FFB plugin based on realistic Mz curve

[LBodnar]

We have been discussing some aligning torque simulation issues at the end of this thread
http://forum.racesimcentral.com/showthread.php?t=287267 but since we started to deviate from the original thread intent, I have opened a new thread.

As a result of the discussion, here is a very quick attempt at rewriting the steering FFB effects from scratch based on more or less accurate Pacejka curve fit for Mz (self-aligning torque).

Parameters are set for default BMW F1

To install, unzip BOTH files into Plugins folder and take BMW F1 fro a spin
To uninstall, just delete LeoFFB.dll and LeoFFB.ini

I tried it on Caterham with the following settings:

25.0 Front Wheel lock angle, degrees (max deflection to one side from straight on)
0.635 Front wheel distance to centerline (half the front wheel track) in metres, from .pm file
1.125 Front axle distance to body reference point (about half the wheel base) in metres, from .pm file
10.5 Relative strength of FFB effects
6.0 Aligning moment curve shape / feeling of tyre losing grip. Lower numbers are duller, range 1.0 .. 10
0.99 Negative aligning moment parameter for high slip angles, range 0.0 ... 1.0
10.0 Slip angle where aligning moment reverses direction - typically between 8 and 15 degrees
0.01 Caster (dimetionless)
0.0 Camber ratio (not degrees!)
1.0 Road bumps feeling on the wheel
11500.0 maximum FFB force for stationary and low speed effects
10.0 stiffness of stationary wheels
1.5 wheel rubber hardness (1..5)
8.0 how quickly static forces drop off when rolling


Cheers!

[Niels_at_home]

Hey Leo, good work, wish I could drop by for a cuppa and a look at your wheel..

"my" corvette has, imho, far better tyres and seems to handle more like a car, if you want the latest beta of it to test a 'realish' roadcar with 'realish' road tyres, I'd be happy to send you the car..

[LBodnar]

Hi Niels,
Yes, please!
I can find some real Mz curve data for road cars but at quite a loss how would an F1 tyre behave at high slip angles with regard to steering torque. This would define the handling behaviour in a fast spin.

[TechAde]

Hi Leo,

My initial impressions are that it feels good, especially with the BMW F1. I tried a few other cars and the effects felt exaggerated, but I guess this would just need some tweaking of the parameters? (All I changed was steering lock to 25 to match the other cars I drove)

One thing I noticed is that it doesn't seem to respond to suspension geometry changes (such as changing caster from +5 to -3), or if it does the effect is very slight. Would this be correct?

Edit: Btw my default lock on the BMW F1 07 seems to be 15 degrees rather than 14.5 - do we have different versions?

[Niels_at_home]

Would that mean the FF just follows the Mz 'curve'? Suspension geometry can have a big effect because you lift the car a bit when steering. The tyres indeed (assuming pacejka etc have it right, very very very likely) have this Mz alligning curve, but thats only half the FF of course, though I feel Leo is quite aware of this..

[TechAde]

Thanks Niels, that would make sense

[LBodnar]

Originally Posted by TechAde

Hi Leo,

My initial impressions are that it feels good, especially with the BMW F1. I tried a few other cars and the effects felt exaggerated, but I guess this would just need some tweaking of the parameters? (All I changed was steering lock to 25 to match the other cars I drove)

One thing I noticed is that it doesn't seem to respond to suspension geometry changes (such as changing caster from +5 to -3), or if it does the effect is very slight. Would this be correct?

Edit: Btw my default lock on the BMW F1 07 seems to be 15 degrees rather than 14.5 - do we have different versions?

Hi there!
Yes, correct, the suspension geometry is not there, particularly the caster as it should be "inbuilt" in Mz curve via offsets and shape alteration. Also, the Mz is assumed to be translated into steering via linear relationship while it should also take into account the fact that a tyre centre is offset from the kingpin axis.

I have possibly got the F1 lock settings wrong as I have derived it from .hdv file where it says
SteerLockRange=(5.0, 0.5, 51)
SteerLockSetting=20
I must have assumed that the result is 5+0.5*(20-1). Possibly "-1" is not needed...

[LBodnar]

Originally Posted by Niels_at_home

Would that mean the FF just follows the Mz 'curve'? Suspension geometry can have a big effect because you lift the car a bit when steering. The tyres indeed (assuming pacejka etc have it right, very very very likely) have this Mz alligning curve, but thats only half the FF of course, though I feel Leo is quite aware of this..

Of course
I am still struggling to understand how rFactor would calculate Mz as
1) it's not tabulated like slip curves
2) I see no reliable way to derive Mz from slip curves. Mz is an integral force and cannot be reconstructed from another integral variable (Fy or Fx)

So I wanted to see what happens if some sensible curve is used... If we go all the way along this path then the plugin might probably turn into an independent tyre simulator...

However, this only affects handling, i.e. steering feedback - it does not affect car motion physcis (apart from FFB feeding back into the control) so a mismatch between the behaviour and a feeling would be very annoying.

However, I have found that it gives a better feel during the slide and cornering. I use ALL the driving aid off (TC, ABS, steering help, etc)

[TechAde]

Originally Posted by LBodnar

However, I have found that it gives a better feel during the slide and cornering. I use ALL the driving aid off (TC, ABS, steering help, etc)

Agreed, however I also get this better feel using RealFeel. The only real difference in feel (see what I did there?) during sliding between this and RealFeel (with good pneumatic trail values) is that with RealFeel Mz never drops below zero - however it will reach almost 0 if the entire contact patch is sliding. Whilst racing I think this difference has a negligible effect as you shouldn't really be reaching such high slip angles.

If there was a way you could have your plugin react to geometry changes as RealFeel does (not just caster, scrub radius also plays a huge part in braking feedback) then I think you'd be onto a winner. Possible?

[Niels_at_home]

If I understand it correctly, the Mz curve as you see on pacejka plots etc is purely that of the tyre. I expect it to depend heavily on the load on the tyre; so one would probably have to scale the Mz curve with load for starters.. Then the effect of suspension geometry on experienced steering torque will depend heavily on the load as well..

How many of these things does LeoFF do at the moment?

[LBodnar]

If we knew how rFactor calculates what it does calculate then it won't be such a hard task to correct what it doesn't do right.

I am tempted to read back the .hdv and .pm files and account for jacking torque (as Niels suggested), castor and scub radius but it gets too complicated too quickly.

Again, when designing the model one should trust their input variables.
All the parameters can be modelled to a higher degree of accuracy and dependant on normal force, speed, temperature, etc.
However, I am interested to which degree the normal tyre forces are accurately modelled by rFactor. If they do not accurately reflect the transient effects in suspension then most of the efforts with tyre modelling will be wasted.

TechAde, can you suggest a RealFeel settings for BMW Sauber F1 + Essington GP Circuit combination to get the best result from RealFeel that I can use as a benchmark?

[LBodnar]

Originally Posted by Niels_at_home

If I understand it correctly, the Mz curve as you see on pacejka plots etc is purely that of the tyre. I expect it to depend heavily on the load on the tyre; so one would probably have to scale the Mz curve with load for starters.. Then the effect of suspension geometry on experienced steering torque will depend heavily on the load as well..

How many of these things does LeoFF do at the moment?

Sure, Mz is linearly scaled with normal force.
AligningTorque = NormalLoad * Koef(SlipAngle)
There is some subtle higher order dependence on normal load but it can be ignored for now.

The actual aligning torque for each wheel is proprtional to normal load on this wheel and its own slip angle which is different for both wheels when you are spinning or turning. This is where overall wheelbase geometry that I have mentioned before comes into effect.

[TechAde]

Originally Posted by LBodnar

TechAde, can you suggest a RealFeel settings for BMW Sauber F1 + Essington GP Circuit combination to get the best result from RealFeel that I can use as a benchmark?

Will do, might take a me a little while though sorry, rather busy at the mo.

How about using the Caterham Caterbusa or Powertec for comparison rather than the F1? The default RealFeel parameters should be spot on for those (1500N MaxForce), though I also run 0 smoothing and 11500 damper.

Edit: Or maybe the latest C6, though the large amount of caster my make direct comparison tricky

[kangaloosh]

Originally Posted by LBodnar

TechAde, can you suggest a RealFeel settings for BMW Sauber F1 + Essington GP Circuit combination to get the best result from RealFeel that I can use as a benchmark?

LBodnar - RealFeel won't work very well with that mod - the suspension is not how you would design it if you were designing suspension with RealFeel in mind. This highlights the biggest problem with RealFeel - you need a mod designed in a similar way to a r/l car, to generate the forces into the steering that you want to feel, you can't just rely on rFactor to simulate the forces it thinks you should feel.

Chris

[LBodnar]

Originally Posted by kangaloosh

LBodnar - RealFeel won't work very well with that mod - the suspension is not how you would design it if you were designing suspension with RealFeel in mind. This highlights the biggest problem with RealFeel - you need a mod designed in a similar way to a r/l car, to generate the forces into the steering that you want to feel, you can't just rely on rFactor to simulate the forces it thinks you should feel.

Chris

Is this purely due to the steering knuckle/spindle geometry? From BMWF106_susp.pm front suspension detail the wheel rotation axis and the kingpin rotation lie in the same plane. This basically means that tyre side force has almost zero effect on steering arm force because its angular momentum is close to zero due to zero arm. Free rolling front tyre will not generate any lateral force either (ignoring rolling resistance) so what is left is pure Mz that is of questionable quality anyway...

[Texas-Tornado]

Hello Grandmasters of FFB!

I just tested the LeoFFB. Once with uninstalling the RealFeel first and once with RealFeel and LeoFFB.

With RealFeel and LeoFFB all the Mods with combined FFB Values (GP79 with 75% RealFeel f. Ex.) are much better to drive.

When i set RealFeel to 0% i just got the LeoFFB (which sounds logical to me).

But when i drive with 100% Real Feel (V8Factor 100% RealFeel f. Ex.) i have the strange Feeling that it feels different from the Version without LeoFFB. It feels a bit better.

This is all from the Feeling point of view, no technical Knowledge here. So can you tell me if my theorie is true or false. How does rFactor handle both Plugins at the same time???

Thx for answering my question!

Cheers David

[LBodnar]

As far as I understand multiple plugins are loaded sequentially and just pass each other the value for FFB:
Pluging 1:
receive FFB from rFactor (A)
process/alter FFB if needed
submit new FFB value to rFactor (B)

Pluging 2:
receive FFB from rFactor (B)
process/alter FFB if needed
submit new FFB value to rFactor (C)

My plugin calculates the FFB from scratch, i.e. it does not use incoming value so if it is loaded last it will discard the previous values. It looks like RealFeel uses a weighted average of the calculated and incoming FFB values so if it is loaded second it should combine two together according to its % setting.

I don't know how one can tell or control the order...

[TechAde]

Originally Posted by LBodnar

It looks like RealFeel uses a weighted average of the calculated and incoming FFB values so if it is loaded second it should combine two together according to its % setting.

Yup, RealFeel ignores the incoming value unless RealFeelMixerPercent is less than 100% (the default). The new SmoothingLevel parameter in v0.9 dictates how many previous values are used to calculate the weighted average, between 0 and 9.

I had an issue reported the other day where RealFeel would sporadically stop producing any force, it turned out that they had 2 versions of the dll in the plugins folder. Quite what happens in that situation I really don't know.

[LBodnar]

TechAde, may I ask you a question?..

I am trying to model aligning force dependence on braking slip and geometry. How one can deduce the braking slip from telemetry data? .mGripFract does not make much sense to me...

Following the definition I'd need a distance from wheel's centre to the contact patch to derive the slip but I can't figure out how to calculate it. The only available parameter is the wheel's .mRideHeight but I can't quite understand what it actually represents.

It seems that there is no reliable way to extract a braking (longitudinal) force for each wheel - neither directly nor by calculating the braking slip and referencing the slip curve...

[TechAde]

Originally Posted by LBodnar

TechAde, may I ask you a question?..

I am trying to model aligning force dependence on braking slip and geometry. How one can deduce the braking slip from telemetry data? .mGripFract does not make much sense to me...

Following the definition I'd need a distance from wheel's centre to the contact patch to derive the slip but I can't figure out how to calculate it. The only available parameter is the wheel's .mRideHeight but I can't quite understand what it actually represents.

It seems that there is no reliable way to extract a braking (longitudinal) force for each wheel - neither directly nor by calculating the braking slip and referencing the slip curve...

Unfortunately I don't think the required info is available via the plugin interface.

[Niels_at_home]

Originally Posted by LBodnar

TechAde, may I ask you a question?..

I am trying to model aligning force dependence on braking slip and geometry. How one can deduce the braking slip from telemetry data? .mGripFract does not make much sense to me...

Perhaps you can do something like wheelspeed / car speed? Although it may not be perfect as turned wheels

Following the definition I'd need a distance from wheel's centre to the contact patch to derive the slip but I can't figure out how to calculate it. The only available parameter is the wheel's .mRideHeight but I can't quite understand what it actually represents.

Is that the slip definition? I thought it was the difference in 'speed', i.e. the speed the tyre would be rolling at (circumference * revolutions/sec) and the speed it is actually going at?

[LBodnar]

The side force vs. braking force graph is very difficult to model...

The definition includes the ratio of circumferential speed of freely rotating wheel in the same configuration and the actual braking wheel. Knowing the hub-to-road distance is an intermediate parameter that can help estimating free wheel RPM.

I have more or less accurate estimate now based on tyre stiffness including internal pressure and its radius and the correction for the RPM expansion. Will try it tonight and, maybe update the plugin.

[LBodnar]

I have updated some simulation parameters and added low speed handling.
Cheers

[irousso]

Leo - just tried your plugin and I really like it.
I have couple of questions though:
1. do I need the low speed plugin to work with this one?
2. when the car loose grip, the FFB drops - what parameter should I change in order to reduce the FFB drop?

Thanks!
Itay

[LBodnar]

Originally Posted by irousso

1. do I need the low speed plugin to work with this one?
2. when the car loose grip, the FFB drops - what parameter should I change in order to reduce the FFB drop?

Hi!
1. No, I have put the low speed code in already
2. Depending on what you want to acieve

a) you can add a bit of caster (understeer):
0.01 Caster (dimetionless)

b) you can extend the aligning moment loss curve to a higher angles, e.g.
12.5 Slip angle where aligning moment reverses direction - typically between 8 and 15 degrees

c) you can change the shape of it, make it more "mushy", e.g.
4.0 Aligning moment curve shape / feeling of tyre losing grip. Lower numbers are duller, range 1.0 .. 10