Hello.

I've been thinking about chassis flex for a while now and come to some conclusions, while on other sides I didn't come to a conclusion yet. That's why I want to start a discussion here. I'm aiming at a high technical level of discussion, so some knowledge of vehicle mechanics, dynamics and elasto-mechanics might come in handy. If we can sort out everything we could move the finished discussion to the improvement suggestions as a finished chassis flex model.

When speaking about chassis flex we speak about deformation of all kinds.
In elasto-mechanics we destinguish between bending, twisting (torsion) and stretching.
Let's go through them one after another:
1. Stretching:
Stretching occurs when on both ends of the chassis opposing forces are acted. In other words: The front wheels accelerate and the rear wheels decellerate or vice versa. This is a very uncommon situation. I can't imagin a stable situation where forces like this would act upon the car. Also the effects of stretching are minor. All stretching does is increase or decrease the wheel-base by a small fraction. Thus I came to the conclusion that stretching can safely be ignored, unless someone can come up with arguments that indicate otherwise.
2. Bending:
When would the chassis bend? You have to apply torque upon the chassis.
The torque can be either applied around the vertical or lateral axis (in the case of the longitudinal axis we talk about twisting, third section).
Torque around the lateral axis leads to bending of the chassis towards or away from the ground. This can only have an effect on the car when the car is very low and the suspension is very stiff (so the body of the car doesn't touch the ground anyway). This can typically be found in singleseaters. But when does torque around the lateral axis appear? Torque around the lateral axis means that there are opposing but apart forces in the plane drawn by the longitudinal and vertical axis of the vehicle. In LFS we have upwards pointing forces from the wheels and downwards pointing forces from downforce or from inertia (while travelling over a bump, f.e.). The locations of the forces are always the same for each car. Front wing, rear wing, undertray downforce, front and rear axis, center of gravity. From this we can calculate the complete line of the deformed vehicle or the maximum bending as a function of these forces within a few minutes. No magic here.
The other possible way of applying torque would be around the vertical axis. This would mean that there are opposing (but again apart) forces in the plane drawn by lateral and longitudinal axis of the car (this is the plane that the car moves in while on flat ground).
This would mean that f.e. the right front of the car accelerates while the left front decellerates. This is a untypical situation for a car and can only appear in a situation where the player has no control over the car anyway (as in hitting a wall after a crash). I'm open to all arguments that indicate otherwise, but until then I will ignore this kind of deformation.
3. Twisting (torsion):
The chassis of the car is twisted when there is torque around the lateral axis of the car, which means that there are opposing but apart forces in the plane drawn by the lateral and vertical axis. Typically this would be the forces of the left front and left rear wheel on one side and the right front and right rear wheel on the other side.
This leads to twisting of the body (the thing we try to decrease with ARBs) and obviously needs to be modelled because of the effects of body roll. The maths behind this is easy. There are two places where torque applies (front and rear axis) and there is no torque outside the wheel base.
There can be another kind of torque around the longitudinal axis that the upwards pointing forces from the wheels and the downwards pointing forces from inertia and downforce create. But the lever arm of these forces is short and most of the forces on the axis is absorbed into the body by the suspension, which should keep the results small. I'm not yet sure wether this needs modelling or not.
Also I'm not sure about the effects. Due to the suspension which is mounted directly at the wheels the axis, which dictates the orientation of the wheels (camber), isn't a loaded part. So even when the chassis bends this doesn't have to result in added dynamic camber because the axis doesn't have to bend. I would need more technical knowledge of the mounting of the axis to speak about this.
Let's just assume that there is bending that leads to added dynamic camber to complete the discussion. To calculate the occuring bending we need all vertical forces that act on the two chassis parts at front and rear axis. LFS already knows these and every student of mechanics can calculate them. Thus three forces act upon each the front and rear axis-parts: The forces from each wheel and the downwards pointing sum of forces from the body (namely mass and downforce). Again we precisely know where these forces apply and can calculate the maximum bending as a function of these forces as well as the angle of the axis at the two wheels (which adds to dynamic camber).
Again I'm open to all suggestions and arguments here.

In a nut shell:
Stretching doesn't need to be simulated.
Bending in vertical direction needs to be simulated.
Bending in lateral direction doesn't need to be simulated.
Twisting needs to be simulated.
Bending of the axes might need to be simulated.

Not yet added:
The mathematical equotations.
(I will add these once we can agree on what needs to be simulated and what not, and where we need to take something else into consideration, etc.)
The influence of dynamic behaviour. The chassis acts like a spring and bends dynamically. It oscillates at a specific rate with specific dampening. The depths of simulations needs to discussed because the spring rates are very high and would require many iterations in LFS to avoid stability problems.

Guideline to replying (or: what I want you to discuss on):
1. The above arguments. Are there errors, do we need to take something else into consideration? Did I go into too much detail somewhere and discussed effects that are too minor? If you are knowledgable enough about vehicle mechanics you could comment on the dynamic camber part from section 3 and wether high load would lead to deforming of the axis.
2. The yet missing parts. If you have the necessary knowledge you can write about the typical oscilation frequencies of the chassis and in what way it should be simulated.
What you should not post:
Anything that begins with "I think", "I believe", "I heard that", "As far as I know" etc. Please use facts. Otherwise you will only distract from the discussion.

Vain

Quote from Vain :

This leads to twisting of the body (the thing we try to decrease with ARBs) and obviously needs to be modelled because of the effects of body roll. The maths behind this is easy. There are two places where torque applies (front and rear axis) and there is no torque outside the wheel base.

uhm ... no arbs counteract body roll not body flex

i guess your missing the mark a bit with not pointing the discussion into the direction of whether those can be modelled with simple models for small bending or if youd nee more elaborate models

Also. To what extent can you simulate chassis flex. Would you integrate each and every component and work out the forces and the interactions between each compenent part. Or would you just come up with some arbitrary figures and implement them into the car model. Could the car model accomodate chassis flex and all it's movements and interactions without a dramatic influence on processor resourses.

Also, bearing in mind that LFS generally uses fully tooled up racing models is chassis flex such an issue. These cars generally have any chassis flex engineered out of them to make for the most stable platform possible.

I think, possible, is just how far you want to open Pandora's chassis flex box. Surely it is an all or nothing intergration. Either simulate chassis flex (Which include chassis simulation, ehich could lead to some interesting future car design. Particularily by us, the general public.) in it's entirety which could potentially eat up precious resources. Or as m,entioned just arbitrarily enter in some figures and roughly give some chassis flex influence.

But then I could be wrong about everything. just trying to sound clever . . . .

Quote from Funnybear :

Also, bearing in mind that LFS generally uses fully tooled up racing models is chassis flex such an issue. These cars generally have any chassis flex engineered out of them to make for the most stable platform possible.

All the cars with rollcages are effectively spaceframe cars, all the strength is in the cage and the original shell, which is normally where the strength of the car is becomes almost totally unstructoral. The road cars in LFS atm are really like this and with a cage added (only a visual enhancement) could be passed off as Group N racing cars, which they are a much more accurate representation of.

So does that mean they do suffer from Chassis flex or can you safetly say that chassis flex is minimal in Rollcaged cars?

I think the maths are all the same. You get a inhomogene differential equation of second order you have to integrate.
The problem are the high frequenzies you get and therefore you need short time steps.

Edit: I think that is the main reason we dont have chasiflex yet

The bending frequencies of modern regular production cars seem to be in the 15Hz-50Hz range if manufacturer data can be trusted, and probably quite a bit higher for race cars. As Hahni said, a somewhat complex structure oscilating at 100Hz or more would probably be quite a CPU hog.

All you need to care is how much the tire loads and angles are affected. Anything else is just for looks. And this basically requires to do some calculations on how suspension joints "move" under certain conditions.

But it gets quite complex once you have to make the car body "oscillate" realistically in-game. Static situation is not hard, but we need dynamix chassis flex

How actually this would make handling and the cars different or better? Honest guestion, not sarcastic note or something nasty against it.

Quote from Blackout :

How actually this would make handling and the cars different or better? Honest guestion, not sarcastic note or something nasty against it.

I would not make the cars handle better, but more realistically. It is similarly overlooked feature amogst racin simulators, like the differentials.

I suppose as this is a simulation that ever aspect of a cars performance and makeup should be simulated. Including Chassis Flex.

I think this is about whether certain elements of Chassis Flex could or should be implemented into LFS.

I for one think that Chassis Flex's actual effect on car performance within LFS is so negligable atm and the actual calculations needed too resource intensive to make it viable at this present moment.

But you never know. Lord Scawen might be thinking about it and have a work in progress racing about on the holy hard drive. But then Ithink he might be focussing on other things for the time being.

Bending in vertical direction:
Simply speaking: The car's belly moves towards the ground. This means the car bottoms out at the undertray. We will someday recieve correct undertray downforce which means that downforce is increased while the car bends downwards and decreased when bend upwards. Also the wooden plank below the BF1 should simulate wear when we get correct downforce so that players with a too low suspension recieve the penalty they would in real F1.
Twisting:
This is similar to body roll. With the difference that this isn't caused by the suspension springs that shorten but by the chassis that flexes. Same effect as driving with a loose ARB.
Bending around longitudinal axis:
The wheels are mounted on the chassis. You put pressure on the middle of the chassis (looking at the cut car at the front axis from the front, f.e.) and the chassis bends which causes additional camber.

@bal00:
Do you have a source for the frequencies or did you just recall that from memory?

Vain

So is Chassis flex doable from a CPU point of view?

When the thread about that truck simulation which had flex Lord Scawen (best title ever ) seemed very interested.

And I understand it obviously would make LFS more realistic, but can't understand how. Would it be more roll in roadcars or something like that? Im sorry I ask such idiotic guestions, but I lack to see how you can feel it there is something so elegant modelled and calculated.
edit. Ok, Vain's post cleared that up

Surely chassis Flex would effect suspension occilations. Not only would the springs be compresing and releasing but if the chassis was flexing at the same time then that compresion/depression (sic) would be amplified or effected in some way that is way more technical than I can adequatly blag my way out of by sounding like I know what I'm talking about.

You can't generate a comprehensive setup with a car that suffers from too much chassis flex. Take you average family estate car. Crap engine, crap tyres and a chassis taken staight out of a John Deere. You take that around a corner at any speed and the Chassis will twist and flex acting like a spring, throwing you around the road like a pogo stick on acid. The Tyres don't have a constant force operating on them from 'above' for them to react against. So rather than dealing with the road surface alone you are giving them twice as much work to do.

Generate a chassis which is stiffer than a vicar at a swingers club and shove that car around a corner you can tune the suspension to the nth degree because you have reduced the variables.

Very layman esk and I would love for someone to come in with something techical so I can shut up and stop sounding so damn ignorant.

-Edit. Oh and I think it should be the Lord Scawen. Although that might get my into trouble for worshiping false idols. . . . No. I've given it a few seconds and no nasty lighting bolts or omnipotent voices and angelic hosts appearing from the ether. I reckon I'm safe.

Quote from Funnybear :

So is Chassis flex doable from a CPU point of view?

I don't see why it would not be doable. After all, it's all about optimization and leaving the 1% stuff out. Just like active camber/caster calculations

EDIT: Unluckily I don't have enough time right now, but the chassis flex is lot more complex than just having the chassis twisting or bending around some axles. You get different kind of chassis flex under braking, cornering and driving over bumps. But there must be someone who could do a little heads up post about it

EDIT: What about "teh Red Baron Scawein"?

Quote from Blackout :

When the thread about that truck simulation which had flex Lord Scawen (best title ever ) seemed very interested.

Yes, but while that way of doing it (the whole car body was basically a bunch of springs welded together) was probably the most accurate, it was also an enourmous CPU hog. Also you need insane calculation frequencies, or else all ends up in an oscillating mess.

The best way is probably to "just" somehow define a few axes/bending directions and have one simple calculation for each. Not the most accurate way, but accurate enough to serve its purpose (make karts realistically possible). Knowing Scawen he'd probably even code visual body flex when he finally implements it

But if you modelled all the compenents needed you talking of thousands of interactions throughout the system. Scawen would have to become an automotive engineer to construct the chassis required. Sorry. The Lord Scawen would need too, which I'm sure he's already doing. Maybe. But with all those interactions not only within the chassis to give you an output but you have to factor that output into the suspension and vice versa.

Thats alot of calculations. All happening very quickly.

-Edit. Arrrgggg. Loosing the thread post plot. Can't keep up.

I for one doubt it is realistic to think LFS should calculate every possible
forces acting upon the whole chassis. LFS has no physical chassis for
starters, basically what looks like anchor/pivot points held together in
a solid structure by an invisible 'chassis'. The simplest and most effective
way of simulating chassis flex would be to move the anchors/pivots
slightly according to forces. A simple box where each corner is at the
suspension anchor points would work well as a simplified structure. Having
the forces acting upon the box and deforming it would move the
suspension points accordingly which would simulate most of the side-
effect of chassis flex we are interested in anyways.

That would certainly get the idea across. Like I mentioned earlier it's either got to be something that you do whole hog or you implement in a very simple but effective fashion.

But even that box needs to have differering degrees of flex. Not knowing how LFS deals with this does the 'invisible' chassis have wieght (ovboiusly it needs to but bear with me) and if so can that wieght be distributed fore and aft (Of course it can do this too, but I'm getting there) so you have a nominal wieght at the front and a nominal wieght at the rear, thus you can generate twist between the two wieghts given a certain resistance value within the 'box' frame. Thus this twist value could easily transmit itself to the suspension mounts in a very simplistic and already available way (I summise). I am assuming that LFS doesn't cheat at all with the suspension mountings and that the suspension can take information from both directions. If so then that would give you the impression of mass induced flex within the chassis structure.

But again, is it actually nessesary and would anyone other than the extreme hardcore, technophile ubergeek actually notice any disernable difference. I would hazard at not. Maybe within the remit of Rallycross it might be more noticable with the softer suspension and rougher surfaces.

Maybe it's a case of what we don't have we don't miss.

well its do-able (look at RoR), but whether it meets Scawens super-standards i don't know

Quote from Funnybear :

So does that mean they do suffer from Chassis flex or can you safetly say that chassis flex is minimal in Rollcaged cars?

They'll still suffer from chassis flex, as will a carbon tub car and it will still make a difference but compared to a road car they are in a different league and without the chassis flex simulated there isn't such a great issue as with the road cars.

Quote from Jakg :

well its do-able (look at RoR), but whether it meets Scawens super-standards i don't know

RoR does chassis flex very well but even the dev notes that even with the VERY high frequencies the maths are run at its still not high enough to make the body ridgid enough. It does look very impressive though.

The chasis flex maths would have to run at far higher frequencies that most of the core physics maths to keep it accurate which would raise the bar on CPU requirements a load. It you dont run it VERY fast you get errors similar to the old pinball crash type physics but as this would be in the chassis would have serious impacts on the physics if not done right

RoR is VERY CPU heavy even with 1 truck.

It would be interesting to see how many springs are needed to model a car chassis though.

Quote from Woz :

It would be interesting to see how many springs are needed to model a car chassis though.

3.
That is because we don't use springs, aka rods, but bars. We simplify the car chassis to three bars of material. One from front to rear axis, one at the front axis, one at the rear axis, 3 in total. In elasto-mechanic theory a bar can do all deformations that a car chassis can do. And we know very well how that bar aka chassis behaves. The simplification is minor, but the change in CPU usage is major. 'Static' chassis flex (without swinging at up to 50 Hz) would utilize less CPU clocks than a single LFS tyre does.
Even without the effects of dynamic oscillation we get several nice effects:
Single seaters can bottom out under chassis flex, camber increases under high load, body roll increases due to chassis twisting.
If someone has knowledge about the effects of chassis oscillation, please share.

Vain

But you can't use static calculations in a sim. The forces on the car change with every millisecond but you don't want the modelled chassis to instantly change it's state to the values the static calculations give you. Static calculations tell you what would happen to the chassis if you applied constant forces to it and then waited a bit. It tells you the final position of the chassis parts, not how they got there, how long it took or whether it even gets there before the forces chang etc. That's why you end up with Differential Equations along with all the problems of stability and oscilations.

Maybe static calculations can be used to get a reasonanble model with less CPU load but I guess we need someone with some experience in simulation programing to comment on that.