Tire analysis with MoTeC i2

[tarnheld]

I made a small experiment to check the different tire compounds. I did an arcade single race on Motegi Oval with the Scion FRS courtesy car, AT, no steering, full throttle with DS3 X button until hitting a wall. All tests were on GT6 version 1.17 . You find the different MoTeC logs in the attached zip.

I then analyzed the tire response by calculating slip ratio from the wheel speeds and the vehicle speed:

SlipRatio = (Wheel Speed - Vehicle Speed) / abs(Vehicle Speed)

You can read more about it here. The essential fact is that Slip Ratio is positive when car is accelerating forward and negative when braking/accelerating backward.

Here is the outcome from start of log until just before hitting the wall.

Red = CH
Green = CM
Light Blue = CS
Blue = SH
Violet = SM
Pink = SS
Yellow = RH
Orange = RM
Green = RS

You will notice that the slip ratios for tires above CS are below zero, indicating deceleration of the car, yet we have strictly positive longitudinal G Force. So something is fishy about the wheel speeds. For all tires above CS the rear driven wheels are slower than the vehicle speed, yet the undriven front wheels (not shown here) are at vehicle speed, with a slip ratio of zero indicating free rolling.

The engine RPM seems to indicate that the rear wheels are directly coupled to the engine, as they have the same RPM at the same rear wheel speed for all tires. Yet the vehicle is travelling at different speeds for same RPM, generally faster for higher grade tires.

One explanation could be that the tires have different diameters. The wheel speed can be calculated from the angular wheel velocity times wheel diameter. This would mean that (with comparable slip ratios) that RH tires are 2-3% smaller in diameter than CH tires. But the front wheels don't show this difference, so they seem to be equal in diameter.

I plan to investigate further, with more powerful engines and different drivetrains. If you have done similar experiments or can explain the phenomenon, please post, but keep on-topic.

EDIT: here are the numeric values from a slice in the middle from CH at the left to RS at the right, you see that the front wheel speeds are almost the same as vehicle speed, but the rear wheel speeds are less than vehicle speed for all tires above CS:

 

[MatskiMonk]

Great work. I've wanted to look in to this myself, but I've not had the time, it doesn't make sense that the driven wheels are under-rotating like that, over-rotation you might expect. I think drive train maybe a factor, I noticed back in my first tests with the MoTeC software that the Veyrons rear wheels show a faster speed than it's fronts, which might be correct, but I then tested on something RWD and it was the other way around.

 

[azidahaka]

MMh do tire deformation and rim size matter in case the game actually calc them?

 

[gluxoid]

My conclusion after fitting a straight line acceleration model (power, drag) was te same: negative slip while accel. --> Physically incorrect.

 

[tarnheld]

MMh do tire deformation and rim size matter in case the game actually calc them?

Could be tire deformation, but the front wheels should deform too, and they do not show differences in speed with regard to tire compounds. Maybe PD measures tire surface velocity, models tire deformation and measures vehicle speed as undriven wheel speed. Then the test would indicate that higher grade tires expand compress significantly when driven/accelerated. Still the whole vehicle is faster when using higher grade tires.

Here is a nice video of what happens to the driven tires IRL (go to 2:03):



You see that the wheels go faster than the vehicle -- first the wheel turns, then the car moves.

EDIT: Oops got that tire surface speed wrong: bigger tire means higher tire surface speed. So RH tires would be compressed compared to CH tires -- totally against RL.

 

[azidahaka]

More compression more contact patch more grip?

 

[tarnheld]

More compression more contact patch more grip?

But what make the tires compress when they are driven or revolve at speed? IRL they would expand -- front and rear.

By the way the strange wheel speeds would be a measuring artefact, because the angular velocity (revolutions per second) of the driven axle would always be higher than the undriven axle when accelerating -- it revolves more over time. I suppose PD calculates the wheel linear speed by multiplying with the tire's unloaded radius -- a constant. If they really determined the (expanding or compressing) tire radius over time it would be overkill and deliver numbers that are useless for analysis -- but you never know.

 

[MatskiMonk]

I think there's two questions:-

Why are the driven wheels slower, when you'd expect them to be faster.
How does the tyre grade affect that.

May be a small piece of the puzzle worth investigating:-

Let's say you have a lot of down force at the back, and accelerate hard. We know weight shift will transfer load to the back wheels, therefore compressing the tyre slightly, and possibly so does downforce at high speed.... so, in a RWD car, at most points during a run, there is more downward force on the back tyres. This will have the effect of decreasing it's rolling circumference:-

Case 1: smaller radius = faster rotational speed for a given ground speed
Case 2: smaller radius = slower ground speed for a given engine speed

So, the game says, my wheel is turning at 20 rotations per second (rps) (because that's dictated by the engine and transmission), and it knows the distance traveled in one second should be rps(π×Wheel Ø). So lets say a wheelØ of 570mm, each second should cover 35.81m (128.91 km/h). Know lets say, the game says "my tyre is being deformed by 5mm thanks to downforce/weightshift" It now makes a calculation that says 20rps×(π×565mm) = 35.50m each second (127.8 km/h) - and reports this as it's wheel speed....which is Slower.

edit: this would indicate that since the discrepency gets higher with softer tyres, the game is allowing for more compression deformation in racing tyres, than comfort tyres.

I'm not really stating this as the answer ... I'm just stating this as an example of how they might have got the maths right, but still come to the wrong answer.

edit: This would also only be the case if they determine the vehicle speed first, and then work back to the wheel speeds... where as it should be the other way around.

So, a couple of other things to consider:

In the real world, tyres get hot, this increases the pressures and I would think, reduces the deformation, and therefore would maintain a more correct speed. But GT6 doesn't handle tyre temps properly, we know this.

Perhaps one of the 3 tyre variables that @Griffith500 (I think...) mentioned somewhere is this compression factor. "Stiffer" tyres may compress less. This isn't the same error from car to car, so I don't think it simply comes down to which grade of tyre you select.

Edit: Might have time tonight to do a few runs and see if for a given car tyre and indicated speed, changing the downforce alters the driven wheel speeds.

 

[azidahaka]

Greart topic, wel discussed and properly tested.

Not like some other high speed ones that complain with no interest toi understand what is wrong so it can be pointed out. I would signal this topic for front page if i could.

 

[Griffith500]

Since this data is generated from replay data, it is quite possible wheel speeds are calculated from the car's linear speed. This error wouldn't be noticeable in replay playback.

Another possibility is that rolling resistance is being factored in somewhere and affecting the reported wheel speed. It might be worth trying deceleration both in gear and in "neutral" (handbrake and throttle to "engage neutral" if sequential car) to see if the relationship is the same, but slip is likely to be much smaller.

 

[MatskiMonk]

Did a really quick test last night, not that scientific.

Z4M Coupe (RWD). RS/RS.

1st Run: Minimum downforce, front to rear wheel discrepency at Vmax was about 7 mph.
2nd Run: Maximum downforce, front to rear wheel discrepency at Vmax was about 7 mph.
3rd Run: Maximum downforce, suspension dropped and stiffened to the max, front to rear wheel discrepency at Vmax was about 7 mph.

So my guess work from the previous post doesn't hold true, at least in this case. The rear wheels still are reported as turning slower, by about the same amount, no matter what the aero load is.

I've had a couple of other theories which I might test tonight, thanks @Griffith500, I'd always wondered about getting the car into neutral.

 

[Griffith500]

Did a really quick test last night, not that scientific.

Z4M Coupe (RWD). RS/RS.

1st Run: Minimum downforce, front to rear wheel discrepency at Vmax was about 7 mph.
2nd Run: Maximum downforce, front to rear wheel discrepency at Vmax was about 7 mph.
3rd Run: Maximum downforce, suspension dropped and stiffened to the max, front to rear wheel discrepency at Vmax was about 7 mph.

So my guess work from the previous post doesn't hold true, at least in this case. The rear wheels still are reported as turning slower, by about the same amount, no matter what the aero load is.

I've had a couple of other theories which I might test tonight, thanks @Griffith500, I'd always wondered about getting the car into neutral.

The handbrake trick only works something like that (it's muscle memory from the GT3 days); I'm sure you'll figure it out. It doesn't work for FWD cars, of course.

What the game really does is engage the auto-clutch, the gear selection remains the same but "greyed out" - I don't think the drivetrain model is sophisticated enough for it to matter which gear you use, not that it should make a huge difference anyway.

 

[tarnheld]

I looked at a seasonal superlap that i did with the 330 P4 at Suzuka, they are the only replays you can check in the Data Logger. Data Logger shows rear wheel speeds faster. I then fiddled with the wheel speeds to bring them to the Data Logger values. First i tried to multiply wheel speed with a correction factor of 1.05 for RH tires -- that brings the wheel speed in correspondence on straights with full throttle. But if you lift throttle, the wheel speed should go below zero, that didn't happen with the correction factor. I then noticed the strange jumps in wheel speeds when shifting gear and blipping throttle. This also seems unrealistic. So i came up with this weel speed correction:

Corrected driven wheel speed = Wheel Speed * C * Throttle Pos + WheelSpeed * (1 - Throttle Pos)

This is a blend between wheel speed corrected with a factor C (=1.05 in my tests) and normal Wheel speed by throttle input -- when not on throttle it's the normal speed and when on full throttle it's the corrected speed.

So this works out pretty well:

Notice the strange spikes while shifting are gone, but the different grades are still there. Strange thing is, it seems to work with all tires. I have tried it on my hotlap replay and it seems to bring the tire speeds into normal regions.
Here is a little example with the slip ratios based on the new wheel speeds:

This is the last part of the Spoon curve -- you see i have to countersteer at the exit because the rear tires spun. But notice the different spikes when shifting up and down, and the altogether more realistic slip ratios.

So it seems PD takes the throttle -- or something different, maybe engine torque into the tire speed calculations. I will check if this matches the Data Logger later.

I will add some more comparison diagrams later, so now you have to do your own experiments. I encourage you to try it out on a hotlap replay yourself.

 

[Griffith500]

Subtle simplification:

Corrected driven wheel speed = WheelSpeed * (1 + ThrottlePos * (C - 1))

Define K = C - 1, i.e. 0.05:

Corrected driven wheel speed = WheelSpeed * (1 + ThrottlePos * K)

This gives the impression of adding 5% extra wheel speed for full-throttle application. The question is, why would that be necessary?

 

[tarnheld]

This gives the impression of adding 5% extra wheel speed for full-throttle application. The question is, why would that be necessary?

That's the prize-question.

I always thought this has something to do with different wheel/tire diameters, but it seems the MoTeC log wheel speeds are different than the data logger wheel speeds. Why? No clue. I will check later if the corrected wheel speeds match up with the data logger. If this holds true, then i guess PD does something wrong when exporting the MoTec logs. Why this has something to do with throttle escapes me.

 

[azidahaka]

If i remember correctly, back then there was a clutch/throttle glitch that would have allowed cars to accelerate past intended values... Might this be a leftover of the bug correction that messes up with datas? Maybe it just shows in motec because it's badly patched just "ingame"?

 

[tarnheld]

Another curious thing: some of you might remember my comparison of 1.00 vs 1.16, there i also presented a diagram of the wheel speed differences. The strange effects and differences when shifting and lifting throttle between versions are gone when using the corrected speeds!

Here is a diagram of the 1.00 vs 1.16 comparison with corrected speeds:

You see there are no strange jumps in the wheel speeds anymore, and the difference between the versions show only in the divergence of wheel speeds and vehicle speed.

 

[Griffith500]

So the 5% difference under throttle applies to both versions, but it "explains away" some of the differences in the data between versions. Very interesting, and good find!

I wonder if the 5% thing is related to some kind of bogus "transmission wind-up", i.e. it's part of how the rpm differences due to throttle position at a constant linear speed are achieved in the game?

The transmission always seemed to be based on resistances, rather than inertias - i.e., the inertia would be represented as additional friction in that case. When inertia is small relative to the power available, this isn't a problem; when it is a substantial part of the total "resistance", and power is low, it is: i.e. low-power, lightweight cars with high-revving engines.

 

[ALB123]

I made a small experiment to check the different tire compounds. I did an arcade single race on Motegi Oval with the Scion FRS courtesy car, AT, no steering, full throttle with DS3 X button until hitting a wall. All tests were on GT6 version 1.17 . You find the different MoTeC logs in the attached zip.

I then analyzed the tire response by calculating slip ratio from the wheel speeds and the vehicle speed:

SlipRatio = (Wheel Speed - Vehicle Speed) / abs(Vehicle Speed)

You can read more about it here. The essential fact is that Slip Ratio is positive when car is accelerating forward and negative when braking/accelerating backward.

Here is the outcome from start of log until just before hitting the wall.

View attachment 336341

Red = CH
Green = CM
Light Blue = CS
Blue = SH
Violet = SM
Pink = SS
Yellow = RH
Orange = RM
Green = RS

You will notice that the slip ratios for tires above CS are below zero, indicating deceleration of the car, yet we have strictly positive longitudinal G Force. So something is fishy about the wheel speeds. For all tires above CS the rear driven wheels are slower than the vehicle speed, yet the undriven front wheels (not shown here) are at vehicle speed, with a slip ratio of zero indicating free rolling.

The engine RPM seems to indicate that the rear wheels are directly coupled to the engine, as they have the same RPM at the same rear wheel speed for all tires. Yet the vehicle is travelling at different speeds for same RPM, generally faster for higher grade tires.

One explanation could be that the tires have different diameters. The wheel speed can be calculated from the angular wheel velocity times wheel diameter. This would mean that (with comparable slip ratios) that RH tires are 2-3% smaller in diameter than CH tires. But the front wheels don't show this difference, so they seem to be equal in diameter.

I plan to investigate further, with more powerful engines and different drivetrains. If you have done similar experiments or can explain the phenomenon, please post, but keep on-topic.

EDIT: here are the numeric values from a slice in the middle from CH at the left to RS at the right, you see that the front wheel speeds are almost the same as vehicle speed, but the rear wheel speeds are less than vehicle speed for all tires above CS:

View attachment 336362

Fantastic work! I really wish we had more posts like these on GTP. I find this type of discussion interesting.

 

[tarnheld]

I checked the Data Logger wheel speeds in my superlap with the 330 P4 at Suzuka, and these are the same as the corrected wheel speeds discussed above. So with this and the 1.00 vs 1.16 data i would say that the MoTeC wheel speeds have to be corrected for the driven wheels like this:

Corrected driven wheel speed = WheelSpeed * (1 + ThrottlePos * K)

with K=0.05 (anything else breaks the 1.00 vs 1.16 data discrepancy removal).

Thanks @Griffith500 for the simplification. 👍

I think this applied to FWD too, if anybody wants to check 4WD, go ahead.

I really don't know how the formula can be explained other than PD made a mistake in the MoTeC export. Maybe the exported 'wheel speeds' are really something else in the drive train -- engine torque is a function of throttle, wheel angular acceleration is a function of engine torque, wheel angular velocity is wheel angular acceleration integrated over time...

 

[Polsixe]

What's the best curve in the game for getting a good lateral g number? Should be flat, little banking, not too large a radius and fairly high constant speed? The ones at Willow Springs or ?

 

[Rotorist]

Good lateral g number means one that corresponds to data from car real life testing maybe? I would choose the 1st turn in Suzuka, the last in Tsukuba or the 1st in Laguna Seca to try out which is the most useful one for this test.

 

[Le_Manuz]

all that study is interesting,
especially that corrected spped factor, is that thing which corrected the tire pitstop bug ?
not the same value before/after pitstop (before being patched) and induced a strange effect about the tire speed ?

i wonder if the results are really the same with tires before and after a pitstop

it would be nice to do some test like said Griffith500
take values when the car is speeding and set it to neutral
should be the same, i really don't think that pd is going so far as taking transmission friction in account

 

[tarnheld]

all that study is interesting,
especially that corrected spped factor, is that thing which corrected the tire pitstop bug ?
not the same value before/after pitstop (before being patched) and induced a strange effect about the tire speed ?

i wonder if the results are really the same with tires before and after a pitstop

it would be nice to do some test like said Griffith500
take values when the car is speeding and set it to neutral
should be the same, i really don't think that pd is going so far as taking transmission friction in account

The tests so far were done in arcarde mode, so without tire wear and fuel consumption. And the correction is only needed for the exported MoTeC logs -- it seems to be only a log artifact, and nothing about physics effects in game. That said, maybe the exported wheel speeds are taken from internals affected by some bug fix -- nobody knows except PD.

Will try the driven wheel slip ratio in neutral later , should be zero.

I plan to do more analysis of the tire compounds, next up will be slip ratio under braking. Sadly the lateral response of the tires is not easy to analyse, as you need some measure of sideslip, for example the angle between where the tire is rolling and where the vehicle is heading. A comparison of the maximum lateral Gs may be all you can do, and it depends on the vehicle and not only on the tires.

 

[tarnheld]

As promised some breaking data, tested with the Corvette Stingray (C7) '14 Courtesy car on Apricot Hill. I just hit the brakes full with controller button, AT, no ABS, grip real, all other aids off. The brakes in this car are strong enough to lock the tires with all compounds. Attached you find all the logs used in making these figures.

First all compounds on dry road:

CH: baseline
CM: green
CS: light blue
SH: blue
SM: violet
SS: pink
RH: yellow
RM: orange
RS: green

You see that the vehicle stops after 88m on CH and after 54m on RS. You see the wheels can only produce a maximum braking acceleration when taken outside their grip boundaries. They are separated by about 0.1 G for the different compounds.

The slip ratios over long. acceleration for a driven wheel (rear left wheel here):

Slip Ratios for the front right wheel:

The tires can produce more acceleration when there is just the right braking force applied, for details see here. For example the SS tires had -2.1 G acceleration at a slip ratio of 43%, more than the RH tires. But the slip ratio is already too high for stable braking, and the wheels will eventually lock up when trying to hold that braking pressure. The Racing tires can hold that braking forces for much longer, they stay over -25% slip ratio, whereas the other compounds drop below that and lock up quickly. For (a lot) more details about stable braking see here -- this one is really in depth about all variables involved in braking.

The racing front tires lock up when the speed has dropped under a certain threshold. Could be that the racing tires have gotten too hot to handle the braking pressure.

The rear tires have not that much load on them because of weight transfer, so they lock up quickly on all compounds. The car remains straight except for the racing tires, where the faster lockup of the rear tires let the car turn a little.

Now the same test for a 20% wet road. Do you think the car takes longer to stop there?

Ha, the car stops faster for all compounds! Ok, it stops almost at the same length for the racing compounds, but almost 10m less than on a dry road for CH! You see why in the Slip Ratio diagrams:

Rear left wheel:

Front left wheel:

Now all compounds stay above -25% slip ratio most of the time, and generate more acceleration. It seems the tires are not that prone to lockup at high speeds when the road is slightly wet, this seems to be realistic according to the paper linked above. But the higher G values especially at the start of braking seem a little unrealistic, as the friction between the road and the tires should be lower than on dry tarmac -- and with it the acceleration generated.

Sorry for the long post, but who thought that full braking brings so many insights and little tidbits?

 

[ALB123]

@tarnheld Please, do not apologize for the length of your post when you're providing such interesting information for the masses to read and absorb. I find this fascinating. I'm quite shocked and dismayed at the difference between the three comfort tires. Personally, I like to drive on CM or CS tires, usually, and while I always knew there were differences, obviously, I would never have guessed the stopping force of the tires was that different. Wow...

 

[Greycap]

Then the test would indicate that higher grade tires expand compress significantly when driven/accelerated. Still the whole vehicle is faster when using higher grade tires.

Perhaps the "wheel speed" is measured at the hub? That would explain it.

If a stationary tyre has a circumference of exactly two metres - 200 cm - and under acceleration it expands, as tyres tend to do, its circumference can quite easily increase by several centimetres. Now if the figure is, say, 205 cm the entire wheel is in fact turning around 2,5% slower than the stationary one and measured at the hub it indeed gives a "slower" speed despite travelling at the exact same speed. It would also explain the increase in speed if higher grade tyres expand more.

 

[Polsixe]

As promised some breaking data, tested with the Corvette Stingray (C7) '14 Courtesy car on Apricot Hill. I just hit the brakes full with controller button, AT, no ABS, grip real, all other aids off. The brakes in this car are strong enough to lock the tires with all compounds. Attached you find all the logs used in making these figures.

First all compounds on dry road:

View attachment 341207

CH: baseline
CM: green
CS: light blue
SH: blue
SM: violet
SS: pink
RH: yellow
RM: orange
RS: green

You see that the vehicle stops after 88m on CH and after 54m on RS. You see the wheels can only produce a maximum braking acceleration when taken outside their grip boundaries. They are separated by about 0.1 G for the different compounds.

The slip ratios over long. acceleration for a driven wheel (rear left wheel here):

View attachment 341220

Slip Ratios for the front right wheel:

View attachment 341232

The tires can produce more acceleration when there is just the right braking force applied, for details see here. For example the SS tires had -2.1 G acceleration at a slip ratio of 43%, more than the RH tires. But the slip ratio is already too high for stable braking, and the wheels will eventually lock up when trying to hold that braking pressure. The Racing tires can hold that braking forces for much longer, they stay over -25% slip ratio, whereas the other compounds drop below that and lock up quickly. For (a lot) more details about stable braking see here -- this one is really in depth about all variables involved in braking.

The racing front tires lock up when the speed has dropped under a certain threshold. Could be that the racing tires have gotten too hot to handle the braking pressure.

The rear tires have not that much load on them because of weight transfer, so they lock up quickly on all compounds. The car remains straight except for the racing tires, where the faster lockup of the rear tires let the car turn a little.

Now the same test for a 20% wet road. Do you think the car takes longer to stop there?

View attachment 341281

Ha, the car stops faster for all compounds! Ok, it stops almost at the same length for the racing compounds, but almost 10m less than on a dry road for CH! You see why in the Slip Ratio diagrams:

Rear left wheel:

View attachment 341282

Front left wheel:
View attachment 341283

Now all compounds stay above -25% slip ratio most of the time, and generate more acceleration. It seems the tires are not that prone to lockup at high speeds when the road is slightly wet, this seems to be realistic according to the paper linked above. But the higher G values especially at the start of braking seem a little unrealistic, as the friction between the road and the tires should be lower than on dry tarmac -- and with it the acceleration generated.

Sorry for the long post, but who thought that full braking brings so many insights and little tidbits?

The fairly constant equi distance difference between compounds show PD programmers are using same formulae with slightly varied constants.