speed manual, gets this 12.84 sec @ 112.0mph (assuming 82% drivetrain efficiency and a 0.4 second gear change time).
That's with a drag coefficient of 0.42 and frontal area of 24 ft^2, which is more along the lines of a big, boxy, American sedan from days of old than what your Dad probably is running. Just to give an idea of how much the aerodynamics alone effect the picture, if the drag coefficient is changed to 0.36 it gets 12.78 @ 113.6mph (much less of a difference than I expected, actually).
It's traction limited all the way up to 60 mph, btw.. Drop the tire friction coefficient to 0.9 and we get 13.33 @ 111.1 mph. Reduce it to 0.85 and we get 13.63 @ 110.5 mph. The point is, just a tiny change in grip here produces a massive change in time.
If we give it as much traction as it wants (1.19 friction coefficient), the time plummets to only 12.29 @ 112.7mph. With better gearing and more grip, it'd go even quicker.
You sure your dad's car is really making 260bhp?
I'm only using a 2.73:1 differential here and the Muncie's first gear ratio is only 2.2:1, so rather tall gears. He ought to be spinning the tires like crazy up to 60+mph if he wants too. Am I right? If not, the 260bhp number is probably just some bench racing fantasy. I wouldn't say anything is wrong with LFS on this one. The LFS car, especially the particular tires it uses, are different enough from your dad's to account for the time and speed difference quite easily. Power to weight ratio is, as others pointed out, only part of the picture. Most important is traction in this case, it appears.
Last edited by jtw62074, .
It can step out during transient periods. For instance, the car might have been in a tight turn with lots of weight transfer and loads of inside rear wheel spin. If you suddenly straightened the car out, eliminating the weight transfer, the inside rear tire will remain at a very high slip ratio, producing a lot of traction force. The torque bias ratio will determine the torque that is bringing the wheel speeds back together, but it's finite and doesn't result in the wheels immediately locking together.
The diff's locking torque is acting to slow down the (formerly) inside rear and speed up the outside rear. So for a brief period of time the torque bias ratio is indeed being exceeded. This is one of the transient situations. Here, after you straighten the car out, you get a kick to the right as the inside wheel slows back down.
The driver says, "the inside wheel suddenly regained its traction in a big hurry and almost spun me back the other direction."
Anyway, the Torsen should be the same way just by virtue of the fact that they describe it as having a fixed torque bias ratio just like a clutch pack diff has. So it shouldn't do anything different as far as what is happening at the tires. Internally of course it's handling things its own unique way, but mathematically speaking the diff can be considered to be a magical black box unless you really want to model every shaft and gear in the device. That's not really necessary for vehicle dynamics sims any more than attempting to model every molecule or square inch of rubber in the tire. Once you have the characteristics, you're good.
It doesn't limit the speed difference. If it did, as soon as the speed difference hits its maximum, the torque bias ratio is free to go to infinity just like a locked diff can. Torsen states the torque bias ratio is constant, though, so this can't be the case.
Really the main thing you have in any diff is a locking torque that's trying to bring the wheel speeds back together into a solid axle. Whether it has literally locked or not doesn't really matter all that much. The maximum torque bias ratio is not allowed to be exceeded by the mechanism. Sometimes that means it will be locked, other times it won't.
The clutch pack and torsen effectively work the same way in the end from what I've gathered so far. The locking torque is a function of the two road reaction torques described in another post I wrote a couple of weeks ago or so. The only difference I'm aware of is that the torsen does it through a clever gear arrangement instead of having clutches sliding, which could be subject to changes in friction coefficient (which effects locking torque) due to wear, heat, and so forth. I'd expect the torsen to maintain its torque bias ratio virtually perfectly. It also likely produces the same torque bias ratio under power and coasting, but I'm not 100% sure on that. There's also no preload at all apparently given the description of what happens when you lift one wheel into the air. It then acts like a fully open diff and the car doesn't go anywhere.
Viscous diffs produce a locking torque as well. However, the locking torque isn't a function of the input torques like it is with the other two diffs discussed here. Instead, it's a function of slip velocity, or how much faster one shaft is spinning than the other, as well as rather strongly dependent on the temperature of the fluid which usually changes quite significantly during all this.
Last edited by jtw62074, .
It didn't seem to me that that's what he was suggesting, but yes, that sounds good
The torsen does "slip" in the sense that the wheels are allowed to differentiate in order to maintain a given torque bias ratio, which is precisely what a clutch pack differential does. The mechanism for doing this is very different of course, using gears instead of allowing anything to literally slip internally, but the end result is the same from what I've gathered so far.
Right, yes. That's an important distinction to make.
Imagine you and I are holding on to a shaft. You're the front or rear wheels and I'm the diff, if I twist with __ torque you'll get the same torque.
After doing a bit of checking, it appears the Torsen diff, mathematically speaking, is identical to the clutch pack diff without the preload. So for all practical purposes, until the preload was added to the diffs recently, everyone using a clutch pack diff has already been driving a Torsen diff.
The Torsen appears to do exactly the same thing as the clutch LSD, it's just doing it through a beautiful arrangement of gears without any clutches or preload. Mechanically it's a great design, but for simulation purposes at this level it's identical.
This was indeed done perhaps ten or fifteen years ago by a company whose name escapes me at the moment. They had a universal automotive/engine calculator for sale that did relatively simple suspension calculations and all sorts of handy things. The last thing I recall them adding to it was an engine simulation that was indeed AI driven, very likely by a neural net. The benefit is that it's lightening fast. The drawback is that in order for it to be accurate you need to hand it a lot of test data as with any ANN, generally. However, that sort of data is no big secret and is easy to come by, so it's not really that far fetched of an idea.
I don't know how well it worked or how accurate it was, but it was extremely fast at finding a result. Just a tiny fraction of a second.
I don't know how well it worked or how accurate it was, but it was extremely fast at finding a result. Just a tiny fraction of a second.
I'm not too sure about that. There should be, I suppose, but have seen no mention of it being important enough to consider anywhere. If the change is small enough it wouldn't be noticeable so might be disregarded. Add in that temperature, clamping force, sliding velocity, and perhaps other factors might be changing the friction coefficient(s) too in an unknown way and it might not be worth considering. I don't bother with it and I'm not aware of any full vehicle simulations used in engineering that do either.
FGED GREDG RDFGDR GSFDG