Ruud van Gaal wrote:
> First, a question about tire friction; how does a graph look when
> showing the friction coefficient vs. slip angle
It's probably easier to find a reference book and look at a graph
from experimental results than wandering around in the dark. ;)
> Take for example a motorbike spinning its rear
> wheel into smoke, while holding the front steady using the brakes.
> How would I describe this function, maxFriction/(|v|+1) or something
> like that?
There are several often used ways to relate tire longitudinal force
to slip ratio (several definitions of slip ratio) based on experimental
observations. "Race Car Vehicle Dynamics", for example, lists
4 or 5 of them and even shows how to convert between them.
> Is there an end to the lowering of the frictioncoefficient when
> spinning madly, or is there a certain minimum?
Notions about how this should look for various tires
are in every reference book with a chapter on tire modeling
(which is every text book about cars, and most books
on how to drive them ;)
> I now use a function which linearly lets the friction dive from
> max..min depending on |v|, and I calculate min(Friction) as
> maxFriction/4.
It depends on the normal force and the tire properties
of course, but a 75% reduction sounds like far more
than is supported by what I recall from the literature.
More like 20%, often less. It depends on the tire. ;)
It's not something that is easy to test for, BTW.
Test sessions destroy tires, they heat up quickly
and the results change markedly with temperature, etc., etc.
Once you get to where you can use the results,
you then can begin to distrust them ;)
With that 75% reduction you will probably still be in wheelspin
up to a very high speed... All these modeling parameters
affect many many things. Getting even one of them very wrong
will cause large problems somewhere. Probably several places...
> But I'm not an expert on suspensions.
I wasn't (and am not). That's why I read books. ;)
No point in guessing when it's right there on page 12 ;)
Even for a hobby, my time is worth more to me than the
cost of a reasonable amount of reference material.
You could spend literally hundreds of hours in confusion.
A hundred hours of study would be a better choice. ;)
> First hmm: when calculating the force of the spring on the body and
> the wheel, I now divide the force into force/2 on the body, and
> -force/2 on the wheels.
> Should I instead apply the total spring force
> on both ends?
Yes, most emphatically.
As an added check, the sum of the forces upwards on
the springs should equal the weight of the car (less the
weight of the wheels themselves.) This keeps the car from flying
off the ground, or sinking into the ground. ;)
> What would an appropriate real value be? GPL uses 'bump==2' for
> example, which doesn't say a whole lot. Seems the damper unit is
> something like N/(m/s), so Newton per meter per second.
Well, calculate the N/(m/s) to give a damping ratio of say 0.6
in jounce and 1.0 (critical) in rebound as a first guess.
For a generic sedan racing car, give it say a 2 to 3 Hz heave
frequency.
Seriously, assuming the mass properties are known,
the two sentences above are enough for first estimates
of the spring stiffnesses and damper settings.
I realize I haven't defined some of those terms and
ideas, but they are very very standard. (see reference below)
There are simple equations (first year undergraduate physics)
that can help you analyze and simulate spring and damper
suspension systems and even tell you what values are
typical for various classes of vehicles.
�Fundamentals of Vehicle Dynamics,� Thomas Gillespie,
SAE, Danvers, MA, 1992. ISBN 1-56091-199-9
Covers:
1. Tire models
2. Definitions of common simulation terms and coordinate systems
(an SAE Recommended practice as an appendix)
3. The SAE "Ride and Vibration Data Manual" appendix that
with study will answer the questions you asked about
suspension settings.
4 through 100... ;);)
You are progressing to areas where
the answers are well known and well described in
even the first undergraduate text books on
vehicle dynamics. That level of interest, study, and understanding
is what it will take to do a good job from here on out.
If you don't have a technical background, you will also have
to soak up engineering ways of thinking and working along the way...
Enjoy ;)
- Matt
Notable threadHobbyist car physics: tyre slip ratios and low-speed instability
Continuing Ruud van Gaal's physics thread, Matt Jessick of Motorsims, Todd Wasson, Gregor Veble and Doug Arnao explained slip ratio definitions, pneumatic trail and tyre load sensitivity. Monster Games' Dave Pollatsek described fixing Viper Racing's hill-sliding bug in NASCAR Heat by modelling sidewall shear springs. Later posts dealt with numerical instability and wheel-lock reversal at low speeds, and F1 2000's high friction values.
With Dave Pollatsek (Monster Games); Matthew Jessick (Motorsims)
physics/realism debateRacer
Automated summary.
On Mon, 04 Sep 2000 23:19:59 GMT, Doug Arnao <[email hidden]>
wrote:
...
>For longitudinal traction? Thats called a slip ratio. Max traction is
>about at 8% tire slip in general. It builds rather sharpley from 0-8% then
>levels off with a slight down slope to about 20%, then falls abruptly
>after that ..........see Carroll Smith "Tune to Win" chapter two.....will
>tell you all you need to know about this including graphs.
Thanks. That book is high on the list. I've ordered "Race Vehicle
Dynamics" from the Milliken bunch and am anxiously awaiting it.
I think it may have some similar graphs in it.
Apart from that, I must say I'm puzzled a bit, because:
slip_ratio=rotational_speed_of_wheel/actual_speed_of_wheel_vs_ground
where rotational_speed_of_wheel is the speed at which the wheel is
trying to move the vehicle (=ang.vel.around_axle*radius).
I would say the *difference* of the 2 speeds is what you would use to
describe the friction coefficient (CF). If the wheel is trying to push
the vehicle forward at 200km/h, and the vehicle is actually moving
100km/h, the slip ratio is 200%. But there's 100km/h of tire surface
moving across the road. When driving 10km/h, but the wheels are doing
20km/h, the slip ratio is also 200%, and this leads to the same CF?
Seems intuitive to me.
...
>If you look at the graph in tune to win, it's certainly not linear and it
>never goes to zero. The max percent of slip is only 27% and at that point
>it is generating a CF of .75. A spinning tire is still making a good
>amount of traction - sliding traction.
Ok, good to know. I guess after the 27% the CF more or less (less)
remains at .75. Well, a little below that.
Yes, I have 4 wheel objects, that actually have a seperate suspension
system (spring+damper) for each. Todd states I have to exert the force
fully both ways.
To explain more simply, suppose I hold a spring in my hands and
compress it using both hands. The force can be calulated using
k(x0-x). Would the forces on BOTH my hands be that number or is the
force divided over the two hands? I wondered about that because if you
begin to think about energy (W=F*d) then the total energy if both
hands start to move out (so the spring stretches again) would be twice
the total energy available in the spring. But I'm not taking into
account the fact probably that if one hand starts to move, the force
on the other hand becomes less and the total sum of W=F*d remains
equal to the energy stored in the spring.
Hope this is clearer? :)
...
>The real life damper only sees the velocity of the shock shaft - that the
>piston is attached to. If your simple model is traveling the damper at the
>same travel (1:1) with the spring and the tire, then its velocity will be
>the same.
Yes, my damper is stuck to the body. So the shaft velocity is the
velocity at which the tire is moving to or away from the damper.
...
<g> no problem.
I'm using the metric system at all places, because that's a whole lot
less constants everywhere (and I'm European, so used to it). I
downloaded a fantastic conversion program btw yesterday at
http://www.joshmadison.com/software/ (Convert), which really has a
zillion units to convert from and to. Used it to convert the air
density from lb-mass/ft^3 to kg/m^3. It's so easy to get things buggy
when you think in the wrong unit system.
> To answer your question - somewhere between 1 and 3
>times the spring value should get you close. Try 1x on the compresion and
>2x on the extension.
Ok. And real-life spring values? I see about 150Nm wheelrate values
(which isn't exactly spring rate, but comes close) in GPL, but in
F12000 I see spring coefficients of 140000 (140K), presumably Nm, as
it seems to use the metric system. Or are F1 springs these days indeed
so hard wrt 1967?
Strange thing is, a rebound value (for F1-2000) is 3000 for example,
with spring_coeff=140000. Not even close to the 1-3 times the spring
value? Where do I go wrong here? :)
Ah, in the mean time, the book has arrived! Wow, technical stuff.
Should do some reading tonight. I have DirectInput working btw. Spent
yesterday evening exploring DirectInput. Quite nice, and now I have my
MSFF wheel working (no force feedback yet ofcourse). It just has funny
values for split axis throttle/braking (even though I set them to
-1000..+1000, the Y/throttle axis stays on 1..-1000, yes, no typo).
But I can get a much better sense of how my car is handling. It's far
too sensitive now :) Must be the friction going too low.
...
>Very interesting project Ruud...............buy the Carroll Smith books,
>you'll have alot less questions :)
Thanks, I now have Race Vehicle Dynamics from Milliken & Milliken, so
that will keep me busy (for eternity probably, hehe), and keep me a
little less questionhappy here on RAS.
Thanks to all the help, I'll keep you posted about the progress, will
do some studying in the book, and when a new update appears on my
site,
Ruud van Gaal, GPL Rank +53.53
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (15 lines)
>> Next, I'm doing a suspension system. I'm already seeing a lot of >> parameters in this field, but I want to have a relatively simple >> representation currently. Mainly to see the car (the body rather) tilt >> under power and steering. But I'm not an expert on suspensions. >> I use a plain vertical suspension with a spring and damper. The >> springs have a spring constant (k) so the force generated when >> compressing/stretching is k(x0-x) (a known formula). >> First hmm: when calculating the force of the spring on the body and >> the wheel, I now divide the force into force/2 on the body, and >> -force/2 on the wheels. Should I instead apply the total spring force >> on both ends? (in the end, it applies the force to bring both body and >> wheel together, that's why I divide the force between both ends) > >Not sure.......you're not explaining it clear enough for me to understand. >Do you have 4 seperate springs?Quoted text (11 lines)
>> What would an appropriate real value be? GPL uses 'bump==2' for >> example, which doesn't say a whole lot. Seems the damper unit is >> something like N/(m/s), so Newton per meter per second. > >The Americans use Lbs/(in/sec) typical shaft speed at a 1:1 suspension >could be as high as 30 inches per second > >Bilstein uses N/(m/s). Actually Nascar Heat uses the Bilstein valving >numbers, because most of Nascar use Bilstein shocks! A valving of 300 is >actually 3000nm force @ 20m/sec piston speed...................sorry, I'm >rambling there <g>.
Ruud van Gaal wrote:
Might be totally wrong here, but I would think 100% slip ratio would be
correct. 100% above the traveling speed??? In other words the speed you are
traveling is the starting point or ground "zero" <g>. make sense?
As to weather your two examples have the same CF.........you would think so by
the numbers. But rubber is a very wierd and amazing substance that in my
experience dosen't act linearly.
Just remeber one thing : All the high dollar fancy, trick, perfectley designed
suspension systems eventually have to transmit their numbers through the tire
contact patch. About 15 square inches of nylon and rubber..........This holds
true for your sim ,too. I've always felt that the tire model makes or breaks a
sim. I believe this is one of the things that Nascar Heat *nailed*.
Todd was correct, yes split between the two the total force.
Yes and your last sentance seems correct to me - yeah, what you said <g>.
We typically use wheel rates (the number out at the wheel) on our Porsche race
cars from 200-600 lbs approx. - this after going out thru the motion ratio (the
lever arms) of the suspension/shock.
Last I paid attention to F1 spring rates, they were using springs as high as
5000 lbs and as low as 800. Again, the wheel rates could be 1/4 to 1/2 of that
amount depending on the motion ratio's used. Downforce plays a huge role in
determining the rate, of course
F12000 numbers look a bit outa sight - I wouldn't try to make those fit
anywhere. They
>
> Strange thing is, a rebound value (for F1-2000) is 3000 for example,
> with spring_coeff=140000. Not even close to the 1-3 times the spring
> value? Where do I go wrong here? :)
No........ where did F12K go wrong <g>.
If a 100 lb spring travels 3 inches - it has 300lbs of stored energy that needs
to be dissapated on the way down (rebound) - or absorbed by the damper. That
is in round numbers of course and assuming you know the shaft speed when the
spring is unloading :)
Just as a matter of hit and miss, we've found for a race car using 2-3inches
travel starting about 2 times the spring force in rebound resistance, then 1/3
that for the rear compression and then 1/2 for the front compression. So in US
LBS units If you had 300 front spring and 500 rear spring you'd get
compression/rebound values of:
front 125/450
rear 333/1000
Millikens book is a tough read - for me at least. However, someone like you
should be in formula and constant heaven :)
--Doug
Quoted text (25 lines)
> On Mon, 04 Sep 2000 23:19:59 GMT, Doug Arnao <[email hidden]> > wrote: > > ... > >For longitudinal traction? Thats called a slip ratio. Max traction is > >about at 8% tire slip in general. It builds rather sharpley from 0-8% then > >levels off with a slight down slope to about 20%, then falls abruptly > >after that ..........see Carroll Smith "Tune to Win" chapter two.....will > >tell you all you need to know about this including graphs. > > Thanks. That book is high on the list. I've ordered "Race Vehicle > Dynamics" from the Milliken bunch and am anxiously awaiting it. > I think it may have some similar graphs in it. > Apart from that, I must say I'm puzzled a bit, because: > slip_ratio=rotational_speed_of_wheel/actual_speed_of_wheel_vs_ground > where rotational_speed_of_wheel is the speed at which the wheel is > trying to move the vehicle (=ang.vel.around_axle*radius). > > I would say the *difference* of the 2 speeds is what you would use to > describe the friction coefficient (CF). If the wheel is trying to push > the vehicle forward at 200km/h, and the vehicle is actually moving > 100km/h, the slip ratio is 200%. But there's 100km/h of tire surface > moving across the road. When driving 10km/h, but the wheels are doing > 20km/h, the slip ratio is also 200%, and this leads to the same CF? > Seems intuitive to me.Quoted text (34 lines)
> > > ... > >If you look at the graph in tune to win, it's certainly not linear and it > >never goes to zero. The max percent of slip is only 27% and at that point > >it is generating a CF of .75. A spinning tire is still making a good > >amount of traction - sliding traction. > > Ok, good to know. I guess after the 27% the CF more or less (less) > remains at .75. Well, a little below that. > > >> Next, I'm doing a suspension system. I'm already seeing a lot of > >> parameters in this field, but I want to have a relatively simple > >> representation currently. Mainly to see the car (the body rather) tilt > >> under power and steering. But I'm not an expert on suspensions. > >> I use a plain vertical suspension with a spring and damper. The > >> springs have a spring constant (k) so the force generated when > >> compressing/stretching is k(x0-x) (a known formula). > >> First hmm: when calculating the force of the spring on the body and > >> the wheel, I now divide the force into force/2 on the body, and > >> -force/2 on the wheels. Should I instead apply the total spring force > >> on both ends? (in the end, it applies the force to bring both body and > >> wheel together, that's why I divide the force between both ends) > > > >Not sure.......you're not explaining it clear enough for me to understand. > >Do you have 4 seperate springs? > > Yes, I have 4 wheel objects, that actually have a seperate suspension > system (spring+damper) for each. Todd states I have to exert the force > fully both ways. > To explain more simply, suppose I hold a spring in my hands and > compress it using both hands. The force can be calulated using > k(x0-x). Would the forces on BOTH my hands be that number or is the > force divided over the two hands?Quoted text (9 lines)
> I wondered about that because if you > begin to think about energy (W=F*d) then the total energy if both > hands start to move out (so the spring stretches again) would be twice > the total energy available in the spring. But I'm not taking into > account the fact probably that if one hand starts to move, the force > on the other hand becomes less and the total sum of W=F*d remains > equal to the energy stored in the spring. > Hope this is clearer? :) >Quoted text (41 lines)
> > ... > >The real life damper only sees the velocity of the shock shaft - that the > >piston is attached to. If your simple model is traveling the damper at the > >same travel (1:1) with the spring and the tire, then its velocity will be > >the same. > > Yes, my damper is stuck to the body. So the shaft velocity is the > velocity at which the tire is moving to or away from the damper. > > ... > >> What would an appropriate real value be? GPL uses 'bump==2' for > >> example, which doesn't say a whole lot. Seems the damper unit is > >> something like N/(m/s), so Newton per meter per second. > > > >The Americans use Lbs/(in/sec) typical shaft speed at a 1:1 suspension > >could be as high as 30 inches per second > > > >Bilstein uses N/(m/s). Actually Nascar Heat uses the Bilstein valving > >numbers, because most of Nascar use Bilstein shocks! A valving of 300 is > >actually 3000nm force @ 20m/sec piston speed...................sorry, I'm > >rambling there <g>. > > <g> no problem. > I'm using the metric system at all places, because that's a whole lot > less constants everywhere (and I'm European, so used to it). I > downloaded a fantastic conversion program btw yesterday at > http://www.joshmadison.com/software/ (Convert), which really has a > zillion units to convert from and to. Used it to convert the air > density from lb-mass/ft^3 to kg/m^3. It's so easy to get things buggy > when you think in the wrong unit system. > > > To answer your question - somewhere between 1 and 3 > >times the spring value should get you close. Try 1x on the compresion and > >2x on the extension. > > Ok. And real-life spring values? I see about 150Nm wheelrate values > (which isn't exactly spring rate, but comes close) in GPL, but in > F12000 I see spring coefficients of 140000 (140K), presumably Nm, as > it seems to use the metric system. Or are F1 springs these days indeed > so hard wrt 1967?Quoted text (18 lines)
> > > Ah, in the mean time, the book has arrived! Wow, technical stuff. > Should do some reading tonight. I have DirectInput working btw. Spent > yesterday evening exploring DirectInput. Quite nice, and now I have my > MSFF wheel working (no force feedback yet ofcourse). It just has funny > values for split axis throttle/braking (even though I set them to > -1000..+1000, the Y/throttle axis stays on 1..-1000, yes, no typo). > But I can get a much better sense of how my car is handling. It's far > too sensitive now :) Must be the friction going too low. > > ... > >Very interesting project Ruud...............buy the Carroll Smith books, > >you'll have alot less questions :) > > Thanks, I now have Race Vehicle Dynamics from Milliken & Milliken, so > that will keep me busy (for eternity probably, hehe), and keep me a > little less questionhappy here on RAS.Doug is right and I think a lot fo games miss a few things here. One is
pneumatics trail that every tire has. Lower the aspect ratio Iwould
venture the lower the pneumatic trail of the tire.
also one thing I noticed only one game so far has gotten right and I will
not name which one. In my Solo ll car when I get understeer and I steer
more into the turn the more understeer I get, the less I turn the wheel the
less the understeer is. I think Jackie Stewart explained it that the tire
tread face is not addressing the road properly, but I do think it relates a
bit to the above also. With sticky slicks it becomes very noticable so
much so once I actaully turned in had a bunch of understeer and reduced
steering lock and the car scooted thru the corner so quickly I was at the
exit with my foot nowhhere near the throttle not to mention even thinking
about full throttle.
Dave
Doug Arnao <[email hidden]> wrote in article
<[email hidden]>...
then
> > >levels off with a slight down slope to about 20%, then falls abruptly
> > >after that ..........see Carroll Smith "Tune to Win" chaptertwo.....will
are
> traveling is the starting point or ground "zero" <g>. make sense?
>
> As to weather your two examples have the same CF.........you would thinkso by
> the numbers. But rubber is a very wierd and amazing substance that in my
> experience dosen't act linearly.
>
> Just remeber one thing : All the high dollar fancy, trick, perfectleydesigned
> suspension systems eventually have to transmit their numbers through thetire
> contact patch. About 15 square inches of nylon and rubber..........Thisholds
> true for your sim ,too. I've always felt that the tire model makes orbreaks a
> sim. I believe this is one of the things that Nascar Heat *nailed*.
>
> >
> >
> > ...
> > >If you look at the graph in tune to win, it's certainly not linear andit
> > >never goes to zero. The max percent of slip is only 27% and at thatpoint
tilt
force
> > >> on both ends? (in the end, it applies the force to bring both bodyand
> > >> wheel together, that's why I divide the force between both ends)
> > >
> > >Not sure.......you're not explaining it clear enough for me tounderstand.
<g>.
>
> >
> > ...
> > >The real life damper only sees the velocity of the shock shaft - thatthe
> > >piston is attached to. If your simple model is traveling the damper atthe
> > >same travel (1:1) with the spring and the tire, then its velocity willbe
is
> > >actually 3000nm force @ 20m/sec piston speed...................sorry,I'm
and
race
> cars from 200-600 lbs approx. - this after going out thru the motionratio (the
> lever arms) of the suspension/shock.
>
> Last I paid attention to F1 spring rates, they were using springs as highas
> 5000 lbs and as low as 800. Again, the wheel rates could be 1/4 to 1/2 of that
> amount depending on the motion ratio's used. Downforce plays a huge rolein
needs
> to be dissapated on the way down (rebound) - or absorbed by the damper.That
> is in round numbers of course and assuming you know the shaft speed whenthe
> spring is unloading :)
>
> Just as a matter of hit and miss, we've found for a race car using2-3inches
> travel starting about 2 times the spring force in rebound resistance,then 1/3
> that for the rear compression and then 1/2 for the front compression. Soin US
books,
you
> should be in formula and constant heaven :)
>
>
> --Doug
>
>
Quoted text (10 lines)
> > > Ruud van Gaal wrote: > > > On Mon, 04 Sep 2000 23:19:59 GMT, Doug Arnao <[email hidden]> > > wrote: > > > > ... > > >For longitudinal traction? Thats called a slip ratio. Max traction is > > >about at 8% tire slip in general. It builds rather sharpley from 0-8%Quoted text (20 lines)
> > >tell you all you need to know about this including graphs. > > > > Thanks. That book is high on the list. I've ordered "Race Vehicle > > Dynamics" from the Milliken bunch and am anxiously awaiting it. > > I think it may have some similar graphs in it. > > Apart from that, I must say I'm puzzled a bit, because: > > slip_ratio=rotational_speed_of_wheel/actual_speed_of_wheel_vs_ground > > where rotational_speed_of_wheel is the speed at which the wheel is > > trying to move the vehicle (=ang.vel.around_axle*radius). > > > > I would say the *difference* of the 2 speeds is what you would use to > > describe the friction coefficient (CF). If the wheel is trying to push > > the vehicle forward at 200km/h, and the vehicle is actually moving > > 100km/h, the slip ratio is 200%. But there's 100km/h of tire surface > > moving across the road. When driving 10km/h, but the wheels are doing > > 20km/h, the slip ratio is also 200%, and this leads to the same CF? > > Seems intuitive to me. > > Might be totally wrong here, but I would think 100% slip ratio would be > correct. 100% above the traveling speed??? In other words the speed youQuoted text (9 lines)
> > >it is generating a CF of .75. A spinning tire is still making a good > > >amount of traction - sliding traction. > > > > Ok, good to know. I guess after the 27% the CF more or less (less) > > remains at .75. Well, a little below that. > > > > >> Next, I'm doing a suspension system. I'm already seeing a lot of > > >> parameters in this field, but I want to have a relatively simple > > >> representation currently. Mainly to see the car (the body rather)Quoted text (7 lines)
> > >> under power and steering. But I'm not an expert on suspensions. > > >> I use a plain vertical suspension with a spring and damper. The > > >> springs have a spring constant (k) so the force generated when > > >> compressing/stretching is k(x0-x) (a known formula). > > >> First hmm: when calculating the force of the spring on the body and > > >> the wheel, I now divide the force into force/2 on the body, and > > >> -force/2 on the wheels. Should I instead apply the total springQuoted text (23 lines)
> > >Do you have 4 seperate springs? > > > > Yes, I have 4 wheel objects, that actually have a seperate suspension > > system (spring+damper) for each. Todd states I have to exert the force > > fully both ways. > > To explain more simply, suppose I hold a spring in my hands and > > compress it using both hands. The force can be calulated using > > k(x0-x). Would the forces on BOTH my hands be that number or is the > > force divided over the two hands? > > Todd was correct, yes split between the two the total force. > > > I wondered about that because if you > > begin to think about energy (W=F*d) then the total energy if both > > hands start to move out (so the spring stretches again) would be twice > > the total energy available in the spring. But I'm not taking into > > account the fact probably that if one hand starts to move, the force > > on the other hand becomes less and the total sum of W=F*d remains > > equal to the energy stored in the spring. > > Hope this is clearer? :) > > > > Yes and your last sentance seems correct to me - yeah, what you saidQuoted text (15 lines)
> > >the same. > > > > Yes, my damper is stuck to the body. So the shaft velocity is the > > velocity at which the tire is moving to or away from the damper. > > > > ... > > >> What would an appropriate real value be? GPL uses 'bump==2' for > > >> example, which doesn't say a whole lot. Seems the damper unit is > > >> something like N/(m/s), so Newton per meter per second. > > > > > >The Americans use Lbs/(in/sec) typical shaft speed at a 1:1 suspension > > >could be as high as 30 inches per second > > > > > >Bilstein uses N/(m/s). Actually Nascar Heat uses the Bilstein valving > > >numbers, because most of Nascar use Bilstein shocks! A valving of 300Quoted text (13 lines)
> > >rambling there <g>. > > > > <g> no problem. > > I'm using the metric system at all places, because that's a whole lot > > less constants everywhere (and I'm European, so used to it). I > > downloaded a fantastic conversion program btw yesterday at > > http://www.joshmadison.com/software/ (Convert), which really has a > > zillion units to convert from and to. Used it to convert the air > > density from lb-mass/ft^3 to kg/m^3. It's so easy to get things buggy > > when you think in the wrong unit system. > > > > > To answer your question - somewhere between 1 and 3 > > >times the spring value should get you close. Try 1x on the compresionQuoted text (9 lines)
> > >2x on the extension. > > > > Ok. And real-life spring values? I see about 150Nm wheelrate values > > (which isn't exactly spring rate, but comes close) in GPL, but in > > F12000 I see spring coefficients of 140000 (140K), presumably Nm, as > > it seems to use the metric system. Or are F1 springs these days indeed > > so hard wrt 1967? > > We typically use wheel rates (the number out at the wheel) on our PorscheQuoted text (12 lines)
> determining the rate, of course > > F12000 numbers look a bit outa sight - I wouldn't try to make those fit > anywhere. They > > > > > Strange thing is, a rebound value (for F1-2000) is 3000 for example, > > with spring_coeff=140000. Not even close to the 1-3 times the spring > > value? Where do I go wrong here? :) > > No........ where did F12K go wrong <g>. > If a 100 lb spring travels 3 inches - it has 300lbs of stored energy thatQuoted text (20 lines)
> LBS units If you had 300 front spring and 500 rear spring you'd get > compression/rebound values of: > > front 125/450 > > rear 333/1000 > > > > > > > Ah, in the mean time, the book has arrived! Wow, technical stuff. > > Should do some reading tonight. I have DirectInput working btw. Spent > > yesterday evening exploring DirectInput. Quite nice, and now I have my > > MSFF wheel working (no force feedback yet ofcourse). It just has funny > > values for split axis throttle/braking (even though I set them to > > -1000..+1000, the Y/throttle axis stays on 1..-1000, yes, no typo). > > But I can get a much better sense of how my car is handling. It's far > > too sensitive now :) Must be the friction going too low. > > > > ... > > >Very interesting project Ruud...............buy the Carroll SmithQuoted text (7 lines)
> > >you'll have alot less questions :) > > > > Thanks, I now have Race Vehicle Dynamics from Milliken & Milliken, so > > that will keep me busy (for eternity probably, hehe), and keep me a > > little less questionhappy here on RAS. > > Millikens book is a tough read - for me at least. However, someone like
Ruud van Gaal wrote:
Your instincts are very good!
One common definition of Longitudinal Slip Ratio is
(rotation_rate_wheel_rad_per_sec * tire_radius_meters -
velocity_meters_per_sec) / velocity_meters_per_sec
This would be -1 for locked braking, and 0 for no slip,
and a high positive number for wheelspin.
See Race Car Vehicle Dynamics, Chap 14, Page 479, eq. 14.12 ;)
As you suggest, the longitudinal_force / normal_force
can be estimated as roughly linear with Longitudinal Slip Ratio
but only within the linear regime ;), (before the force peaks).
At higher slip, you lose force similarly to lateral force.
Have fun with Race Car Vehicle Dynamics Chapter 14 ;)
For the future, look carefully at how to vary the tire
charactersistics with normal force. This is important
because if you don't have sufficient correct variation
from this effect, normal setup techniques like
adjusting the anti-roll bars to modify the
distribution of lateral load transfer between the
front and rear tires won't cause proper
changes in the stability (oversteer/understeer)...
--
Matthew V. Jessick Motorsims
[email hidden] http://www.motorsims.com
Vehicle Dynamics Engineer (972)910-8866 Ext.125, Fax: (972)910-8216
Quoted text (14 lines)
> > On Mon, 04 Sep 2000 23:19:59 GMT, Doug Arnao <[email hidden]> > wrote: > > ... > >For longitudinal traction? Thats called a slip ratio. > Apart from that, I must say I'm puzzled a bit, because: > slip_ratio=rotational_speed_of_wheel/actual_speed_of_wheel_vs_ground > where rotational_speed_of_wheel is the speed at which the wheel is > trying to move the vehicle (=ang.vel.around_axle*radius). > > I would say the *difference* of the 2 speeds is what you would use to > describe the friction coefficient (CF).On 5 Sep 2000 18:51:21 GMT, "David"
<[email hidden]> wrote:
>Doug is right and I think a lot fo games miss a few things here. One is
>pneumatics trail that every tire has. Lower the aspect ratio Iwould
>venture the lower the pneumatic trail of the tire.
I'm reading RcVD (Race car vehicle dynamics) and I saw yesterday what
pneumatic trail is for the first time. Seems to have the answer to my
otherwise upcoming force-feedback questions (well, somewhere in that
area). ;)
Not to mention the gyroscopic precession that the tire generates when
rotating (see http://www.accs.net/users/cefpearson/gyro.htm)! That
will be done probably only when the feel isn't right.
But what aspect ratio are you talking about? width/length of contact
patch?
>also one thing I noticed only one game so far has gotten right and I will
>not name which one. In my Solo ll car when I get understeer and I steer
>more into the turn the more understeer I get, the less I turn the wheel the
>less the understeer is.
This would make perfect sense, but depending on slip angle. At first,
when running straight, you have no understeer. Then, when starting to
turn, understeer sets in. Understeer gets *relatively* less upto a
certain angle (the car gets more lateral acceleration), and then
you're over the top, so lateral acceleration actually becomes less,
and thus more understeer (because of reduced friction).
Well, I might be looking at it too simple.
In the end, you want to turn, so you'll have to compromise between
much-understeer-but-fair-lateral-acceleration and
less-understeer-but-also-crappy-lateral-acceleration.
Wasn't it somewhere in Cart where they say 6 degrees of slip angle is
the best? Which would probably mean 'has the largest lateral
acceleration'.
Sounds like too much understeer indeed. Ofcourse, having too much slip
angle will just make understeer worse, as you can see too in GPL, for
example, where sometimes the slip angle doesn't make a whole lot of
difference when slipping through a corner.
Ruud
>Dave
>
>Doug Arnao <[email hidden]> wrote in article
><[email hidden]>......
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (7 lines)
> I think Jackie Stewart explained it that the tire >tread face is not addressing the road properly, but I do think it relates a >bit to the above also. With sticky slicks it becomes very noticable so >much so once I actaully turned in had a bunch of understeer and reduced >steering lock and the car scooted thru the corner so quickly I was at the >exit with my foot nowhhere near the throttle not to mention even thinking >about full throttle.On Wed, 06 Sep 2000 01:43:39 GMT, mjessick-Motorsims
<[email hidden]> wrote:
...
>Your instincts are very good!
Thanks. :)
Ah, ofcourse. The difference in respect to the one or the other. Make
sense now. Thanks!
>See Race Car Vehicle Dynamics, Chap 14, Page 479, eq. 14.12 ;)
I'm still on page 22! ;)
>As you suggest, the longitudinal_force / normal_force
>can be estimated as roughly linear with Longitudinal Slip Ratio
>but only within the linear regime ;), (before the force peaks).
>At higher slip, you lose force similarly to lateral force.
>
>Have fun with Race Car Vehicle Dynamics Chapter 14 ;)
I will. I have to read on a bit though. :)
Tonight I will first do fullscreen on Win32, I think.
I will keep this in mind.
Thanks,
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (7 lines)
>One common definition of Longitudinal Slip Ratio is > >(rotation_rate_wheel_rad_per_sec * tire_radius_meters - > velocity_meters_per_sec) / velocity_meters_per_sec > >This would be -1 for locked braking, and 0 for no slip, >and a high positive number for wheelspin.Quoted text (8 lines)
>For the future, look carefully at how to vary the tire >charactersistics with normal force. This is important >because if you don't have sufficient correct variation >from this effect, normal setup techniques like >adjusting the anti-roll bars to modify the >distribution of lateral load transfer between the >front and rear tires won't cause proper >changes in the stability (oversteer/understeer)...On Tue, 05 Sep 2000 18:03:39 GMT, Doug Arnao <[email hidden]>
wrote:
...
>> If the wheel is trying to push
>> the vehicle forward at 200km/h, and the vehicle is actually moving
>> 100km/h, the slip ratio is 200%....
>Might be totally wrong here, but I would think 100% slip ratio would be
>correct. 100% above the traveling speed??? In other words the speed you are
>traveling is the starting point or ground "zero" <g>. make sense?
Yep, makes sense. I understand it now; it is just (v2-v1)/v1, while I
just did v2/v1. Still, intuitively, I would feel that for example in
the 200 vs. 100 km/h case, you have 100km of tire strip slipping along
every hour, while at 20 vs. 10km/h you only have 10 km of tire length
per hour, so in the 100km situation, I would expect more friction
(more forward force) over the hour than in the 10km case. But for both
cases, the long. slip ratio is 2 ( (200-100)/100 and (20-10)/10), so
the CF is the same if you look it up in a table.
Perhaps this will soon be explained to me in RCVD.
>As to weather your two examples have the same CF.........you would think so by
>the numbers. But rubber is a very wierd and amazing substance that in my
>experience dosen't act linearly.
I know. When first starting out with the car sim, I thought things
would be much easier. Now for a little force-feedback, I'll quickly
have to look at pneumatic trail and such to get a good feel perhaps.
Or just use the CF for now. But FF is funny, it seems you can
generates forces based on wheel position, velocity and acceleration.
So you can resist movement like a normal tire would when standing
still, but also based on how FAST you are moving the wheel, or even
*accelerating* the wheel. Cool stuff, but complicated.
>Just remeber one thing : All the high dollar fancy, trick, perfectley designed
>suspension systems eventually have to transmit their numbers through the tire
>contact patch. About 15 square inches of nylon and rubber..........This holds
>true for your sim ,too. I've always felt that the tire model makes or breaks a
>sim. I believe this is one of the things that Nascar Heat *nailed*.
Hm, after Nascar 1, 2 and 3 and some 1999 pack I gave up a bit on
Nascar titles, since they are so similar. It seems Nascar Heat might
deserve a try, although I'm ofcourse really anxious to see WSC (I'm
more into road tracks).
...
Uhm, you say 'split' here. Todd says *don't* split, but apply force to
both ends. I do that now, because in effect, that's what the state is
at time t. Makes more sense to me now. The total *energy* will be
divided over both ends automatically (by 2).
...
Funny, all the other values seem to be in the right unit system
though, well, as far as I can tell now. Would have thought they'd do
juggling with numbers internally to the program, so that the physics
files LOOK ok, but internally they can screw up (or never use some
vars). ;)
The shaft speed is the same as my wheel speed. I've got a damper
working now, and it works fine. Still have to insert some correct
numbers now. And start working on contact patch size, since the wheel
is bouncing oh so quietly so I have to insert a threshold for when the
tire actually touches the road.
But first I want to do a fullscreen, then get back to the theory.
Hopefully Voodoo2's will then be supported.
You mean the springs have only 2-3 inches of travel available? Seems
very short. Or perhaps that's customary for race/Porsche cars.
...
>Millikens book is a tough read - for me at least. However, someone like you
>should be in formula and constant heaven :)
It starts out relatively easy. Or well, with stuff I have heard about
before. But it looks to be a book where things quickly turn into a lot
of theoretical info. Which is good in a way, because I have little use
for statements 'rubber gets less tractive power when getting warmer'.
;)
Thanks for all the numbers and info,
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (9 lines)
>> Yes, I have 4 wheel objects, that actually have a seperate suspension >> system (spring+damper) for each. Todd states I have to exert the force >> fully both ways. >> To explain more simply, suppose I hold a spring in my hands and >> compress it using both hands. The force can be calulated using >> k(x0-x). Would the forces on BOTH my hands be that number or is the >> force divided over the two hands? > >Todd was correct, yes split between the two the total force.Quoted text (11 lines)
>We typically use wheel rates (the number out at the wheel) on our Porsche race >cars from 200-600 lbs approx. - this after going out thru the motion ratio (the >lever arms) of the suspension/shock. > >Last I paid attention to F1 spring rates, they were using springs as high as >5000 lbs and as low as 800. Again, the wheel rates could be 1/4 to 1/2 of that >amount depending on the motion ratio's used. Downforce plays a huge role in >determining the rate, of course > >F12000 numbers look a bit outa sight - I wouldn't try to make those fit >anywhere. TheyQuoted text (9 lines)
>> Strange thing is, a rebound value (for F1-2000) is 3000 for example, >> with spring_coeff=140000. Not even close to the 1-3 times the spring >> value? Where do I go wrong here? :) > >No........ where did F12K go wrong <g>. >If a 100 lb spring travels 3 inches - it has 300lbs of stored energy that needs >to be dissapated on the way down (rebound) - or absorbed by the damper. That >is in round numbers of course and assuming you know the shaft speed when the >spring is unloading :)Quoted text (9 lines)
>Just as a matter of hit and miss, we've found for a race car using 2-3inches >travel starting about 2 times the spring force in rebound resistance, then 1/3 >that for the rear compression and then 1/2 for the front compression. So in US >LBS units If you had 300 front spring and 500 rear spring you'd get >compression/rebound values of: > >front 125/450 > >rear 333/1000Ruud van Gaal wrote:
You were not far off, you should have only subtracted 1 ;).
The dependence of tire forces is dependent on slip ratios/angles rather
than direct velocities because the friction forces are largely
independent of the velocity. That way only the geometry of the
gripping/sliding of the steering patch matters and not the velocity at
which it happens.
The power losses, however do increase with the velocity, so the tires do
get heated more when operating at the same slip angle/ration but with a
higher overall speed. The forces involved are largely similar at the
same loads, however.
Always keep in mind, though, that these are all models, and that reality
may not always behave exactly like that. There will always be the
dependence of the tire forces on the speed, it's just that these can be
treated as corrections to an otherwise well functioning slip angle/ratio
dependence.
-Gregor
Quoted text (11 lines)
> > Yep, makes sense. I understand it now; it is just (v2-v1)/v1, while I > just did v2/v1. Still, intuitively, I would feel that for example in > the 200 vs. 100 km/h case, you have 100km of tire strip slipping along > every hour, while at 20 vs. 10km/h you only have 10 km of tire length > per hour, so in the 100km situation, I would expect more friction > (more forward force) over the hour than in the 10km case. But for both > cases, the long. slip ratio is 2 ( (200-100)/100 and (20-10)/10), so > the CF is the same if you look it up in a table. > Perhaps this will soon be explained to me in RCVD. >Quoted text (16 lines)
>Doug is right and I think a lot fo games miss a few things here. One is >pneumatics trail that every tire has. Lower the aspect ratio Iwould >venture the lower the pneumatic trail of the tire. > >also one thing I noticed only one game so far has gotten right and I will >not name which one. In my Solo ll car when I get understeer and I steer >more into the turn the more understeer I get, the less I turn the wheel the >less the understeer is. I think Jackie Stewart explained it that the tire >tread face is not addressing the road properly, but I do think it relates a >bit to the above also. With sticky slicks it becomes very noticable so >much so once I actaully turned in had a bunch of understeer and reduced >steering lock and the car scooted thru the corner so quickly I was at the >exit with my foot nowhhere near the throttle not to mention even thinking >about full throttle. > >DaveQuoted text (7 lines)
>>One common definition of Longitudinal Slip Ratio is >> >>(rotation_rate_wheel_rad_per_sec * tire_radius_meters - >> velocity_meters_per_sec) / velocity_meters_per_sec >> >>This would be -1 for locked braking, and 0 for no slip, >>and a high positive number for wheelspin.On 07 Sep 2000 06:37:18 GMT, [email hidden] (JTW62074) wrote:
How about that! ;)
I must say now that I'm reading RCVD, I will probably not base my code
so much on the slipvector, but more on slip angles and ratios.
The suspension is giving me some pain now, because for the first time
my body and wheels are listening to gravity all by themselves, and my
wheels quickly disappear into the ground once my body starts
accelerating down to the ground (and I'm not even dancing! ;).
So I'll have to get that right for non-banked, non-elevated surfaces
first, and then perhaps do another intermediate release.
Force feedback seems within reach, now I've learned a bit about
pneumatic trail. Just have to figure out how to combine those general
forces (a 'friction' force in DirectInput terms) and steering torque
into 1 force (in force feedback terms).
And a quick revlimiter. 500,000 rpm is not really within everyday
limits ;)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (20 lines)
>>>One common definition of Longitudinal Slip Ratio is >>> >>>(rotation_rate_wheel_rad_per_sec * tire_radius_meters - >>> velocity_meters_per_sec) / velocity_meters_per_sec >>> >>>This would be -1 for locked braking, and 0 for no slip, >>>and a high positive number for wheelspin. > >Ruud, I have given you incorrect/difficult to work with information and >apologize! I've been using something similar to: > >slip_ratio = true_rot_vel / free_rolling_rot_vel - 1 > >--rather than the equation Matt Jessick just posted using a difference in the >ratio. This led to a wild goose chase and a few IF blocks in getting slip >ratios to compute correctly and work properly in strange situations (wheels >spinning backwards while car is moving forwards, etc..) > > I'll switch to the "difference and ratio" equation above as well. Learn >something new everyday. Thanks to both of you on this subject.On Wed, 06 Sep 2000 16:38:08 +0200, Gregor Veble
<[email hidden]> wrote:
Damn, the off-by-1 bug keeps haunting me until... forever! ;)
Hm, strange, but true. Car turning forces were more complicated than I
figured. Duh.
>Always keep in mind, though, that these are all models, and that reality
>may not always behave exactly like that. There will always be the
>dependence of the tire forces on the speed, it's just that these can be
>treated as corrections to an otherwise well functioning slip angle/ratio
>dependence.
Yes, I've thought about that. Creating a curve editor of some sorts
and then having a few extra parameters to affect the X and Y axes to
adjust for things like extra loads for example. Sort of coefficients
to modify where the peaks are located.
Oh, I see btw that pound-forces are used in the pound-inch-second
system, and 1 pound-force (lbf) = 4.448222 Newton. I thought 1 pound
was about .45 kg. Oh ofcourse, .45kg = .45*9,81 = 4,4N. Damn. It's not
that difficult, Ruud. ;)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (14 lines)
>Ruud van Gaal wrote: >> >> Yep, makes sense. I understand it now; it is just (v2-v1)/v1, while I >> just did v2/v1. Still, intuitively, I would feel that for example in >> the 200 vs. 100 km/h case, you have 100km of tire strip slipping along >> every hour, while at 20 vs. 10km/h you only have 10 km of tire length >> per hour, so in the 100km situation, I would expect more friction >> (more forward force) over the hour than in the 10km case. But for both >> cases, the long. slip ratio is 2 ( (200-100)/100 and (20-10)/10), so >> the CF is the same if you look it up in a table. >> Perhaps this will soon be explained to me in RCVD. >> > >You were not far off, you should have only subtracted 1 ;).Quoted text (10 lines)
>The dependence of tire forces is dependent on slip ratios/angles rather >than direct velocities because the friction forces are largely >independent of the velocity. That way only the geometry of the >gripping/sliding of the steering patch matters and not the velocity at >which it happens. > >The power losses, however do increase with the velocity, so the tires do >get heated more when operating at the same slip angle/ration but with a >higher overall speed. The forces involved are largely similar at the >same loads, however.>How about that! ;)
>I must say now that I'm reading RCVD, I will probably not base my code
>so much on the slipvector, but more on slip angles and ratios.
Yes, this is how I'm doing it too. See why? :-) There's real tire data
available (in rare places) that use these inputs, and nondimensional tire
theory is based on this as well. I just completed the third dimension. All I
can say is it's neat to see the car roll up and jump off sine curve generated
hills (no polygons, but then again, my hills are invisible ;-)). Still have a
couple details to mess with there, such as calculating the slip ratio along the
actual plane the tire is moving instead of along the ground plane (when a hill
steepens past a point, it generates slip ratios closer and closer to zero, then
changes sign, causing the car to accelerate up the hill.. The faster it hits
the hill, the more rapidly it accelerates up it... Oops!)
There's another little problem here that I knew was coming and now must deal
with it, so figured I'd warn you ahead of time so you can fix it before me :-).
It's called the "stopping on a hill" bug.
If the car is stopped on a hill (pointed straight towards the top or bottom,
no special angle), the slip ratio model doesn't generate any force to keep the
car in place. Oops again. Both velocities are 0 (free rolling and true
rolling) hence, 0 slip ratio and 0 force. The car slides very gently down the
hill even with the tires locked. Ponder that one for a while and let me know
what you think :-)
To: Matt Jessick, Doug Arnao, or anybody who'd know:
Given these problems, I'm considering doing a full blown 3-D numerical
analysis of a tire to generate slip angle and ratio curves at different
pressures, loads, and possibly temperatures. Have vehicle dynamics engineers
had much success with this type of project, or would it be a waste of time (a
LOT of time, probably)?
Thanks,
Todd Wasson
Well, we shipped Viper Racing with that same hill-sliding bug--I didn't fix
it until after we started working on our current project (NASCAR Heat), and
got a little smarter about tire modeling. (Ended up fixing by modeling the
shear-spring effect of the sidewalls) But just using slip ratios, slip
angles, and camber angles will get you pretty far for starters--the
magnitude of pneumatic trail is more of a factor in steering wheel rim force
than in the car's yawing moment.
-Dave P.
MGI
"JTW62074" <[email hidden]> wrote in message
news:[email hidden]...
<snip>
> If the car is stopped on a hill (pointed straight towards the top orbottom,
> no special angle), the slip ratio model doesn't generate any force to keepthe
> car in place. Oops again. Both velocities are 0 (free rolling and true
> rolling) hence, 0 slip ratio and 0 force. The car slides very gently downthe
> hill even with the tires locked. Ponder that one for a while and let meknow
> what you think :-)
>
Ruud van Gaal wrote:
> Oh, I see btw that pound-forces are used in the pound-inch-second
> system, and 1 pound-force (lbf) = 4.448222 Newton. I thought 1 pound
> was about .45 kg. Oh ofcourse, .45kg = .45*9,81 = 4,4N. Damn. It's not
> that difficult, Ruud. ;)
I would call 0.45 kg a pound mass (lbm).
Be careful with them. The easiest English mass unit to use
like you use kg in SI units is the "slug". (32.174 lbm or so).
1 slug acted on by 1 pound force is accelerated at 1 ft/s/s.
Keep to SI and stay sane ;)
- Matt
JTW62074 wrote:
> There's another little problem here that I knew was coming and now must deal
> with it, so figured I'd warn you ahead of time so you can fix it before me :-).
> It's called the "stopping on a hill" bug.
I'm always amused by this one, because in all the various simulators
I used in aerospace, only one of them stopped. (And we mostly
ignored spurious brake lockups because we knew it wasn't working right.)
We were mostly interested in hitting something or going somewhere fast,
and not concerned with stopping when we got there. This seems to
be a problem unique to manned simulators specializing in operations,
as opposed to perfomance analysis and control design tools I
mostly worked with.
> If the car is stopped on a hill (pointed straight towards the top or bottom,
> no special angle), the slip ratio model doesn't generate any force to keep the
> car in place.
> The car slides very gently down the hill even with the tires locked
Yes, that's annoying. ;) You need to include static friction in some way.
We change the equations of motion to take into account each tire being
constrained or not, but that is rather complex with one to four tires stopped
and the various redundancies. You will need a specialized model for slow
speed anyway, so try and get the slow speed and stopped modeling to
work together.
> Given these problems, I'm considering doing a full blown 3-D numerical
> analysis of a tire to generate slip angle and ratio curves at different
> pressures, loads, and possibly temperatures. Have vehicle dynamics engineers
> had much success with this type of project, or would it be a waste of time (a
> LOT of time, probably)?
You'd think that top racing teams could get great masses of data
from tire manufacturers about their fancy racing tires. No such
luck apparently (based on the comments of various consultants.)
Without good data you won't know how correct you are anyway.
So you will have to do things as the teams do, and use "telemetry"
to tune your models until they seem as correct as possible.
(If you know the trajectory, you can guess what the tires must be
doing. Sort of. If you have to guess at both.... ;)
It would indeed take some time. You might want to make nerdy friends fast. ;)
Do you know any graduate students? (They are used to working for
low pay and eating inexpensive food ;);)
- Matt
On 07 Sep 2000 21:06:53 GMT, [email hidden] (JTW62074) wrote:
>>Yes, I've thought about that. Creating a curve editor of some sorts
>>and then having a few extra parameters to affect the X and Y axes to
>>adjust for things like extra loads for example. Sort of coefficients
>>to modify where the peaks are located.
>
> Ruud, look into the Pacejka Magic Formula.
With a name like that the button 'I'm feeling lucky!' in Google might
actually work for the first time! ;)
> I believe chapter 14 in Race Car
>Vehicle Dynamics has a little blurb in it (relating it to non-dimensional tire
>theory, something I still haven't got to match real tire data acceptably.)
>Aside from that, search the net for Pacejka's work. He's figured out a snazzy
>formula that lets you modify 4 coefficients to generate a realistic curve.
Sounds good. I'll look into it.
First now a weekend of good social sailing. ;)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
On 07 Sep 2000 21:27:23 GMT, [email hidden] (JTW62074) wrote:
I see why, lol. In fact, it's more than that. I now implements forces
in every which way, without regarding lat/long directions. Just
assuming 'well, friction behaves the same in every way'. Not true,
because the tires are 3-dimensional springs of sorts.
I know the feeling with my slipvector. I'll keep it, but more for a
statistical reason, or for maintaining lateral and longitudinal
velocity in 1 structure.
I think I thought about that one. :)
No really, that's why I came up with the slipvector. Well, not coming
from the direction you're coming from, but I see resemblances. The
thing that bogged me was this:
- Throw the car forward with V=10, the wheels not turning. No throttle
or brakes. Now the car touches the ground. The wheels should start
rotating in real life.
I use the slipvector for this problem. The wheels are doing V=0 road
speed, the car is doing V=10. Slipvector is V=-10 (all along the
forward axis, which I call Z).
This leads to friction force along the Z axis. This friction force
generates torque to the wheel. The wheel starts turning.
I've had success with this, and you soon get an equilibrium where the
wheels are turning just as fast as the car is moving.
Now for brakes; these generate a torque from 0..-maxTorque, depending
on the torque actually produced at the wheel (so for a still-standing
car it won't start rolling when you press the brakes, just like
friction never exceeds the original force).
So when sliding forward, the slipvector creates torque for the wheel,
which decelerates the car (well, the friction force does).
Now suppose you're on the hill, hm, picturing. The gravity force gives
a force in the longitudinal direction (to the downside of the hill).
This force generates torque around the wheel, which then starts
rotating so your car goes down the hill. But your brakes are still on,
the torque generated of the gravity force is nullified (upto
maxBrakingTorque;
totalTorque+=max(maxBrakingTorque,currentOtherTorque)). So no torque,
no moving of the wheel, no velocity.
Hmm, where did the slipvector come into this? ;) More coffee please.
No, the slipvector is bugging me. Because of slipratio.
The reason I came up with it was more when wheels are rotating in the
air, and then you let the car down on the ground. In which case the
slipvector is generating a forward force. Problem is in how I support
throttle, which just lets the wheel rotate with friction, then you get
a slipvector, and from that a force is generated which actually pushes
the car forward. This doesn't work correctly, because the wheel is
always turning faster than the actual car velocity. But when
accelerating the car, somewhere the slip vector remains greater always
than the car velocity. So the car starts accelerating by itself if
only touched by a mild throttle.
Don't try to follow this, its bad, and I had put in rolling resistance
so that it stays stable at a certain point. I'm going to revise the
lot though, so I'm not putting too much thinking in it. It looked
great when throwing the car forward, but it was bugging me at low
speeds where the slipvector's accelerating bug was bigger than the
car's rolling resistance.
I'm just rattling about. I'm going to revise the stuff so that I do
lateral and longitudinal forces in seperate ways. And then figure out
the problems when they come to me.
Seems like you would have to approximate this sort of thing to do
anything like GPL, where tire temps are very important, for instance.
Ah well, more time late next week. First a little more reading, and
more, and thinking. Back to the drawing board. (well, the slipvector
looks nice and feels nice with my MSFF wheel, but is has side-effects
because of a bad way of thinking in too many places)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (7 lines)
>>How about that! ;) >>I must say now that I'm reading RCVD, I will probably not base my code >>so much on the slipvector, but more on slip angles and ratios. > > Yes, this is how I'm doing it too. See why? :-) There's real tire data >available (in rare places) that use these inputs, and nondimensional tire >theory is based on this as well.Quoted text (8 lines)
> I just completed the third dimension. All I >can say is it's neat to see the car roll up and jump off sine curve generated >hills (no polygons, but then again, my hills are invisible ;-)). Still have a >couple details to mess with there, such as calculating the slip ratio along the >actual plane the tire is moving instead of along the ground plane (when a hill >steepens past a point, it generates slip ratios closer and closer to zero, then >changes sign, causing the car to accelerate up the hill.. The faster it hits >the hill, the more rapidly it accelerates up it... Oops!)Quoted text (10 lines)
> There's another little problem here that I knew was coming and now must deal >with it, so figured I'd warn you ahead of time so you can fix it before me :-). > It's called the "stopping on a hill" bug. > > If the car is stopped on a hill (pointed straight towards the top or bottom, >no special angle), the slip ratio model doesn't generate any force to keep the >car in place. Oops again. Both velocities are 0 (free rolling and true >rolling) hence, 0 slip ratio and 0 force. The car slides very gently down the >hill even with the tires locked. Ponder that one for a while and let me know >what you think :-)Quoted text (7 lines)
> To: Matt Jessick, Doug Arnao, or anybody who'd know: > > Given these problems, I'm considering doing a full blown 3-D numerical >analysis of a tire to generate slip angle and ratio curves at different >pressures, loads, and possibly temperatures. Have vehicle dynamics engineers >had much success with this type of project, or would it be a waste of time (a >LOT of time, probably)?Quoted text (10 lines)
>So when sliding forward, the slipvector creates torque for the wheel, >which decelerates the car (well, the friction force does). >Now suppose you're on the hill, hm, picturing. The gravity force gives >a force in the longitudinal direction (to the downside of the hill). >This force generates torque around the wheel, which then starts >rotating so your car goes down the hill. But your brakes are still on, >the torque generated of the gravity force is nullified (upto >maxBrakingTorque; >totalTorque+=max(maxBrakingTorque,currentOtherTorque)). So no torque, >no moving of the wheel, no velocity.Quoted text (7 lines)
>Problem is in how I support >throttle, which just lets the wheel rotate with friction, then you get >a slipvector, and from that a force is generated which actually pushes >the car forward. This doesn't work correctly, because the wheel is >always turning faster than the actual car velocity. But when >accelerating the car, somewhere the slip vector remains greater always >than the car velocity.>> It's called the "stopping on a hill" bug.
>
>I'm always amused by this one, because in all the various simulators
>I used in aerospace, only one of them stopped. (And we mostly
>ignored spurious brake lockups because we knew it wasn't working right.)
That's funny!
>Yes, that's annoying. ;) You need to include static friction in some way.
>We change the equations of motion to take into account each tire being
>constrained or not, but that is rather complex with one to four tires stopped
>and the various redundancies.
Constrained? I don't understand what that means.
>You will need a specialized model for slow
>speed anyway, so try and get the slow speed and stopped modeling to
>work together.
I've heard of someone else doing this as well. Seems the individual switched
from a slip angle/ratio based model to a tire deformation model at low speed,
then completely abandoned the slip angle/ratio approach and all hill stopping
problems vanished. The only problem with this approach that I can think of is
it may be difficult to get real tire data to switch formats like that.
Yes, the slip angle/ratio approach does not work well at low speed. I get
instabilities and slightly jumpy motion at speeds below (I think) about 4mph
when running at 166 Hz. Running low weight tires, car, anything at all really,
makes the problem worse.
Yes, that is a problem. Telemetry and actual race cars are beyond anything I
can get access to :-) Perhaps a numerical analysis model could be tuned to
match real data for one tire, then nondimensional technique could be used to
predict slip angle/lateral force at some variation of load, then the same
change could be run through the numerical analysis model for comparison.
Either way, if I go for the tire simulation, I could sure use everybody's input
:-)
>It would indeed take some time. You might want to make nerdy friends fast. ;)
>Do you know any graduate students? (They are used to working for
>low pay and eating inexpensive food ;);)
>
>- Matt
Hmm... Not happening! I'll have to become the ultimate nerd myself for
awhile to make that work, but then again, I'm half way there already :-)
Todd Wasson
Quoted text (13 lines)
>> Given these problems, I'm considering doing a full blown 3-D numerical >> analysis of a tire to generate slip angle and ratio curves at different >> pressures, loads, and possibly temperatures. Have vehicle dynamics >engineers >> had much success with this type of project, or would it be a waste of time >(a >> LOT of time, probably)? >Without good data you won't know how correct you are anyway. >So you will have to do things as the teams do, and use "telemetry" >to tune your models until they seem as correct as possible. >(If you know the trajectory, you can guess what the tires must be >doing. Sort of. If you have to guess at both.... ;)>Well, we shipped Viper Racing with that same hill-sliding bug--I didn't fix
>it until after we started working on our current project (NASCAR Heat), and
>got a little smarter about tire modeling. (Ended up fixing by modeling the
>shear-spring effect of the sidewalls)
Were you doing the shear spring effect by treating the tire as a sort of
torsion spring at low speeds? For example, sitting on a hill, perhaps the car
could roll backwards far enough to rotate a theoretical free rolling tire 5
degrees (or maybe much further for a drag slick, simulating wrinkling to a good
extent).
This would conflict with a slip ratio model (in my mind) as the rotational
angle difference between the theoretical free rolling tire and the actual tire
would keep increasing whenever a torque was applied at the hub (the rotational
speeds are different, so the angle difference would keep increasing at any slip
ratio except for 0). Perhaps this angle "offset" could be reset anytime the
tire rotates far enough to move a piece of rubber through the contact patch.
But then again, this would created tiny bursts of torque every few degrees.
Hmm...
>the
>magnitude of pneumatic trail is more of a factor in steering wheel rim force
>than in the car's yawing moment.
>-Dave P.
> MGI
Yes, I had a problem when I started working on this model where the car would
not straighten out if the gas or brakes were applied in even the minutest
amount. Someone suggested that my neglecting the effects of tire aligning
torque might be the problem. After implementing the effects on the CG of the
car, the change in handling was nearly undetectable. (it's still in the code,
but remarked out as it's fairly expensive for no noticable change). I learned
during that process that the tire aligning torques at all four tires more or
less cancel each other out when examining their effects on the CG of the car.
As you said, it's great for force feedback, but it doesn't do much to the
handling of the car.
The "won't straighten out" problem turned out to be caused by the way I was
limiting sideforce at the tire when torque was applied at the wheel hub.
Todd Wasson
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com
(oh here I am giving away hard-won techniques)
I figured that the empirical tire modeling formulas that I was using were
incorporating the effect of building up of tire shear, and then parts of the
tire "springing back" as the bits that were making up the patch moved off
the ground (and of course transitioning to a sliding state at the tail end
(and head, I suppose) of the contact patch--as you said, this is what a slip
angle/ratio based system is modeling. At non-zero tire rotation rates, the
"shear" is constantly being relieved, making act more like a damper than a
spring. So the trick was to continuously dissipate the stored "shear"
energy as a function of rotation, and blend between the two types of model
as the tire rotation went from zero to a small epsilon rate, and then use
just slip-based modeling above that speed. It worked out pretty well, as
you get nice effects like the "kickback" when you screech to a stop and then
they toss the car a little bit backwards, a phenomenon we all experienced
when learning how to drive...
I think that's what's going on when you drive a normal car to a stoplight,
and then pull off the brake right as you come to a stop; this lets the tire
"unwind" itself in a smooth and controlled fashion.
Since this was all for road/oval racing with no standing starts (except
for pitstops) this seemed like plenty good, especially given current CPU
speeds--I'm sure for doing high-end drag racing simulations, you'd need
fancier deformation modeling with all that crazy tire crinkling and stuff
I've seen in the shots of dragsters shooting out of the hole.
Anyways, it's nice to hear that other people care about this kind of
stuff...
-Dave P.
MGI
"JTW62074" <[email hidden]> wrote in message
news:[email hidden]...
> >Well, we shipped Viper Racing with that same hill-sliding bug--I didn'tfix
> >it until after we started working on our current project (NASCAR Heat),and
> >got a little smarter about tire modeling. (Ended up fixing by modelingthe
> >shear-spring effect of the sidewalls)
>
>
> Were you doing the shear spring effect by treating the tire as a sort of
> torsion spring at low speeds? For example, sitting on a hill, perhaps thecar
> could roll backwards far enough to rotate a theoretical free rolling tire5
> degrees (or maybe much further for a drag slick, simulating wrinkling to agood
> extent).
>
> This would conflict with a slip ratio model (in my mind) as therotational
> angle difference between the theoretical free rolling tire and the actualtire
> would keep increasing whenever a torque was applied at the hub (therotational
> speeds are different, so the angle difference would keep increasing at anyslip
> ratio except for 0). Perhaps this angle "offset" could be reset anytimethe
> tire rotates far enough to move a piece of rubber through the contactpatch.
> But then again, this would created tiny bursts of torque every fewdegrees.
> Hmm...
>
JTW62074 wrote:
> >We change the equations of motion to take into account each tire being
> >constrained or not, but that is rather complex with one to four tires stopped
> >and the various redundancies.
>
> Constrained? I don't understand what that means.
If one tire happens to be stopped, and you deicde to use that
fact, you can reformulate the equations of motion taking that into
account. Add equations that force the acceleration at that tire
contact point to be zero, along with the other usual equations.
The added constraint equation will change the results you get.
For example, if you pin one tire fixed on a hillside, you know that
all the others are forced to move only on circular paths around the
fixed tire...
> Yes, that is a problem. Telemetry and actual race cars are beyond anything I
> can get access to :-)
I was thinking lap times that are fairly available on the web, and give a good
gross check. Pickdifferent kinds of courses. Road America and Laguna Seca
will stress the errors in different parts of the model.
On 09 Sep 2000 23:37:16 GMT, [email hidden] (JTW62074) wrote:
...
> It works a little differently in mine. When sitting on a hill at a dead
>stop, gravity pulls straight downwards (I don't split it into seperate
>vectors). This compresses the springs. Since the springs are at an angle to
>the hill, they accelerate the car along the ground plane (x/z in your system).
>Since the elevation at each wheel changes after each time step movement, the
>springs extend a little, and the car drops.
Hopefully your springs will stabilize quickly then at a point where
the dropping of the car the rebounding of your springs match up.
Because in fact, the car is not as heavy on the hill to the springs
(the gravity force component of the car perpendicular to the hill
surface is only part of that of the gravity force pulling at the car's
CG). So for the springs, it looks as if the car is not as heavy
anymore and start to decompress. Movement occurs because of the
gravity force component in the direction of the slope ultimately.
...
> By hitting the brake pedal, the tires can be locked, but my model doesn't
>care how the car is moving in relation to the tires' rotation.
Meaning you have no spinning wheels? (say the wheels are spinning at
twice the velocity that the car is actually moving)
...
It does make sense yes, but I'm using it in conflicting ways. When
applying a torque to a wheel. That makes it rotate. The next
integration step a slipvector is noticed, which creates a forward
force on the car, and the car starts accelerating (linearly).
However, on the other hand I had the problem of spinning wheels on a
car floating in air, then dropping the car onto the ground. The car
should move forward and the wheels stop decelerate their rotation
(quite quickly).
Combining the two is what I do wrong at the moment; in the latter
case, the wheels turn faster than the car does, so I apply a force to
the car (translation). However, somewhere the wheel is also angularly
accelerated, and the two result in the wheel always turning faster
than the car velocity. Until roll resistance or something like that
comes in.
The problem I had basically was that I wanted to know, for a given
torque at the wheel axle, how fast the *wheel* itself is going to
rotate. This is dependent on the body mass; a heavy car will
accelerate its wheels more slowly than a light car. But what is the
ratio of wheel rotation and car acceleration? If you apply the torque
as a force at the outer radisu of the wheel (to calc its rotation),
the wheel would accelerate (angularly) much faster if it is on the
ground. Problem is that I want to have spinning wheels, so just taking
the car acceleration and calculating needed wheel velocity out of that
isn't enough.
There are certain principles I have to review carefully. That's ok,
its my first sim and things are still relatively small.
...
> Check out GPL. Drive along in 3rd or 4th gear where there won't be any
>wheelspin under full throttle (use chase cam to really see the rpm's.) Drive
>at a constant speed (a little throttle). Floor the gas and watch the rpm's
>jump fairly dramatically even though the car hasn't sped up more than 3-4 mph
>yet. Perhaps they will spin up by 10% or so beyond what we'd calculate for
>that speed? It looks like they do to me.
Hmm, interesting. But I haven't seen yet where I would get a clue of
how spinning percentage relates to the torque.
...
> Bottom line, I think your slip vector and my slip ratio are fundamentally
>the same thing, only expressed in different units so slip angle and ratio can
>be combined in your model. Your thinking may not be as far off as you suspect.
I hope so. :)
It does fine indeed on just skidding around, but it has side effects
which I want to pull out by taking an approach that fits more into the
RCVD line of things. To keep from going mad. ;)
The simpler things stay (in relation to known material), the more
improvements I can make.
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (11 lines)
>>Problem is in how I support >>throttle, which just lets the wheel rotate with friction, then you get >>a slipvector, and from that a force is generated which actually pushes >>the car forward. This doesn't work correctly, because the wheel is >>always turning faster than the actual car velocity. But when >>accelerating the car, somewhere the slip vector remains greater always >>than the car velocity. > > According to Race Car Vehicle Dynamics, this is correct! It seems counter >intuitive to me as well, but after reading about the longitudinal workings of >the tire over and over and thinking about it a lot, it started to make sense.Quoted text (17 lines)
>> It works a little differently in mine. When sitting on a hill at a dead >>stop, gravity pulls straight downwards (I don't split it into seperate >>vectors). This compresses the springs. Since the springs are at an angle >to >>the hill, they accelerate the car along the ground plane (x/z in your >system). >>Since the elevation at each wheel changes after each time step movement, the >>springs extend a little, and the car drops. > >Hopefully your springs will stabilize quickly then at a point where >the dropping of the car the rebounding of your springs match up. >Because in fact, the car is not as heavy on the hill to the springs >(the gravity force component of the car perpendicular to the hill >surface is only part of that of the gravity force pulling at the car's >CG). So for the springs, it looks as if the car is not as heavy >anymore and start to decompress. Movement occurs because of the >gravity force component in the direction of the slope ultimately.Quoted text (29 lines)
>>>Problem is in how I support >>>throttle, which just lets the wheel rotate with friction, then you get >>>a slipvector, and from that a force is generated which actually pushes >>>the car forward. This doesn't work correctly, because the wheel is >>>always turning faster than the actual car velocity. But when >>>accelerating the car, somewhere the slip vector remains greater always >>>than the car velocity. >> >> According to Race Car Vehicle Dynamics, this is correct! It seems counter >>intuitive to me as well, but after reading about the longitudinal workings >of >>the tire over and over and thinking about it a lot, it started to make >sense. > >It does make sense yes, but I'm using it in conflicting ways. When >applying a torque to a wheel. That makes it rotate. The next >integration step a slipvector is noticed, which creates a forward >force on the car, and the car starts accelerating (linearly). >However, on the other hand I had the problem of spinning wheels on a >car floating in air, then dropping the car onto the ground. The car >should move forward and the wheels stop decelerate their rotation >(quite quickly). >Combining the two is what I do wrong at the moment; in the latter >case, the wheels turn faster than the car does, so I apply a force to >the car (translation). However, somewhere the wheel is also angularly >accelerated, and the two result in the wheel always turning faster >than the car velocity. Until roll resistance or something like that >comes in. >Quoted text (11 lines)
>> Check out GPL. Drive along in 3rd or 4th gear where there won't be any >>wheelspin under full throttle (use chase cam to really see the rpm's.) >Drive >>at a constant speed (a little throttle). Floor the gas and watch the rpm's >>jump fairly dramatically even though the car hasn't sped up more than 3-4 >mph >>yet. Perhaps they will spin up by 10% or so beyond what we'd calculate for >>that speed? It looks like they do to me. > >Hmm, interesting. But I haven't seen yet where I would get a clue of >how spinning percentage relates to the torque.On 18 Sep 2000 22:55:38 GMT, [email hidden] (J. Todd Wasson) wrote:
...
> Sorry it took so long to get back here ;-)
Well what do you know, lol, I asked a question today here about the
exact thing you're explaining in this post in great detail. :)
Thanks so much for elaborating the engine/road torque process, I think
I've got it now! ;-)
> You're right about this. I
>didn't give it a whole lot of thought since the springs are nearly
>perpendicular to the ground anyway, and make a force in the same direction the
>ground will because of this. I just wanted to get 3-dimensional and see what
>would happen :-)
If you decide into vectors for 3D by the way, be sure to also check
out matrices, since they are very useful into doing calculations like
transforming tire coordinates to world coordinates in just a single
matrix multiplication step.
> However, my blurb about the springs isn't really right now that you mention
>it. If it was, all you'd have to do is mount the springs at an angle and the
>car could accelerate by itself on a flat road... Hmmm... ;-) That's like me
>saying drag racers jack up their cars because they go faster when they're
>pointing downhill!! LOL!!
Lol, so that's why! ;-) Would be difficult to reverse park such cars
in a busy street. :)
...
Weird indeed, but a bit understandable looking at the slip vs. grip
pictures in RCVD. I get a better understanding every day now, and the
slipvector was indeed doing a simulation of some sorts of slip ratio,
how funny.
However, the slipvector was also accelerating the wheel, at a ratio
that was determined by sprung mass : wheel acceleration : wheel
angular acceleration. Some bogus theory I wrote about in my README.txt
from the source zip at my site. Although well, it uses energy, not
forces, to come up with something.
The ratio was too big though, I think, so the car kept on speeding up,
even though it should eventually get to a point where engine torque
equals reverse road torque. This was not the case, so it was a
perpetuous acceleration. Thankfully I understand now why it didn't
work. Besides, lateral and longitudinal forces are generated in
different ways. Even long. forces are generated differently for
negative and positive slip ratios. So the slipvector is not good
enough for me. (although it is used in RARS, where the force is
directly a scaled slipvector, for every direction)
> In fact, take a look
>at your front wheels under hard acceleration. Do they actually turn *slower*
>than you'd think? They should be developing a negative slip ratio.
In fact my front wheels rotate faster, lol. Sorry, I have a front
wheel drive car. ;-)
Never noticed this though. Never paid enough attention to it.
Yes, I was getting crazy this weekend, when I was not getting any
further, thinking about this problem. Seems it had no ending.
...
> Here's what I'm doing. We've got two torques at the wheel. One is caused by
>the brakes/engine torque *to the road*, and the other is caused by the road
>*to the tire* in the other direction. [long text about torques]
> You get the idea?
I think I do yes, thank you very much. I must prove this to myself by
actually implementing it in software, which seems quite simple now,
given that I have the curve part of the slip ratio handled (I've
created a curve editor).
...
> By spinning percentage, I mean slip ratio times 100. One tire may exert it's
>maximum force at .1 slip ratio (or 10% slippage). The tire's speed will *jump*
>to the point where the torque FROM the road TO the tire equals the torque FROM
>the tire TO the road. (Actually, it may jiggle around in there a little, but
>it's so subtle I can't see it.)
Would be fun perhaps to have a view in the simulation of the slip
ratio traction curve, and see where you are with the wheel. Or things
like that to analyse what you're doing. Well, that's for when it feels
correct. I'll try the 2 torque thingy tonight.
Thanks for helping me cross this steep bridge, I would be offputting
to say the least when not being able to get help like this over the
net. :)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (10 lines)
>>Combining the two is what I do wrong at the moment; in the latter >>case, the wheels turn faster than the car does, so I apply a force to >>the car (translation). However, somewhere the wheel is also angularly >>accelerated, and the two result in the wheel always turning faster >>than the car velocity. Until roll resistance or something like that >>comes in. >> > > Exactly right. The wheel *should* be spinning faster than the car velocity. >This is what slip ratio *is*. Seems wierd, doesn't it?Quoted text (11 lines)
>>The problem I had basically was that I wanted to know, for a given >>torque at the wheel axle, how fast the *wheel* itself is going to >>rotate. This is dependent on the body mass; a heavy car will >>accelerate its wheels more slowly than a light car. > > It will, but you don't need to calculate that. > >>But what is the >>ratio of wheel rotation and car acceleration? > > You don't need to calculate this either. (Are you getting mad yet ;-) ?)On 18 Sep 2000 22:55:38 GMT, [email hidden] (J. Todd Wasson) wrote:
...
> Ignore the weight of the car completely right now. In the first time step,
>the car is at a dead stop with no wheel rotation. Now apply torque to the
>driving wheel. Accelerate the wheel according to torque/(moment of inertia).
>It speeds up a little and therefore has a slip ratio. The slip ratio will
>accelerate the car, yes, but forget that for now. Assume the car is cemented
>to the ground and can't move....
Ok, got it working last night. Works reasonably fine, but mostly at
higher speeds with less varying slip ratios. Numerical instability is
quickly present at low speed, so the front wheels for example see that
they're not moving and the ground is, so SR=-1. This gives a great
force which rotates the wheel; a little too fast, so the next step (I
use 100Hz mainly for simulation, I tried 250Hz and 1000Hz and the last
one worked much better already, but at quite a cost) the wheel
suddenly rotates very fast, so it has a slip ratio of >>0 (say, 10 or
100). So it reverses back up, and pingpongs around 0. Nice though to
see that if I throw the car with a longitudinal velocity (your eh,
X-axis according to SAE axes) the velocity at which things stabilize
is the same for all integration frequencies. Phew. That takes care a
bit of correct laptimes. :)
So it's shaky when <250Hz. I think I'll have to soften the forces for
low speed situations. It even happens and not so low speeds, even, so
all the more need for that.
After some more reading I can probably do the simulation of the tire
seen as a spring for low velocities (other good-working ideas for
getting rid of low-speed numerical instability?).
And do the fabulously feeling of the car bouncing back a bit when you
brake in front of a traffic light, for example, which Dave Pollatsek
described. Well, in the first place it would be to keep sim frequency
down will maintaining a correct feel. You are racing, so mostly speeds
will be high, so 99% of the time is spent with nicely working slip
ratio's.
> Time 0----Driving torque is 100 Newton*Meters
> At 0 km/hr and 0 slip ratio, you'll have 100 Newton*Meters accelerating the
>wheel.
> Time .001 --- Driving torque is still 100, the wheel has sped up and now
>there is a slip ratio of maybe .02.
How would you get to that .02?
As V=0, the slip ratio in essence becomes infinite at my place. That's
a main reason why my tires jerk around at low speeds, because SR
bounces back from negative to positive all the time, and the
integration step is a bit too big, so the tires starts shaking in
forward and backward rotation.
Currently I do a simple if(v<.001){ SR=10 }, sort of.
> This may mean (according to load, total
>grip, however you model the tire) that there is 20 Newton Meters being applied
>to the road at that instant. This force will accelerate the car a little.
I currently have a normalized traction force curve, which goes from
0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a
friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit
unrealistically high?).
So the force becomes:
curve->GetValue(slipRatio)*frictionCoefficient(1.7)*forceNormal
Where forceNormal is just load * gravity. I'm reading up on ride and
roll rates and suspensions to get some load transfer later on.
Just wondering still how one would select axis systems in a car, where
you have a body rolling and pitching, and the plane through the 4
wheels doing a similar type of rolling. When would roll lead to an
actual change of roll in a car, instead of just being a temporary roll
of the body wrt the wheels?
Probably they are both one and the same thing, but I must read up on
roll and ride to understand that. Before I let my body roll turn into
camber, which doesn't feel right.
...
> You get the idea? And yes, the driving wheels will be turning a little
>faster than the car velocity during this process(this IS correct, according to
>the texts). When hitting the brakes, they will be turning a little slower than
>the car velocity. This difference is the slip ratio.
Brake torque is calculated by something like:
- maxBrakingTorque=5000 Nm (a constant)
- brakingTorque=controllerBraking*maxBrakingTorque
(where controllerBraking=0..1)
Right?
Anyway, it's working now, albeit with some numerical jumping. Back to
getting it all working in world coordinates, as I moved back to basics
for the new slip ratio model.
OpenAL for the sound seems easy to use and plugs in nicely to OpenGL,
but DirectSound3D offers better 3D support, esp. probably when on
audio cards that support it. Haven't decided yet.
Thanks,
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
<snip>
> I currently have a normalized traction force curve, which goes from
> 0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a
> friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit
> unrealistically high?).
I seem to remember numbers of 0.8 - 1 for road tyres, and 1.2-1.4 for race
tyres. This was A level maths, so may be inaccurate, hope it helps though.
I've been following this thread with quite some excitement, I hope to see
the finished result (I'm trying your efforts so far at this moment). Good
luck.
Andy
--
Visit my site... http://mail.wm-europe.com/~andrew/
Read about my Mini... http://mail.wm-europe.com/~andrew/mini/
Use console pads on a PC ... http://www.consolepads.co.uk
Music videos and DC games ... http://mail.wm-europe.com/~andrew/ftp.html
On Tue, 19 Sep 2000 13:45:36 +0100, "Andy Laurence"
<[email hidden]> wrote:
Yes, thanks. It seems F1-2000 is bending its physics rules quite a bit
to get things going. How odd; I would have done such things with magic
factors in the source code, or at least not show it this bluntly in
the .veh files. Seems 2.7 is way too much, even for an F1 tire.
>I've been following this thread with quite some excitement, I hope to see
>the finished result (I'm trying your efforts so far at this moment). Good
>luck.
That version does quite nicely. ;)
However, I've broken things down and am rebuilding it for the slip
ratio model, so I'll have to patch things up to get it to turn again.
With a slip angle model, that is.
Some teasers from the Log:
- Slip ratio/slip angle modeling
- Better integration; fixed simulation frequency (100Hz); editable
(the downloadable version is in slomo; you can edit out the
fixed_timespan=.01 line to get realtime)
- Body stub graphics
- Frames/second counter (fps)
- Wheel graphics may be replaced by .ASE model files (using 3D Studio
Max for example)
- Front wheel drive works reasonably well (modify testcar.ini)
- Joystick/wheel controller input (Win32/DirectX7 needed); MSFF only
- Curve editor (Win32/MFC) to edit force curves
There still a million things to be done though. Have done some
thinking about a track format, BSP trees (or octrees or just 2D BSP
trees, because tracks are mainly 2D), matrices, DirectPlay etc. etc.
etc.
GPL's track format is great for nicely banked and pitched tracks (have
studied it some time ago), but somewhere I would like it if you could
just drive where you wanted. First things first, a flat track with no
banking or whatever, will probably be the first step. And then
something with pitch and height, so a Stunt Car Racer type should be
possible. Still reasonably far away though. But fun if it goes into
that direction.
But ofcourse the car physics remain at least half of the simulation's
fun, especially for the people here at RAS! ;-)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (8 lines)
><snip> >> I currently have a normalized traction force curve, which goes from >> 0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a >> friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit >> unrealistically high?). > >I seem to remember numbers of 0.8 - 1 for road tyres, and 1.2-1.4 for race >tyres. This was A level maths, so may be inaccurate, hope it helps though.Quoted text (7 lines)
> ><snip> > >> I currently have a normalized traction force curve, which goes from > >> 0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a > >> friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit > >> unrealistically high?). > > > >I seem to remember numbers of 0.8 - 1 for road tyres, and 1.2-1.4 forQuoted text (17 lines)
> However, I've broken things down and am rebuilding it for the slip > ratio model, so I'll have to patch things up to get it to turn again. > With a slip angle model, that is. > > Some teasers from the Log: > > - Slip ratio/slip angle modeling > - Better integration; fixed simulation frequency (100Hz); editable > (the downloadable version is in slomo; you can edit out the > fixed_timespan=.01 line to get realtime) > - Body stub graphics > - Frames/second counter (fps) > - Wheel graphics may be replaced by .ASE model files (using 3D Studio > Max for example) > - Front wheel drive works reasonably well (modify testcar.ini) > - Joystick/wheel controller input (Win32/DirectX7 needed); MSFF only > - Curve editor (Win32/MFC) to edit force curvesQuoted text (11 lines)
> There still a million things to be done though. Have done some > thinking about a track format, BSP trees (or octrees or just 2D BSP > trees, because tracks are mainly 2D), matrices, DirectPlay etc. etc. > etc. > GPL's track format is great for nicely banked and pitched tracks (have > studied it some time ago), but somewhere I would like it if you could > just drive where you wanted. First things first, a flat track with no > banking or whatever, will probably be the first step. And then > something with pitch and height, so a Stunt Car Racer type should be > possible. Still reasonably far away though. But fun if it goes into > that direction.Quoted text (20 lines)
>Ok, got it working last night. Works reasonably fine, but mostly at >higher speeds with less varying slip ratios. Numerical instability is >quickly present at low speed, so the front wheels for example see that >they're not moving and the ground is, so SR=-1. This gives a great >force which rotates the wheel; a little too fast, so the next step (I >use 100Hz mainly for simulation, I tried 250Hz and 1000Hz and the last >one worked much better already, but at quite a cost) the wheel >suddenly rotates very fast, so it has a slip ratio of >>0 (say, 10 or >100). So it reverses back up, and pingpongs around 0. Nice though to >see that if I throw the car with a longitudinal velocity (your eh, >X-axis according to SAE axes) the velocity at which things stabilize >is the same for all integration frequencies. Phew. That takes care a >bit of correct laptimes. :) > >So it's shaky when <250Hz. I think I'll have to soften the forces for >low speed situations. It even happens and not so low speeds, even, so >all the more need for that. >After some more reading I can probably do the simulation of the tire >seen as a spring for low velocities (other good-working ideas for >getting rid of low-speed numerical instability?).Quoted text (14 lines)
>> At 0 km/hr and 0 slip ratio, you'll have 100 Newton*Meters accelerating >the >>wheel. >> Time .001 --- Driving torque is still 100, the wheel has sped up and now >>there is a slip ratio of maybe .02. > >How would you get to that .02? >As V=0, the slip ratio in essence becomes infinite at my place. That's >a main reason why my tires jerk around at low speeds, because SR >bounces back from negative to positive all the time, and the >integration step is a bit too big, so the tires starts shaking in >forward and backward rotation. >Currently I do a simple if(v<.001){ SR=10 }, sort of. >Quoted text (7 lines)
>I currently have a normalized traction force curve, which goes from >0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a >friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit >unrealistically high?). >So the force becomes: >curve->GetValue(slipRatio)*frictionCoefficient(1.7)*forceNormal >Quoted text (12 lines)
>> You get the idea? And yes, the driving wheels will be turning a little >>faster than the car velocity during this process(this IS correct, according >to >>the texts). When hitting the brakes, they will be turning a little slower >than >>the car velocity. This difference is the slip ratio. > >Brake torque is calculated by something like: >- maxBrakingTorque=5000 Nm (a constant) >- brakingTorque=controllerBraking*maxBrakingTorque > (where controllerBraking=0..1) >Right?Quoted text (15 lines)
>><snip> >>> I currently have a normalized traction force curve, which goes from >>> 0.0 to 1.0 and back to .60 or so (taken from RCVD). I also have a >>> friction coefficient, set to 1.7 for now (2.7 from F1-2000 seems a bit >>> unrealistically high?). >> >>I seem to remember numbers of 0.8 - 1 for road tyres, and 1.2-1.4 for race >>tyres. This was A level maths, so may be inaccurate, hope it helps though. > >Yes, thanks. It seems F1-2000 is bending its physics rules quite a bit >to get things going. How odd; I would have done such things with magic >factors in the source code, or at least not show it this bluntly in >the .veh files. Seems 2.7 is way too much, even for an F1 tire. > >On Tue, 19 Sep 2000 17:29:44 +0100, "Andy Laurence"
<[email hidden]> wrote:
...
>> Seems 2.7 is way too much, even for an F1 tire.
>
>I have to agree, although I expect values are quite high in F1 (maybe 1.7).
Yep, in RCVD, page 27 it says something like 'a Grand Prix tire can be
as much as 1.8'.
...
>> That version does quite nicely. ;)
>
>I've had a play, and it handles superbly. It seems to be very slow, but
>that might be my PC, or did you program it that way (easier to see what's
>going on)?
It is slow indeed, it uses a fixed timespan, so it runs the sim at
100Hz. From below: ;-)
(the downloadable version is in slomo; you can edit out the
fixed_timespan=.01 line to get realtime)
This however means that it will run in the frequency given by how fast
your card draws things. So it is a bit more jerky. But realtime. Fixed
that some time ago.
...
>What are the chances of us being able to design our own tracks? Maybe to
>the extent that we could copy a city (or, rather, part of) to drive around,
>similar to what they did in Driver. That would be great, I'd certainly try
>Bristol, UK if we could.
Let's not move forward too soon! :)
I'm looking to create an MFC/OpenGL app to create 'tracks' (in
whatever sense). What is needed is that I can incorporate a racer.lib
so I can just flexibly draw a 'track' from where I want (which also
means correct orthogonal views; top/left etc). Same would be if I were
to make a camera editor.
The problem is that are a zillion ways to create tracks, and I want to
have some way of being able to integrate most of them into the Racer
class hierarchy in some way. So I don't really have to create an
editor for each type of track layout I want to experiment with (that
would mean all track types would end up being ASCII files probably,
with no GUI editor).
First thing though would be for example a flat kart-like track, which
is small enough so I don't have to do BSP stuff yet. First things
first.
>> But ofcourse the car physics remain at least half of the simulation's
>> fun, especially for the people here at RAS! ;-)
>
>Definitely, please, keep us informed (and entertained).
I will :)
Ruud van Gaal, GPL Rank +53.25
MarketGraph : http://www.marketgraph.nl
Pencil art : http://www.marketgraph.nl/gallery/
Car simulation: http://www.marketgraph.nl/gallery/racer/
Quoted text (17 lines)
>"J. Todd Wasson" wrote: > >> >After some more reading I can probably do the simulation of the tire >> >seen as a spring for low velocities (other good-working ideas for >> >getting rid of low-speed numerical instability?). >> >> When the tire's true velocity gets below a certain value, try.... > >> I actually wrote a whole seperate program to analyze and figure out what >the >> problem was. Can you believe it?? Took about 3 lines of code to make it >go >> away and 300-400 lines to figure out!! (Shaking head.) > >There have been entire engineering studies and papers written on the subject. >It's not as sexy as E = m c**2, but it can still keep one awake at night ;) >Quoted text (7 lines)
>> This is the other thing I mentioned earlier. The slip ratio "difference >> instead of ratio" method at low speed prevented instabilies, but once you >> introduce a big braking force, it starts up again. > >Introducing your difference approximation (you linearized the slip ratio) >kept the frequency of the system from quickly going toward infinity at >low speeds but at realistic force levels it's still pretty high.Quoted text (15 lines)
>> This cause for this started >> (in my model anyway), by torque from the brakes acting in an opposite >direction >> from wheel rotation. If the wheel is turning slowly forward, a big braking >> force makes it suddenly spin BACKWARD, then forward and back, forward and >back, >> etc.. The bigger the force, the bigger the instability/jumping. > >At my shop we term this friction reversal and it can occur and has to >be dealt with almost everywhere. The tires both when stopped and >in some conditions even when moving, all through the drive train, >collisions, ... > >- Matt >
"J. Todd Wasson" wrote:
There have been entire engineering studies and papers written on the subject.
It's not as sexy as E = m c**2, but it can still keep one awake at night ;)
> This is the other thing I mentioned earlier. The slip ratio "difference
> instead of ratio" method at low speed prevented instabilies, but once you
> introduce a big braking force, it starts up again.
Introducing your difference approximation (you linearized the slip ratio)
kept the frequency of the system from quickly going toward infinity at
low speeds but at realistic force levels it's still pretty high.
> This cause for this started
> (in my model anyway), by torque from the brakes acting in an opposite direction
> from wheel rotation. If the wheel is turning slowly forward, a big braking
> force makes it suddenly spin BACKWARD, then forward and back, forward and back,
> etc.. The bigger the force, the bigger the instability/jumping.
At my shop we term this friction reversal and it can occur and has to
be dealt with almost everywhere. The tires both when stopped and
in some conditions even when moving, all through the drive train,
collisions, ...
- Matt
Quoted text (9 lines)
> >After some more reading I can probably do the simulation of the tire > >seen as a spring for low velocities (other good-working ideas for > >getting rid of low-speed numerical instability?). > > When the tire's true velocity gets below a certain value, try.... > I actually wrote a whole seperate program to analyze and figure out what the > problem was. Can you believe it?? Took about 3 lines of code to make it go > away and 300-400 lines to figure out!! (Shaking head.)
"J. Todd Wasson" wrote:
> Friction reversal? That's a good name for it. Any tips on how to handle
> it?
As was suggested in this thread, if it changes sign,
try holding it stopped for awhile. Of course, you then have to
come up with all sorts of tricky criteria for when to
stop and when to release it! ;)
The general problem with 4 wheeled cars is difficult because
of the large number of different possible reversal situations you
can be in at any one time, and also the potential for redundancies.
(Can cause numerical problems when solving the equations of motion.)
- Matt
