Hi,
Ever since I got an FF wheel to use with GPL, I've been using 4ms Force
Feedback latency in core.ini as, IIRC, this was about what someone at
Papyrus suggested we use.
With my old model LWFF this worked like a charm. When I bought the MOMO
though, I got spikes in the FF using the same settings. After a lot of
tinkering I struck a compromise, but the wheel would start oscillating when
I let go, so I upped the damping quite a bit which in turn killed the feel,
so in the end I reached a 2nd and what I thought was a final compromise.
Then Mario Petrinovic suggested 21ms. Tried that eventhough it sounded like
a long time in comparison. No real spiking, but there was a sense of
disjointed FF, so I rejected the idea.
However, tonight, fellow MOMO owner Bart Westra suggested 35ms... even
longer!! Bart is not someone who'd pull your leg in such matters, though, so
I tried it and whattaya know... perfect. Slight tendency to oscillate on
compressions with hands off, but otherwise perfect.
Right, now I'm confused.
If I was smart I'd just shrug and start using 35ms, of course, but I'm not.
I've got to know now. What is a typical response time for a force feedback
wheel (in your opinion)? What are your experiences using low vs high FF
latency settings?
Jan.
=---
"Pay attention when I'm talking to you boy!" -Foghorn Leghorn.
"Jan Verschueren" <[email hidden]> wrote in
news:[email hidden]:
My understanding is that the best value will vary depending on
computer, wheel etc. If the latency is 0 GPL would pass the feedback
that is correct right at the moment, however because of the time
of calculations, overhead of drivers, latency of the communication
line and mechanics of the wheel you would feel feedback through
the wheel much later. Latency parameter should apparently tell GPL
how long does it take from the calculation of feedback value to the
moment when you get it on your wheel. It uses it by predicting what
feedback you should expect in this amount of time and sends it to
the wheel, so that when the wheel will react it will be correct
feedback for the moment it happens. So one should aim to find
correct latency parameter for his system.
The drawback of big latencies is that the further GPL predicts
less correct it becomes. For example,
when I exit Peralta at Mexico very close to the wall sometimes
I feel "touching the wall" effect on the steering wheel while
I didn't actually touch it.
Alex.
Quoted text (28 lines)
> Hi, > > Ever since I got an FF wheel to use with GPL, I've been using 4ms Force > Feedback latency in core.ini as, IIRC, this was about what someone at > Papyrus suggested we use. > > With my old model LWFF this worked like a charm. When I bought the MOMO > though, I got spikes in the FF using the same settings. After a lot of > tinkering I struck a compromise, but the wheel would start oscillating > when I let go, so I upped the damping quite a bit which in turn killed > the feel, so in the end I reached a 2nd and what I thought was a final > compromise. > > Then Mario Petrinovic suggested 21ms. Tried that eventhough it sounded > like a long time in comparison. No real spiking, but there was a sense > of disjointed FF, so I rejected the idea. > > However, tonight, fellow MOMO owner Bart Westra suggested 35ms... even > longer!! Bart is not someone who'd pull your leg in such matters, > though, so I tried it and whattaya know... perfect. Slight tendency to > oscillate on compressions with hands off, but otherwise perfect. > > Right, now I'm confused. > > If I was smart I'd just shrug and start using 35ms, of course, but I'm > not. I've got to know now. What is a typical response time for a force > feedback wheel (in your opinion)? What are your experiences using low > vs high FF latency settings?Quoted text (7 lines)
> > Jan. >=--- > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > > >I'm not sure what is typical for others, but with two different LWFFGP
wheels I have had to use 100 - 120 msec in order to get proper steering
response. Also note that in the GPL readme for the FF patch it says
that the 85 msec used in the core.ini.sample file was based on the MS
wheel. I think it is mostly a function of the wheel's motor, not so
much the computer. I could see maybe 5 - 10 msec latency in computer
processing and USB, but I doubt that it would be more than that.
Whereas DC motors (if that's what they are using) can certainly have
latencies in the 100's of milliseconds. The small, weak, cheap motors
in the less expensive wheels would be likely to have higher latencies
than the more powerful motors in more expensive products. Also the more
gear reduction it is using the more latency is likely, due to the higher
rpm of the motor. I would think a big direct drive servo motor would be
the best thing, but that's probably too expensive for home users. I
know there are some people from a couple of different wheel makers who
lurk here frequently, it would be interesting to hear what they think...
"Jan Verschueren" <[email hidden]> wrote in message
news:[email hidden]...
> Hi,
>
> Ever since I got an FF wheel to use with GPL, I've been using 4msForce
> Feedback latency in core.ini as, IIRC, this was about what someone at
> Papyrus suggested we use.
>
> With my old model LWFF this worked like a charm. When I bought theMOMO
> though, I got spikes in the FF using the same settings. After a lot of
> tinkering I struck a compromise, but the wheel would start oscillatingwhen
> I let go, so I upped the damping quite a bit which in turn killed thefeel,
> so in the end I reached a 2nd and what I thought was a finalcompromise.
>
> Then Mario Petrinovic suggested 21ms. Tried that eventhough it soundedlike
> a long time in comparison. No real spiking, but there was a sense of
> disjointed FF, so I rejected the idea.
>
> However, tonight, fellow MOMO owner Bart Westra suggested 35ms... even
> longer!! Bart is not someone who'd pull your leg in such matters,though, so
> I tried it and whattaya know... perfect. Slight tendency to oscillateon
> compressions with hands off, but otherwise perfect.
>
> Right, now I'm confused.
>
> If I was smart I'd just shrug and start using 35ms, of course, but I'mnot.
> I've got to know now. What is a typical response time for a forcefeedback
> wheel (in your opinion)? What are your experiences using low vs highFF
Quoted text (7 lines)
> latency settings? > > Jan. > =--- > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > >Jan,
It is highly dependant on the speed of your system (CPU, memory, graphics,
etc.) and the speed of your wheel connection (gameport vs. serial vs. USB).
That is, even the same wheel on the same system would require a different
core.ini setting if you were using the serial connection vs. the USB (for
those few wheels that come with more than one connecter).
I spent months crafting the GPL settings and with a fast system I can find
nothing that works better than 0. I don't want any software predicting
(which by definition is imperfect) if I don't need it. But you have to
experiment to see what works for you. Just pick a track with big curbing or
something that you can feel very well and try driving into it at fairly slow
speeds. You'll know if the reaction in your hands is not matching what you
are seeing on the screen. Unless you have a fairly slow machine, the feel
in your hands should be a natural extension of what you are seeing on the
screen--not delayed, or, exaggerated because it is being predicted.
Marc.
"Jan Verschueren" <[email hidden]> wrote in message
news:[email hidden]...
when
> I let go, so I upped the damping quite a bit which in turn killed thefeel,
> so in the end I reached a 2nd and what I thought was a final compromise.
>
> Then Mario Petrinovic suggested 21ms. Tried that eventhough it soundedlike
> a long time in comparison. No real spiking, but there was a sense of
> disjointed FF, so I rejected the idea.
>
> However, tonight, fellow MOMO owner Bart Westra suggested 35ms... even
> longer!! Bart is not someone who'd pull your leg in such matters, though,so
> I tried it and whattaya know... perfect. Slight tendency to oscillate on
> compressions with hands off, but otherwise perfect.
>
> Right, now I'm confused.
>
> If I was smart I'd just shrug and start using 35ms, of course, but I'mnot.
Quoted text (9 lines)
> Hi, > > Ever since I got an FF wheel to use with GPL, I've been using 4ms Force > Feedback latency in core.ini as, IIRC, this was about what someone at > Papyrus suggested we use. > > With my old model LWFF this worked like a charm. When I bought the MOMO > though, I got spikes in the FF using the same settings. After a lot of > tinkering I struck a compromise, but the wheel would start oscillatingQuoted text (9 lines)
> I've got to know now. What is a typical response time for a force feedback > wheel (in your opinion)? What are your experiences using low vs high FF > latency settings? > > Jan. > =--- > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > >Bumping into things is not a very good way of calibrating latency IMO.
I don't even believe it exercises the same part of the physics model as
the parameter that you do want to get synchronized, which is the front
wheel alignment torque. The way to calibrate it is in the readme file -
gradually increase latency until the forces that you get when swerving
back and forth at speed are in synch with your steering motions. If you
don't do it this way then the first time you make a mistake at speed you
will find it very difficult to recover due to the delay in the forces.
That is, you will make a correction, get it straight, and then the
steering force will hit and send you off in the other direction. For me
at least, the cars feel a lot more stable when I set the latency this
way. And setting it in this manner resulted in a much higher latency
correction than the bump method, on my system.
Regards,
Hal
Who makes a lot of mistakes at speed ;o)
"Marc Collins" <[email hidden]> wrote in message
news:Wy8%7.9237$[email hidden]...
> Jan,
>
> It is highly dependant on the speed of your system (CPU, memory,graphics,
> etc.) and the speed of your wheel connection (gameport vs. serial vs.USB).
> That is, even the same wheel on the same system would require adifferent
> core.ini setting if you were using the serial connection vs. the USB(for
> those few wheels that come with more than one connecter).
>
> I spent months crafting the GPL settings and with a fast system I canfind
> nothing that works better than 0. I don't want any softwarepredicting
> (which by definition is imperfect) if I don't need it. But you haveto
> experiment to see what works for you. Just pick a track with bigcurbing or
> something that you can feel very well and try driving into it atfairly slow
> speeds. You'll know if the reaction in your hands is not matchingwhat you
> are seeing on the screen. Unless you have a fairly slow machine, thefeel
> in your hands should be a natural extension of what you are seeing onthe
> screen--not delayed, or, exaggerated because it is being predicted.
>
> Marc.
>
>
> "Jan Verschueren" <[email hidden]> wrote inmessage
> news:[email hidden]...
> > Hi,
> >
> > Ever since I got an FF wheel to use with GPL, I've been using 4msForce
> > Feedback latency in core.ini as, IIRC, this was about what someoneat
> > Papyrus suggested we use.
> >
> > With my old model LWFF this worked like a charm. When I bought theMOMO
> > though, I got spikes in the FF using the same settings. After a lotof
> > tinkering I struck a compromise, but the wheel would startoscillating
> when
> > I let go, so I upped the damping quite a bit which in turn killedthe
> feel,
> > so in the end I reached a 2nd and what I thought was a finalcompromise.
> >
> > Then Mario Petrinovic suggested 21ms. Tried that eventhough itsounded
> like
> > a long time in comparison. No real spiking, but there was a sense of
> > disjointed FF, so I rejected the idea.
> >
> > However, tonight, fellow MOMO owner Bart Westra suggested 35ms...even
> > longer!! Bart is not someone who'd pull your leg in such matters,though,
> so
> > I tried it and whattaya know... perfect. Slight tendency tooscillate on
> > compressions with hands off, but otherwise perfect.
> >
> > Right, now I'm confused.
> >
> > If I was smart I'd just shrug and start using 35ms, of course, butI'm
> not.
> > I've got to know now. What is a typical response time for a forcefeedback
> > wheel (in your opinion)? What are your experiences using low vs highFF
Quoted text (9 lines)
> > latency settings? > > > > Jan. > > =--- > > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > > > > > >Good point. The 0 works well in all situations on my machine, but it may be
fast enough that it doesn't matter. On a system that is closer to the
limit, I am sure there is a difference in the swerving forces (which are a
lot more subtle) vs. the simple spring compression over a curb (kerb).
Marc
"Haqsau" <[email hidden]> wrote in message
news:a1kgfm$r5jjt$[email hidden]...
Quoted text (112 lines)
> Bumping into things is not a very good way of calibrating latency IMO. > I don't even believe it exercises the same part of the physics model as > the parameter that you do want to get synchronized, which is the front > wheel alignment torque. The way to calibrate it is in the readme file - > gradually increase latency until the forces that you get when swerving > back and forth at speed are in synch with your steering motions. If you > don't do it this way then the first time you make a mistake at speed you > will find it very difficult to recover due to the delay in the forces. > That is, you will make a correction, get it straight, and then the > steering force will hit and send you off in the other direction. For me > at least, the cars feel a lot more stable when I set the latency this > way. And setting it in this manner resulted in a much higher latency > correction than the bump method, on my system. > > Regards, > Hal > Who makes a lot of mistakes at speed ;o) > > > "Marc Collins" <[email hidden]> wrote in message > news:Wy8%7.9237$[email hidden]... > > Jan, > > > > It is highly dependant on the speed of your system (CPU, memory, > graphics, > > etc.) and the speed of your wheel connection (gameport vs. serial vs. > USB). > > That is, even the same wheel on the same system would require a > different > > core.ini setting if you were using the serial connection vs. the USB > (for > > those few wheels that come with more than one connecter). > > > > I spent months crafting the GPL settings and with a fast system I can > find > > nothing that works better than 0. I don't want any software > predicting > > (which by definition is imperfect) if I don't need it. But you have > to > > experiment to see what works for you. Just pick a track with big > curbing or > > something that you can feel very well and try driving into it at > fairly slow > > speeds. You'll know if the reaction in your hands is not matching > what you > > are seeing on the screen. Unless you have a fairly slow machine, the > feel > > in your hands should be a natural extension of what you are seeing on > the > > screen--not delayed, or, exaggerated because it is being predicted. > > > > Marc. > > > > > > "Jan Verschueren" <[email hidden]> wrote in > message > > news:[email hidden]... > > > Hi, > > > > > > Ever since I got an FF wheel to use with GPL, I've been using 4ms > Force > > > Feedback latency in core.ini as, IIRC, this was about what someone > at > > > Papyrus suggested we use. > > > > > > With my old model LWFF this worked like a charm. When I bought the > MOMO > > > though, I got spikes in the FF using the same settings. After a lot > of > > > tinkering I struck a compromise, but the wheel would start > oscillating > > when > > > I let go, so I upped the damping quite a bit which in turn killed > the > > feel, > > > so in the end I reached a 2nd and what I thought was a final > compromise. > > > > > > Then Mario Petrinovic suggested 21ms. Tried that eventhough it > sounded > > like > > > a long time in comparison. No real spiking, but there was a sense of > > > disjointed FF, so I rejected the idea. > > > > > > However, tonight, fellow MOMO owner Bart Westra suggested 35ms... > even > > > longer!! Bart is not someone who'd pull your leg in such matters, > though, > > so > > > I tried it and whattaya know... perfect. Slight tendency to > oscillate on > > > compressions with hands off, but otherwise perfect. > > > > > > Right, now I'm confused. > > > > > > If I was smart I'd just shrug and start using 35ms, of course, but > I'm > > not. > > > I've got to know now. What is a typical response time for a force > feedback > > > wheel (in your opinion)? What are your experiences using low vs high > FF > > > latency settings? > > > > > > Jan. > > > =--- > > > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > > > > > > > > > > >Hi Marc,
it's not just a matter of 'working well' :-)
With the right lat setting, you can get the feedback a tad earlier, thus
being able to react quicker.
Achim
"Marc Collins" <[email hidden]> wrote in message
news:Wvu%7.14372$[email hidden]...
> Good point. The 0 works well in all situations on my machine, but it maybe
Quoted text (122 lines)
> fast enough that it doesn't matter. On a system that is closer to the > limit, I am sure there is a difference in the swerving forces (which are a > lot more subtle) vs. the simple spring compression over a curb (kerb). > > Marc > > "Haqsau" <[email hidden]> wrote in message > news:a1kgfm$r5jjt$[email hidden]... > > Bumping into things is not a very good way of calibrating latency IMO. > > I don't even believe it exercises the same part of the physics model as > > the parameter that you do want to get synchronized, which is the front > > wheel alignment torque. The way to calibrate it is in the readme file - > > gradually increase latency until the forces that you get when swerving > > back and forth at speed are in synch with your steering motions. If you > > don't do it this way then the first time you make a mistake at speed you > > will find it very difficult to recover due to the delay in the forces. > > That is, you will make a correction, get it straight, and then the > > steering force will hit and send you off in the other direction. For me > > at least, the cars feel a lot more stable when I set the latency this > > way. And setting it in this manner resulted in a much higher latency > > correction than the bump method, on my system. > > > > Regards, > > Hal > > Who makes a lot of mistakes at speed ;o) > > > > > > "Marc Collins" <[email hidden]> wrote in message > > news:Wy8%7.9237$[email hidden]... > > > Jan, > > > > > > It is highly dependant on the speed of your system (CPU, memory, > > graphics, > > > etc.) and the speed of your wheel connection (gameport vs. serial vs. > > USB). > > > That is, even the same wheel on the same system would require a > > different > > > core.ini setting if you were using the serial connection vs. the USB > > (for > > > those few wheels that come with more than one connecter). > > > > > > I spent months crafting the GPL settings and with a fast system I can > > find > > > nothing that works better than 0. I don't want any software > > predicting > > > (which by definition is imperfect) if I don't need it. But you have > > to > > > experiment to see what works for you. Just pick a track with big > > curbing or > > > something that you can feel very well and try driving into it at > > fairly slow > > > speeds. You'll know if the reaction in your hands is not matching > > what you > > > are seeing on the screen. Unless you have a fairly slow machine, the > > feel > > > in your hands should be a natural extension of what you are seeing on > > the > > > screen--not delayed, or, exaggerated because it is being predicted. > > > > > > Marc. > > > > > > > > > "Jan Verschueren" <[email hidden]> wrote in > > message > > > news:[email hidden]... > > > > Hi, > > > > > > > > Ever since I got an FF wheel to use with GPL, I've been using 4ms > > Force > > > > Feedback latency in core.ini as, IIRC, this was about what someone > > at > > > > Papyrus suggested we use. > > > > > > > > With my old model LWFF this worked like a charm. When I bought the > > MOMO > > > > though, I got spikes in the FF using the same settings. After a lot > > of > > > > tinkering I struck a compromise, but the wheel would start > > oscillating > > > when > > > > I let go, so I upped the damping quite a bit which in turn killed > > the > > > feel, > > > > so in the end I reached a 2nd and what I thought was a final > > compromise. > > > > > > > > Then Mario Petrinovic suggested 21ms. Tried that eventhough it > > sounded > > > like > > > > a long time in comparison. No real spiking, but there was a sense of > > > > disjointed FF, so I rejected the idea. > > > > > > > > However, tonight, fellow MOMO owner Bart Westra suggested 35ms... > > even > > > > longer!! Bart is not someone who'd pull your leg in such matters, > > though, > > > so > > > > I tried it and whattaya know... perfect. Slight tendency to > > oscillate on > > > > compressions with hands off, but otherwise perfect. > > > > > > > > Right, now I'm confused. > > > > > > > > If I was smart I'd just shrug and start using 35ms, of course, but > > I'm > > > not. > > > > I've got to know now. What is a typical response time for a force > > feedback > > > > wheel (in your opinion)? What are your experiences using low vs high > > FF > > > > latency settings? > > > > > > > > Jan. > > > > =--- > > > > "Pay attention when I'm talking to you boy!" -Foghorn Leghorn. > > > > > > > > > > > > > > > > > >Interesting reading guys... thanks.
Jan.
=---
"Pay attention when I'm talking to you boy!" -Foghorn Leghorn.
I like the feedback to feel natural, which it does at 0 on my system. Are
you suggesting that hotlappers can actually get GPL to predict the feedback
a tad early and then take advantage of that? You can't get any lower/faster
than 0 for us regular folks, can you?
Marc
"Joachim Trensz" <[email hidden]> wrote in message
news:a1n81s$rqkv9$[email hidden]...
may
> be
> > fast enough that it doesn't matter. On a system that is closer to the
> > limit, I am sure there is a difference in the swerving forces (which area
as
> > > the parameter that you do want to get synchronized, which is the front
> > > wheel alignment torque. The way to calibrate it is in the readmefile -
> > > gradually increase latency until the forces that you get when swerving
> > > back and forth at speed are in synch with your steering motions. Ifyou
> > > don't do it this way then the first time you make a mistake at speedyou
> > > will find it very difficult to recover due to the delay in the forces.
> > > That is, you will make a correction, get it straight, and then the
> > > steering force will hit and send you off in the other direction. Forme
vs.
can
the
> > > feel
> > > > in your hands should be a natural extension of what you are seeingon
lot
of
high
