The online racing simulator
Quote from wolfshark :
Does not remove the autocentering - i tried js0, event0/1/2/3/4 but it doesn't do it.

js0 is the old joystick interface which doesn't work with FF in any way, you've got to use the "events".

ltwheelconf -b 0 --device=/dev/input/by-id/usb-Logitech_Logitech_Driving_Force-event-joystick

is the right one for me. Alternatively you can run "ls -l" in /dev/input/by-id" to check which eventX belongs to your wheel. I'd like to be absolutely sure the kernel method is not working before I start to hack the driver...
Quote :sudo ./ltwheelconf -b 0 --device=/dev/input/by-id/usb-Logitech_Logitech_Formula_Force_RX-event-joystick

Is not working. It's pointing to event3, but
Quote :sudo ./ltwheelconf -b 0 --device=/dev/input/event3

is not working either


Quote :Wheel autocenter force is now set to 0.

Is what i get from everything, but it isn't working
OK, then I guess it's safe to assume that the kernel glitch I suspect is really the cause.

I have this hack you can try out. It's not really a solution; what it does is that it disables autocentering altogether 'cause the resulting command will always contain zero autocentering force. If you find some spare time, you can give it a try...
Attached files
autocenter_hack.patch.txt - 415 B - 275 views
That the most awesome piece of information! Thanks a bunch for this, it clears things up a lot. The boundaries are however calculated not from the center, but from the fully turned position; the command you explained here sets range to 720 deg, not 180.

Any chance there's more information like this available? It would certainly help a great deal...
Some better info about commands for autocentering force would be helpful. LTWC is supposedly doing it right as it works for all wheels we've tested it with so far, but it needs two input values, rampspeed and centerforce. The kernel driver accepts only the rampspeed assuming that the centerforce is always 0x80. I don't think it's possible to pass both rampspeed and centerforce to the driver without breaking compatibility with the rest of the Linux force feedback stack.

I'm still waiting for someone with Formula EX/RX to test this theory, but it seems to me that EX/RX ignore the rampspeed and use just the centerforce which thus cannot be disabled. Problem is that without really knowing what the fields in the command are for I cannot just blindly patch the kernel driver and hope for the best...
Quote from lbodnar :
Going back to proper terms, do you mean
rampspeed = stiffness coefficient
centreforce = saturation force
and I assume the spring offset is zero?

Yes, exactly.

It appears that the driver adjusts the stiffness coeff proportionally to the saturation force, so the commands look like this

0% centering: FE 0D 00 00 00 00 00
50% centering: FE 0D 06 06 40 00 00
100% centering: FE 0D 0C 0C 80 00 00
150% centering: FE 0D 0F 0F FF 00 00

Knowing this I should be able to fix the driver, the only dilemma that's left is what should be considered the full centering force. Since there isn't any userspace interface yet that would allow to change overall effects strength, I guess I'll go with 150% for now...
With a bit of further investigation I created a proper kernel patch which fixes the autocentering. It applies OK against 2.6.39.1 and so far produces correct results for me. If you can find the time please try it out so I can submit it upstream.
Attached files
lg_autocenter.patch.txt - 1.7 KB - 334 views
cool thing! I'll test it with the G27 tomorrow, if i get myself to get my old wheel from the attic i'll test that, too.
Finaly my wheel is woring under linux! MadCatX you are great.
The strange thning is that outside of the game the center spring is working but the wheel isn't exactly centered I don't really care about that, really. The important think is that ingame the center spring is gone and i finally feel what the car is doing. There is only one tiny issue - roughly every 30 sec the wheel jurks left for, i don't know 0.1s or less. The pedals are still combined, but the patch has nothing to do with that. I don't use brake and throttle at the same time anyways.
Thanks a lot for trying the patch out, let's see if I can get it integrated into the mainline. As for the sudden jerk to the left, does in happen even in native linux games? It's possible that the FF in WINE still needs some improvements.
On the other hand the lg-ff driver might have issues even with the constant force commands, it hasn't been updated for a long while. It'll take some extra USB sniffing to verify that.

At this point I'd welcome some DirectInput sample apps and tutorials and about FF usage so I could generate constant force effects in Windoze and compare the commands to those of lg-ff driver.
Tried Speed Dreams - no force feedback, but i think that game doesn't have FF at all. Tried Vdrift from playdeb, but no force feedback there, i haven't tried to compile it - it's a nightmare app to compile.
Enabling FF in VDrift is a bit tricky. Open ~/.vdrift/VDrift.config file and look for something like

[ joystick ]
calibrated = false
ff_device = /dev/input/DFP
ff_gain = 1.235940
ff_invert = false
hgateshifter = false
two_hundred = false
type = wheel

The ff_device line must point to the event file of your wheel, otherwise FF doesn't work.

The patch has been sent upstream BTW, so let's hope I did everything right this time
I red about that, but that didn't work either, so i am compiling it
It's a b#tch - i hate thes game. I had to broke Ogre and Hugin to be able to install it. I am downloading the data and i will test it but for so many years haven't succeed with that stupid game.
Good luck with that, if you can't get it working, there is always libff. The "ffmvforce" test tool that comes with it could give you a clue if FF works correctly.
I tried the tests of fftest and they are not working. Also the center spring isn't switching off there. When trying to switch the center spring off with LTWheelConf or ffset it goes hard left.
Now that is really confusing. Do both methods of disabling autocentering in LTWC give you the same results?

EDIT: BTW the ""fftest" doesn't work for me either, try the "ffmvforce" instead.
Finaly Vdrift worked. The ffb is working, no autocentering in game But that game absolutely sucks! Very very heavy and not really good-looking graphics, stupid physics, stupid feel.

Now for the autocentering. Yesterday when i tried to disable autocentering both ways failed and made the wheel turn constantly left - this is dangerous for the motor, so i turned it to 90 again to center it.

When i play Vdrift and exit - there is no autocenter. When i play LFS and exit there is strange autocentering - stroing on the right and weak on the left. There are no sudden left jerking problems with Vdrift, so i guess that's wine problem. In LFS the jerks do happen, let's say once a lap for a very short time but they are really strong and ultimately they may destroy my wheel.
I don't seem to experience any sudden snaps to the left. Sometimes the wheel gets a bit shaky around the centre if I set its range low enough. That's just the LFS' physics engine trying to simulate the real force that's affecting the steering axle and because the range of your wheel is much lower than the range of the car's wheel, the effect is a bit garbled. I'm on WINE 1.3.22 BTW...

I confirm the problem with strange FF after exiting from LFS. If I use LTWC -b 0 command to disable autocentering, wheel quickly snaps to the left with kinda great force. -b 100 seems to reduce the tendency to turn to the left a bit, but I didn't find a way to get the wheel completely out of this weird state. I have a theory why this happens though and there's some discussion going on the LKML, so let's see if we can come up with some improvements.

Thanks a lot for all the testing
I've already done that but I didn't see any immediate problem. We'll try to do some more extensive fixes to the FF driver which will hopefully sort things out.
Bumpage:
We have a preliminary version of a bigger patch that seems to solve the weird autocentering after exiting WINE bug.

In order to do things properly we however need some help at this point. It looks like each wheel might be using a slightly different format of effect commands and we'd like to verify that.
For Windows there's a tool called USBLyzer (33 day trial) that can do USB traffic sniffing. If you want to help us out, here's how:

1) Plug your wheel in and let it calibrate.
2) Launch USBLyzer, in the left panel find your wheel and tick it (see screenshot).
3) Click "Start capture" and move you wheel a bit, you should see an avalanche of numbers in the main USBLyzer panel as the wheel sends updated position of it's axes to PC.
4) Open the calibration settings of your wheel in Control Panel (the one where you see a picture of the wheel and it's axes position)
5) Disconnect the AC power from the wheel (where applicable) and center the wheel.
6) Press a button on the wheel, you should see another flood of data in USBLyzer. Try not to move your wheel while the commands are being sent, otherwise the log will get contaminated by axes position updates. (it's not a big problem, but it'll make our job a bit harder)
7) Save the logs and upload them somewhere.

We already know what the commands should look like for DFP and Wii wheel, so what we need most is commands for G25/7 and Formula Force, but any other Logitech wheel won't hurt too.

Thanx.
Attached images
usblyzer.jpg
hm, maybe post that in a new thread, as no windows-users will see that here.

Sorry for not having tested the patch like i said before, i'm kinda busy the last few weeks and don't find much time to 'waste' I'll see to it this weekend.
I made what you said + one more dump for the throttle and brake. First i pressed the throttle, released it, then brake and released it. I guess that could help for separate axis.

When i pushed a button to initiate FFB i am 100% sure that the wheel didn't move, but on the calibrating app it moved slightly. Repeated the process now and the sam thing happened.
I am attaching the file here.
Attached files
FormulaEX.zip - 16.9 KB - 461 views
I guess I should've done it before, I've attached the screenshot how to properly select the capture source in USBLyzer. Unless it's done like that, the report is useless. Can you please try it again wolfshark? You don't have to capture pedals and wheel movement, we know how that's done. It's just the FF we're interested in. Thanks a lot.

@slim.one. That idea crossed my mind, but I don't think anyone would care Perhaps if we're hopelessly unsuccessful I'll open an extra thread.
Quote from MadCatX :That idea crossed my mind, but I don't think anyone would care

yeah that might be true, but never loose hope
I care, but it's a bit different. Here you go - only ffb, no wheel/pedals movement. Started capturing and then opened the calibration utility.
Attached files
FormulaEX FFB only.zip - 17.1 KB - 416 views

FGED GREDG RDFGDR GSFDG