AC FFB Implementation, Improvement Suggestions

Heh, I really like how Kunos takes critique. Some other baldhead would've already banned Pax from playing.

Sorry for OT. I'll stop now.

 

Does this mean that in a wheel like the T500RS, for example, only the Constant and Damper settings in the control panel actually make any difference? And that periodic and spring make no difference?

For reference, here are what the T500RS settings mean:

The four different settings located in the Gain Settings tab have the following purposes:

CONSTANT:
A constant force will keep the same level in time. When a game decides to apply a force of x% of what the wheel can do, the "Constant" will keep that force at the same level.
Example: A driving simulator game will usually use the constant force to simulate the G-Force. It will create a constant force at 0% but, depending on the speed and the wheel angle, the force will be increased accordingly.

PERIODIC:
A periodic force will vary in time according to the type of periodic effect, amplitude and frequency. Because a periodic force includes an offset that has the same behavior as a constant force, some games will create a periodic force and merge an effect they would render for a constant force (like the G-Force) and a periodic effect (bumping road).
Example: A periodic effect gives the shaking effect on the wheel. At high frequency, you will feel it rumble... but in some cases very low frequency are used to create crash effects where the wheel first turns fully to the left, then to the right and then to the left again, before stopping. Of course, that depends on what the game developers decide.

SPRING:
A spring force is a force that increases according to how far you are from a specific position on the wheel. The basic spring force we could think of is a default spring center where the wheel goes back to its center position when you release it. But the force can be set to negative, which will make the wheel go further away from the center position.
Example: Basic spring force where the wheel goes back to its center position.

DAMPER:
A damper force controls how the wheel will react when it's moving. It is usually used as a (dynamic) friction or if you use it while the wheel gets back to its center position (spring effect), then it will behave like controlling the damping on a spring-mass system. A game will usually use the damper force in order to make it harder for the player to turn the wheel while in other circumstance make it feel like if it's very smooth and easy.
Example: Controlling the force you need to apply on the wheel to rotate it.

The SPRING and DAMPER effects are what we call Dynamic Effects, because they rely on information which depends on the wheel (position for spring or speed for damper). The Dynamic Effects are those effects which benefit the most from hardware force-feedback implementation (compare to software) - like in the T500RS - because you want a fast response based on how the player will turn his wheel.

The CONSTANT and PERIODIC effects are what we call Static Effects. Once the game sets them, they act according to the parameters set or modified through time by the game regardless of how the user turns the wheel. The game might adjust them accordingly, but they are not linked directly to the wheel position, speed or acceleration.

 

Yesterday I tested all the steering options to help me feeling the road better and none of them improved my feeling.
I am not good enough to follow your first page discussion, but I am used to Game Stock Car FFB and we feel clearly the road grip with it. I know that these games have a different approach, so it is not to compare. But I don't have many precision at turns exit, I miss some infos like road grip, somy trajectories are not really precise in the end of turns. As if the car is not connected to the road. For fast turns, it is different, good feeling.

 

I guess it´s related to the recent update. After a short time my wheels starts to pull to the right side forever. It´s GT3RS V2 Firmw. 756, Driver Beta 174, Win 7 x64
Don´t have time to try a little more...........

 

Hello again guys,

nice to see the responsiveness of the devs and some good input! I have just been sleeping...
I have limited time to write a reply just now, but here is a little at least:

In the AC version of yesterday which I looked at traffic from, in each loop @250 Hz you send three "structs" of data to the FFB device:

1. Update of Constant Force magnitude. This is fine of course
2. Update of Constant Force effect config, including "Direction Enable"-flag, duration, gain etc. With the FFB driver I use atm, this is a 17 byte structure. It is also never changed. Hence, sending this is unnecessary.
3. Update of Damper effect config, which is the exact same struct as in 2, and its contents is also static. Hence, sending this is also unnecessary.

Stefano, as you mentioned DIEP_DIRECTION, and "Direction Enable" is a member above, setting that and sending it to the FFB device could well be the only reason the struct is sent at all.

Again, it should be stressed that exactly what goes out to the FFB device and what can be set depends on the FFB device and drivers, and one can unfortunately also not expect each FFB device to behave exactly the same for a certain set of FFB calls. So, caution is needed when changing the FFB code not to cause some wheel model to stop working, but removing the above two structs from the data flow I think should cause no issues.

Regarding the "unorthodox" way of setting damper strength, Stefano you use the gain parameter to do this. I suggest you set gain to a constant max and use damper PositiveCoefficient and NegativeCoefficient instead (http://msdn.microsoft.com/en-us/library/windows/desktop/bb321011(v=vs.85).aspx). The result of this is better defined.

As for "overflowing" the USB channel, that is not possible, but what you _may_ end up with by sending redundant data is that you will not reach your intended update rate. Depending on the FFB device driver and device implementation, that could have serious consequences = causing the sim engine not being able to meets its deadlines.

Regarding monitoring of USB traffic, one can use either a HW or SW solution. HW solutions are quite expensive (€1000 and up for good stuff). SW solutions also cost, but are often less expensive (and in some respects less accurate). HW monitors are mostly for implementing USB stacks and USB HW, SW monitors have the advantage you also can view traffic in the USB driver and other OS layers.

I will have a look at the new AC build tonight!

BR Fredric

 

Apart from all these technical matters, there are a couple of things i do expect to feel with the ffb that actually i do not feel at all (don't know if it's my fault in some device setting):
1 - in a stopped car i do expect some force to turn the wheel. At the moment i can turn it all to the end with a finger...
2 - when i ran over a kerb (even at extremely low speed) the steer should rotate. I don't know how to explain but hitting a kerb, going into a hole or putting the car on un-leveled asphalt should let the car modifing its direction for a bit: when in reality we drive on gravel we can feel the steering moving itself left and right depending on what the tires are rolling on... or, another example, when i hit a stone or, on an extreme level, a sidewalk, the steering wheel rotates! On smoother level is what happen in the banks like in the ovals. This force i belive is greater than the force we feel when we go into under or oversteer, which is well simulated.
3 - Is it possible to implement some filter that consider the mechanical feel of each car' steering wheel? We do not rotate the tires, we rotate a steering wheel connected with some mechanics and sometimes a power steering too...

Thank for the attention and you are doing a great job, continue this way!

 

Agreed, ideally all FFB queues should be generated by the host side simulation SW, but as Sebastien mentions, the computation update rate of the simulation SW is in practise much lower than the update rate in the FFB device (at least for quality FFB devices).
A sim update rate of 250-500 Hz, plus FFB device round trip latency, I would say is too low to generate e.g. a really good damper effect.
I have not experimented with this low update rates though (no need for me to do that as I use powerful microcontrollers for my FFB devices), so I cannot say exactly how low you can go with acceptable results. David Tucker at iRacing has made some experiments on this I think though.

So, the available standard "canned" effects in FFB wheels are simple, but they could be used to achieve certain convincing effects hard to produce entirely from the host SW side.

 

Pax, after reading your answer about what you call redundancy according to MS , Logitech and Thrustmaster driver engineer, I don't see any problem in Stefano code. Thie way of sending value are perfectly genuine and IMO must be handle by the driver/firmware.
Also remember that code side we don't send USB data, we call DirectInput and it's direct Input who send USB packet we don't have control on.
For example ,according to your readings, calling 1 setParameters with flags DIEP_DIRECTION and DIEP_SEPCIFICPARAMETERS seems to become 2 USB packet send by Direct Input

 

Good of Fred to offer the results of his tests

Good of kudos to respond so quickly ( his masters voice ? Lol )

Will the AC software fab be upgraded to update higher than 250hz ? ( perhaps 400hz )

I use a high end wheel & can feel the difference in fine resolution ( in effects ) between Iracing & rf2

AC ffb is good but does feel as though it needs some work in the finer detail dept

 

Hello there,

I am aware of DirectInput, yes; as you can see I linked to it in my last post above, on how to best set damper strength.

On your and Thrustmaster's etc. theory that one cannot update Constant Force magnitude in isolation, as you can see in my post above the damper config is updated in isolation of e.g. its coefficients, so they can be split. That should (hopefully) also be evident if one inspects the DirectInput API, but I will not spend time on that.

Furthermore, I just had a look at the FFB traffic generated by iRacing, and sure enough iRacing outputs updates to the Constant Force magnitude _only_, which in my case is 4 bytes every ~17 ms -> nice and tidy.

So, it can be done.

I will now have a look at the latest AC build.

BR Fredric

 

I just checked the new build, and now only the Constant Force magnitude is updated each round; 4 bytes every 4 ms. Very nice!

 

I don't understand your speech there, I never gave a theory or say how things must be done or don't.
I'm no expert in FFB protocol in any way, I'm just telling you what those guys told me how I have to call direct input in the correct way. And just tell it there to share this knowledge to help Stefano.

 

I have no idea how FFB works in all honesty so it's hard to atriculate but something feels off in AC to me. Generally I have a lack of road feel and a sense of resistance/lightening of the wheel as grip changes. FFB actually feels more like a heavy spring is working the wheel, not the motor. Around centre it feels very vague and steering inputs dont feel very precise. The 458 and KTM feel better than others, such as the E30 which feels very numb and limp. There's also a lack of "effects" such as kerb rumble etc.

In Game Stock Car I get a much better sense of the cars balance and general grip levels through the FFB. I hope this is something that will continue to evolve and improve in Assetto Corsa.

 

Kerb "issue" has been answered but it'll be worked upon. Just need to wait.

Sent from my bq Edison using Tapatalk

 

Comparing games is perhaps something to avoid, but like endgame, my ffb reference is game stock car 2013 too. Road grip feeling and mass transfert are perfectly rendered (f1 retro for instance), it can be tested in the demo. But Assetto uses another approach so I really don't know what is possible or not...
Anyway, thank you for your efforts to improve the game.

 

I notice that many of those voicing an opinion (myself included) on the vagueness of the FFB have teh T500RS. Worth looking into, maybe? A game like rF2 the FFB is absolutely sublime and I can feel every nuance of what the car is doing, but in AC There really isn't much going on at all, except aligning forces. I see the FFB-meter in pedals app vibrate a little bit, but I just don't feel those forces in my wheel.

 

In my league ( http://sim-racing-evolution.fr/index.php ) some guys have the fanatec csw and have the same feeling as t500 owners. At this time, they have more sensations in gsc2013 too. I am confident in Kunos ;-)

 

You are guys amazing! Daily I'm operating with people who can barely write their name or send attachment, then there are people like you, understanding all of the physics, programming, studying all of the variables of the FFB process and all those inputs... You are just gods for me and my brain going to explode when I'm trying to realize how everything works. I have finished a couple of laps with a couple of cars and it's simply brilliant. I can feel I'm driving a real car finally. Iracing is perfect, RF2 as well, old LFS is still great too, but here I can feel something like a step ahead ;o) I know that even if real things are 100% implemented into the game, something here and something there must to be changed so the feeling can be identical as much as possible and you implemented everything perfectly all together. Yes, there is a room for improvements and tweaking, but the basic is absolutely awesome for me! I wish you guys all the luck and wish everything will go smoothly for you. Even in this "stage" of the early access, this becoming number one to me ;o)

 

Uh oh/splash.. the graphics is brilliant too. On my 560Ti it looks just great, running smooth v-sync 60, force feedback is smooth, strong, accurate, this is so much perfect ;o) Well today it was a hell, nothing worked, all the day Murphy's law everywhere... and now after 102 minutes in AC my last words was. "Oh everything works like it should works, everything is perfect, you are my best friend now buddy" and I slightly hugged my lcd ;o)

 

lmao