# Comments on STS with some data analysis

Discussion in 'General Discussion' started by JmbJr, Jan 28, 2016.

1. ### JmbJrRegistered

Joined:
Jan 26, 2016
Messages:
10
0
Hello all,

I would like to share a few things I have done in the past few days regarding the Steering Torque Sensitivity implemented in rFactor 2.

After much read and lots of data gathering and analysis, I am a point that I can share what I have done without not being completely wrong or misleading.
So I played around with the STS, using the Skip Barber at Joesville and used TechAde Plugin to gather the SteeringTorque and FFB outputs from the car (BTW, I have a T300RS). For my tests I used STS=0, 0.5, 1 and 2. Also, set both min Force and smoothing to 0.

These next two figures are the comparison from STS = 2 and 1 (first figure) and STS = 1 and 0 (second figure)

View attachment 19201
View attachment 19202

(I've also included a high-res pdf of the figures)
View attachment 19203
View attachment 19204

You can clearly see the influence of STS on the FFB that is sent to the wheel.

My goal (which I still haven't figured out) is to see if I can "linearize" the output of my T300 using some STS value.
So far, I managed to, I think, figure out the mathematical model ISI in applying to the FFB (the description TechAde conjectured seems to be right).
The normalized SteeringTorque vs FFB is the following:
FFB = [1-STS]*x[SUP]2[/SUP] + STS*x, where x is the SteeringTorque. This might not be any news to most of you, but I thought it would be a nice addition to the discussion.

I have included a link to the spreadsheet I used to perform all the analysis. In addition, if someone is interested in seeing the effect of STS has in your particular wheel response, you can use wheelcheck software to do so, however it is important that you change the Max Count to 100, otherwise you will have problems with the calculations.

https://www.dropbox.com/s/c57lknivr2x17cx/STS Analysis - T300 - All Forces 100%_V2.xlsx?dl=0

In a nutshell, this spreadsheet takes the wheel output response from the aforementioned software and applies the STS and the FFB Minimum Torque (FFBMinT) to verify how these to parameters affect our wheel response. (This analysis is inspired by a similar one found in the PCars forum. Just as a disclaimer here, it is beyond the intention of this thread to compare whatever PCars is doing with their FFB systems and what rFactor offers).
One thing you will notice is that both STS and FFBMinT are not enough to "linearized" the wheel response, at least, for my T300RS.

Again, the idea of this thread is to provide some numerical/analytical view on how the rFactor FFB parameters play a role in the final-to-the-wheel output. I know that FFB is very personal and it is beyond the point here.

Hopefully, you will find this useful and bring further discussion.

Cheers

Last edited by a moderator: Feb 2, 2016
2. ### LazzaRegistered

Joined:
Oct 5, 2010
Messages:
8,570
3,852
It's a shame ISI haven't (as far as I know) given information on whether the FFB setting in the plugin interface will ever be functional (like rF1), because if it were you could basically give it any response you wanted. But it's probably fair to guess they have some decent reason for disabling that. (would start many arguments at least, I'm sure)

3. ### ProdigyRegistered

Joined:
Nov 2, 2012
Messages:
947
64
What kind of plugin would that be, something like RealFeel?

4. ### LazzaRegistered

Joined:
Oct 5, 2010
Messages:
8,570
3,852
If you were able to 'write' to the FFB you could come up with an alternative FFB system, or you could just take the incoming values and shape them whatever way you wanted. Like the equalizer analogy above. Potentially you could have some standard correction curves for specific wheels and stuff... but we can't, so probably not worth worrying about.

5. ### unknwnRegistered

Joined:
Aug 11, 2015
Messages:
132
10
Pcars linearity adjustment is quite good, several segments isn't really necessary unless the wheel has many aggressive spikes in the output (T300 and T500 are relatively smooth).

Please see example, right top graph (T300 75% in TM):

Red line is final output after processing compared to orange line which is normal wheel output without any tweaks.

Above approach allows to "improve" deadzone/low forces in the center (with help of DRR/DRF settings), where as scoop knee/reduction allows to flatten the overall curve.

Last edited by a moderator: Jan 28, 2016
6. ### EmeryRegistered

Joined:
Oct 24, 2010
Messages:
2,785
1,377
Rather than trying to re-linearize a nonlinear wheel, the way I'm using STS is sort of like following the concept of HDR.

If we've got a wheel that doesn't produce the peak forces of the real world, then there are two choices for configuration. First is the obvious "scale everything linearly", but that produces weak forces throughout the range because the wheel can't match the real world. The second is to take advantage of the fact that human sense of pressure isn't linear and trick it, just like HDR tricks our eyes.

So after setting car-specific FFB so it almost never clips, start with the linear response we're all taught to configure, then bring the bring up the low level with a touch of minimum torque (1%-2%) and maybe a little dampening & spring (a *little* means likely no more than 10%) to give a nice feel without masking the subtle difference between parked and slowly rolling (or locked brakes and unlocked brakes). The idea is to make it work for YOU, so I'm not going to insist that the levels I find work for me will work for you! That leaves you with more of a U-shape to the force curve and the peaks aren't as high relative to the lows. So now we increase STS to fatten up the middle & high forces... peak forces are still within the software clipping per the car-specific FFB. For myself, 1.2 STS works well on a CSWv2 with BMW GT2 rim.

7. ### JmbJrRegistered

Joined:
Jan 26, 2016
Messages:
10
0
The problem for me with PCars or any other system to "linearize" the output response of the wheel is data the methods so far (including whatever you find in PCars forum and what I provided here - basically the same idea) are just manipulating the "input force" to the wheel to change the shape of the output response. I think that is not the ideal, in fact, far from it. The best way, IMO, is to have a wheelcheck-like-software where the user could change the input force shape prior to performing the test. That would be the only way for me that I could trust what these functions are actually doing. Now that I think of it, I may try to create that. I have a vast experience in MATLAB, so if there is a wrapper for DirectInput to MATLAB then this will be very easy. If not, I will give a shot and try to code in C#. Nonetheless, these spreadsheets at least can give a first order idea about what these linearization methods are doing.

In terms of feeling, I haven't done a through test of the STS to see which one I prefer best. So far, I am quite happy with the rF2 default settings, and, btw, I using all forces in the TM control panel @ 100% with the overall output @ 75%. I can feel what the car is doing throughly T300 as long as the cornering forces stay within 75% of the max, which is the range that the T300 is "mostly linear".

8. ### unknwnRegistered

Joined:
Aug 11, 2015
Messages:
132
10
Another way could be just go after something that gives more consistent and informative feedback across the range. In my case (T300) that is something between 0.7 to 0.8 STS when running 75% force in TM CP. Of course I would prefer Pcars like linearity adjustment as applying STS on already curved output is a guesswork and cant give a precise outcome.

Last edited by a moderator: Jan 30, 2016
9. ### unknwnRegistered

Joined:
Aug 11, 2015
Messages:
132
10
I have seen many times STS ~0.270 being recommended for being more closed to RL feedback. Personally I can't stand this setting on T300 as it "kills" almost all forces at lower end of dynamic range (0-40% force - i.e. driving straight, braking, weaker turns gives low/no details/force but as the force increases it suddenly get's boosted so much that it feels like a waste in the range of 70-100% force considering the loss in other ranges and it feels like there is no smooth force transition according to what the car is doing - nothing, nothing, then suddenly BAM you have all the torque. However I understand that on DD wheels it will feel different as you have enough torque which stills allows to feel details even after reducing lower end.
What I mean by consistent feedback is linear feedback. I.e. You go through corner increasing/decreasing understeer and feel force increasing/decreasing smoothly regardless the force being outputted (i.e. 30% feels like 30%, 60% feels like 60%, 90% like 90% instead of 30% feeling like 60%, 60% feeling like 90% and 90% feeling like 99% or vice versa when setting very low STS i.e. ~0.270).

Last edited by a moderator: Jan 30, 2016
10. ### unknwnRegistered

Joined:
Aug 11, 2015
Messages:
132
10
I am talking about the "linearity" of what physics engine outputs vs final torque felt on wheel. If the rFactor physics outputs 100% force at most optimum slip angle then I want to have 100% torque, if the physics engine outputs 70% force signal I want to feel 70% torque of the wheel. The inbuilt "compression" or STS will mess up the input vs output relation. Depending on where the force is acting in the range of wheel output (0-100% force) - the optimum slip angle might give 100% output as 100% torque and 70force as 70torque (linear), however with "inbuilt" compression or high STS you might feel 100% output as 100% torque and 70% output as 90% torque (high compresion) - meaning you get only 10% torque difference in terms of optimum slip angle vs understeer, where as you should have 30% torque difference.. Where as lowering STS will boost the dynamic range of the range "50-100%" while weakening the "0-50%" range.

Are you saying that rFactor physics outputs force values that gives "unrealistic" feedback?

Last edited by a moderator: Jan 30, 2016
11. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
My guess would be that rFactor physics outputs right force values.

Other things changing how wheel behaves are:
Weight of your steering wheel, delay between game and ffb system, physical friction of ffb system and virtual damping/friction effects. Also power steering being not simulated can have somekind of effect.

maybe using low STS simulates effect of higher inertia of real cars steering system.
With lower force wheels you probably don't need to worry about that because ffb related to steering wheel's mass is not too high.

With my osw wheel i found out, that high (default) STS + high delay setting(lot of filter) didn't work too well. Wheel oscillated with some cars. After going to lower delay settings (no filter) I was able to use defalt STS again. No oscillating or other problems and drifting is easy. Maybe too easy.

Last edited by a moderator: Jan 30, 2016
12. ### JmbJrRegistered

Joined:
Jan 26, 2016
Messages:
10
0
Thanks you all for the contribution in this thread! I do appreciate your time and input to further the discussion!

I have also seem some users saying the were using STS @~ 0.2, and similar to you, unknown, the lower-to-mid end of the force range was way to week, and I didn't like the feeling at all.

Like Paul mentioned, the tire forces when cornering are not linear, in fact, far from it. The lateral force times the mechanical trail + the Aligning Torque VS Slip Angle might produce a peak torque just before the maximum grip and will dip down a little when maximum grip is achieved, depending on the car and tires. In the end, the "Input Force" to the wheel will produce this small dip in torque in maximum grip situations. Depending on the car, again, this small decrease in torque might be in the order of 10% (let's just use this number as an example). The problem with the non-linear response of our commercial wheels, like the T300, coupled with the rF2 in-car multiplier is that if you set the multiplier too high and the maximum experience rF force (say, in TechAde's plugin monitor) reaches 85%-90% of the maximum, we will not have a 10% drop in torque in our wheel (as per out example), due to the high insensitivity of the wheel in the higher end. For my T300, a 10% decrease in "input-to-the-wheel" torque will give me a 2% reduction in the torque, with all the STS and MinForce as default. So the idea is to keep the dynamic range of the produced torque in the region of most "linear" output response. In addition, keep in mind that most wheels, even with a "linear" region exists, the slope response may not be 1, which might deceit our perception of the produced torque.

I will play more with the STS coupled if some settings for the MinForce to see if I can achieve a more "linear response" throughout a higher range of torque, and I will report back.

PS: How do I embed images in the post. I tried to follow the forums instructions but could not do it...

Last edited by a moderator: Feb 2, 2016
13. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
Agree.

Yes, it's the same thing.

14. ### unknwnRegistered

Joined:
Aug 11, 2015
Messages:
132
10
I do not see how the natural inertia of different wheels or max RPM at certain % has impact in the situation I am trying to describe i.e. you go through the fast and smooth corner, keep the wheel in the same position (torque acting, no RPM or oscillations from bumps and such), when the car starts to understeer STS/inbuilt compression will mess up the ratio of torque felt of optimum angle vs understeer i.e. you might feel it harder or easier depending on your settings/natural linearity of hardware and the % of force being outputted from the sim. In this situation inertia/wheel weight/max RPM will have nearly no impact for the main static force that you feel i.e. torque and we have STC setting to make the powerful wheels have the same torque output at certain % (assuming the natural output of the hardware is linear). Inertia/max RPM and other hardware differences will start giving very different feedback when there are spikes of force being outputted from the sim (bumps, kerbs) or you actually let go of the wheel or when you allow it to turn e.g. drifting - where it might become very hard if the wheel is constantly overshooting or is too slow/dampened and etc. As I understand these are the situations that you are trying to fix with STS, which I see as workaround that will change the feedback in other situations. The better way would be to have the correct and adjustable (to suit different hardware) dampening in the system.

Last edited by a moderator: Jan 31, 2016

Joined:
May 22, 2014
Messages:
693
199
True

16. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
Shouldn't you be increasing the STS then?

17. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
I have tried changing it, and i think the effect is the same as the pictures shows in op.
Btw. the pictures are in wrong order in the op. First picture shows STS = 2 and 1 instead of STS = 0 and 1.

So I still think that increasing the STS would be the thing to do, if i understood the problem right.

18. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
Labels inside pictures and in pdf links was right but the pictures was in different order than what was said in the post:"These next two figures are the comparison from STS = 0 and 1 (first figure) and STS = 1 and 2 (second figure)"

And your description of STS 0 and STS 2 is the same as i have experienced it.

19. ### Korva7Registered

Joined:
May 22, 2014
Messages:
693
199
Back to this.
So i thought that you should increase STS from default, but that was because i was concentrating on how torgue rises as slip angle rises. This would make torgue rise earlier, but on the other hand the drop off would happen later. If the torgue drop off when slip angle gets too high is more importat then lowering STS helps.

Joined:
Jan 26, 2016
Messages:
10