RR98ITR

I cannot believe that I've felt compelled to do this. Again. But there it is.

So, to bring us all up to date, sometime back I started a thread about sway bars, not that you could find it in a search easily given my naming conventions:

https://honda-tech.com/forums/road-racing-autocross-time-attack-19/did-you-see-nicholas-cage-movie-national-treasure-2761537/

In that thread I "dramatized" my effort to understand how bars work - how in the quantitative sense. And part of the point of that dramatization was intended to convey that there's alot of misunderstanding and sloppy thinking and writing out there not making arrival at a valid understanding any easier.

So where we got to ultimately was Claude providing a nice derivation of the basic formulas that nicely complemented the authoritative work of Olley & Milliken and a couple of others.

Recapping ever so briefly:

The roll stiffness from the springs = wheel rate * (track ^ 2) / 1375 lbs*ft/degree

The roll stiffness from the bar = wheel rate * (track ^ 2) / 688 lbs*ft/degree
(where the wheel rate here referred to is the single wheel rate of the full bar as is so typically calculated)

Claudes derivation provides a nice description of why we square the track, and how there are two components to the resistance of the springs to roll - that to compress the outboard spring and that to relieve the inboard spring.

If we want to demystify the bar thru similar thinking, shedding more light on the halving of the divisor from the spring equation to the bar equation, we might imagine that in roll the bar actually becomes two springs each with a rate calculated using Half of the bar length. In roll, when the bar twists, the center of the bar (L/2) doesn't get displaced. We could if we wanted use that calculated wheel rate in the spring equation and get the same result we got from the bar equation. Get it? It's not that easy. In fact I think a misunderstanding of just this point is why Staniforth wrote those vexing lines I referred to wherein a bar might provide effects 3 or 4, or 5 or 10, times greater than stiffer springs. No...2 is the correct number.

So, think about it, study Claudes derivation, refer to the big books. This looks like settled science. Yes, I understand that the average racer doesn't know this. Yes, I understand that alot of highly trained racing engineers don't know this. That doesn't change anything about the physical reality of the devices.

And finally, it was just recently that Mark Ortiz wrote about this very thing:

http://www.eviltwinmotorsports.com/w...ter-2011.2.pdf

I even understand that there are people who deny that the above is true. They actually "laugh" at the idea that a bar returns twice the roll stiffness per increment in single wheel rate. They "proved" their understanding with a set of scales. There was an article about such use of scales a while back in Circle Track, and here are a couple of links to similar procedures:

http://www.jameshakewill.com/sway-bar-rates.pdf

https://honda-tech.com/forums/road-racing-autocross-time-attack-19/sway-bar-stiffness-calculations-2887328/#post44469969

A key thing to keep in mind about use of scales - you're measuring single wheel rates, Not roll stiffness. The only thing about roll stiffness you can validate on scales like this is misunderstanding - Unless you apply the single wheel rate data to the correct bar formula.

And remember: "Bars don't reduce 'independence', they Increase it".


Scott, who doesn't know what else to say...ok, that's not true...I do:

Shinny

Lol can't wait to start learning that kind of stuff in college. Makes sense tho glad someone finally put some hard math to it all

solo-x

Why wait until college? Start now. You can learn this stuff on your own if you're motivated enough.

VTECIntegra9

This was what I was looking for in my linked post... The way I saw it was by instead of rolling the body and keeping the wheels flat, you displace the wheels... Now looking back at my post you would have to start at ride height w/ all 4 on scales (or equivalent blocks - had some of these at school for my K&C machine)... Then you would be able to move each up/down an equal and opposite amount. Would this give a closer representation to the stiffness? x2? I would like a little more explanation about this if possible... p.s. Find the treasure.

RR98ITR

Here's a link to the Circle Track article:

http://www.circletrack.com/chassiste...ues/index.html

It's a plan to measure a conventional single wheel rate for a bar and then back out the motion ratio to get a conventional bar rating. No mention of anything relating to the roll stiffness. This suggests that Bob thinks the single wheel rate value expresses the roll stiffness, which I never seem to tire of arguing it doesn't.

So what's this method of measuring a bar good for? Measuring the conventional single wheel rate of a pretzle of a bar? Sure. Get that and then you can calc the roll stiffness - which is what you really ought to be interested in.

Could you set up a direct measurement of roll stiffness using a set of scales? Hmmmm...yes...you'd have to be able to fix the chassis rigidly with respect to a reference plane (ie the floor of your shop) and measure the wheel pair loads at various opposite deflections simulating roll...it'd be alot of work to get an exact number...if you really felt you had to have an exact and direct number.

The purpose of any of this is to get a workable driveable balance. The plus or minus on optimum is probably generous. Again, and I apologize for repeating myself so much, Olley's recommendation of a bar contribution to roll stiffness of 20% is a good choice. And that can be roughly calculated/estimated.

Scott, who has been watching the interviews with Charlie Sheen and Muamar Gaddafi...they remind me of racers I have known...

ginsu2k

When I was designing the anti-roll bars for our Formula SAE car I found out that not a lot of people actually know how they work.

Yes, the bar adds roll stiffness by prevent the wheels from moving in opposite directions, but how does that effect the chassis? After all, it is the chassis that we want to prevent from rolling, right?

I did a lot of research and there wasn't much actual information on how the bars work in terms of Free-Body Diagrams, until I found a small illustration in Fred Puhn's "How to Make your Car Handle". This was an actual Free-Body Diagram showing the forces on the bar and the reaction forces going to the chassis. This was exactly what I was looking for. How does the bar effect the chassis? In reality, the twisting of the bar introduces a force couple on the mounting brackets-one side pushes up and one side pulls down). It is this force couple that creates the Anti-Roll Moment which is what actually prevents the car from rolling. I believe this is exactly what you're talking about Scott.

I agree that in the articles you posted (besides Ortiz), all of the methods of experimentally calculating the contribution of the bar simply take into account a SINGLE FORCE (one side), not the complete force couple (both sides). So I would at least agree that using the methods described by Hakewill and the Circle Track article, you could increase the calculated roll stiffness by a factor of two and hopefully not be far off.

descartesfool

I think we should first start by getting rid of the 1375 divisor and return to the dimensionless one of 1/(pi/180/2) = 114.5916 or 114.6 for short. That way the units you use for the wheel rate and track will always give you the right answer as long as the length units match for both the wheel rate and the track. You get 1375 when you multiply the dimensionless constant of 114.6 by 12 = 1375.1 to convert and answer with all length units in inches to feet for a torque in ft-lbs rather than in-lbs. But now you have a constant which has dimensions inherent in it, which is not as elegant.

So let's use Roll Rate per degree = (Wheel Rate * Track^2)/114.6 or

RR = W*T^2/114.6 for short.

Then if you put W in lbs/in and T in inches, you get the right answer, just as you do if W is in lbs/ft and T in feet, or W in Newtons/m and T in m, or W in kg/mm and T in mm. And W is the single wheel rate = Motion ratio^2 * spring rate.

Now about that factor of 2 to add or not to add for ant-roll bars. Well that depends on how the W for an anti-roll bar is measured. How are anti-roll bars measured/specified? I never know. I measured mine on a bench with one end fixed, and then on the car with one side of the car's suspension blocked and no springs after factoring out the static gas force in the dampers, but I do not know how race bars are specified. So I am really not sure about the factor 2 in the bar equation.

Might be much simpler to use the same equation for both springs and bars and then to define how the anti-roll bar's wheel rate is measured. Springs are rather easier to measure, as long as one has really big strong arms that is.

RR98ITR

I thought that's what I had done with this:

Scott, who has come to depend on Claude to provide the more elegant parts of these discussions...Thanks Claude...

VTECIntegra9

Very good points, you are checking the rate. I wish I could have access to my senior design project. I had two large Parker actuators that could grab the chassis and torque it to a certain +/- force and then measure corner weights, vertical displacements based on force as well as toe and camber change per degree of roll, instead of straight wheel travel... Now i'm stuck eight next to a well equipped race-garage or IN a poorly-equipped one... Now I'm at a point where I'm going to use the bars I have and see how the car works regardless so why all the effort right now except doing my own modeling and simulation (Speed-Wiz).

Maybe I should ignore the integra and start using this experimentation on my Dad's spitfire, (which has no more bumpsteer - woot!), but would benefit much more from a good suspension model..

Now I'm rambling - by a factor of 2?

RR98ITR

Claude - check for PM from me.

Scott, who needs to talk to Claude about theta...

TunerN00b

Does your Spitfire have the later model "swing spring" in the rear (look for a pivot under the center of the leaf spring above the diff)? Mine does, and it was rather frightening to put the front end up on a jack under the cross-member and watch the rear roll 5* without any resistance, at all, from the suspension...

If it does have that rear suspension, you're going to have some real fun trying to model the roll stiffness, and even more so if you have a "camber compensator" back there as well (it really helps with snap oversteer transitioning).

And if you do have the camber compensator back there, I'd love to see how the modeling is done, considering how you end up with 2 springs in parallel preloaded against each other with only 1 resisting roll.

RR98ITR

That's funny...I had a Spitfire with a swing spring. The only snap oversteer I remember was this one time when one of my front sway bar links broke.

Scott, who misses that car sometimes...and is glad it died and I lived...

TunerN00b

I think my front LCAs are on their 4th (and possibly 5th) reweld of the ARB attachment points. That poor front bar is completely overworked.

First corner, terminal understeer. Transition from hard one way to the other, and the tail slides out unexpectedly. Something about the suspension jacking up and positive camber or something.


I love my Spitfire. Every single time I start to feel that Honda screwed up something with the suspension design of my Integra, I go spend 5 minutes under the Spitfire, and that feeling magically vanishes...

RR98ITR

"What??? How hard could it be to understand how a sway bar works??? All you gotta do is put the car on jackstands and jack up on one wheel and you can plainly see that that forces the other wheel up too. Thus a swaybar clearly does bad things: transfering weight that otherwise wouldn't be at all, and linking the two sides of the car together and reducing the independence of the suspension. It's OBVIOUS. And as obvious is that the additional roll stiffness a bar provides comes from the bar and the inside spring in series, parallel with the outside spring. Again, OBVIOUS just from looking." (while you've got the car on jackstands)

I think in Buckaroo Banzai somebody asked the Black Lectroid if the ship the Red Lectroids were building would fly. He said yes, and added that it was a very bad design. Sometimes I just can't help but be impressed by the dedication and bad design of some of the Lectroids. And foremost among them I have to admire the one's that bristle with the utmost confidence to the observation that theirs is a very bad design.

The bar doesn't know about the springs. It only knows about angular displacement twixt the mounts on the chassis and the pickups near the uprights. Ok, if you lift the inside wheel, then then the bar and inside spring are in series, in parallel with the outside spring. But that's it. And a gas shortage and a flock of seagulls. That's it.

Please forgive me Fred, but I just have to show them your nifty drawing. PS - did you know that Arthur Mallock said your book was very good? He did.

Scott, who IS high on something: RR98ITR!...WINNING!(on jackstands)...with two smokin hot...swaybarz...

ginsu2k

That's the exact Fred Puhn pic I was thinking of...somehow the only suspension book in our entire campus library had exactly the diagram I was looking for, although it may not help much in obtaining an accurate roll stiffness value.

Free body diagrams FTW. It is so easy to see how the weight is transferred and how the ARB mounting points resist body roll.

And then all sorts of questions arise? Where do you ideally locate the mounting points on the LCA? Because it would be nice to have them at the very ends, but it introduces some serious packaging difficulties.

Also, the ARB mounts are parallel with the vehicle body plane, but on the Honda Civic/Integra chassis, they are perpendicular in the rear. Does this introduce unwanted deflection in the ARB brackets? And the bolts holding the brackets are not being ideally loaded, in that it introduces shear loads into the bolts instead of tension loads.

I do believe the Realtime Integra did a complete rework of the rear ARB, and it after all these considerations, it looks pretty awesome except that the length and angle of the ARB endlinks is definitely not ideal.

VTECIntegra9

Yes the Spit has the transverse leaf spring but it doesn't do the rear-end jacking anymore - I don't want to quote what has been changed but it has been improved up to a point. The next step would be to create an independent rear suspension, which our friend's spit does have and is wonderful to look at - almost looks like a real car back there. The front has a custom made ARB mount/bracket and the bar is through the front frame rails and can be swapped to different stiffness bars for different setups/tracks and even removed for wet setup if needed. I can take some pictures of the updated setup.

For our cars, I never realized the rear mounting point orientation to have an influence on the bar forces - but I can see where you would run into an undesirable compliance in the ARB system.

descartesfool

What i am suggesting is that if we just remember one equation for roll rate, and adjust the wheel rate for the springs versus the bar, we can do the following;

from the anti-roll bar, we get Wb and from the springs we get Ws for wheel rate.

Then if we use the one equation for roll rate,

for the bar, we get the roll rate due to the bar only:

RRb = Wb*T^2/114.6

for the springs, we get the roll rate due to springs only

RRs = Ws*T^2/114.6

now the total roll rate, RRt = RRb + RRs from both the bar and springs.

RRt = Wt*T^2/114.6 = (Wb + Ws)*T^2/114.6 in whatever unit you choose/degree.

so all we have to do is use the correct value for Wb.

And one must remember that a car just on springs without any bars is not independent in the sense of being able to move one wheel up or down without affecting the others. Just put a car on the scales, and get the corner weights. Now move up just one spring perch, or change one spring, or put a spacer under one tire, and observe all four corner weights change, along with all four of the spring compressions. Same thing happens if you add bars. The four corners are coupled through the stiffness of the chassis.

chrisdavis6

subscribed! Freakin love these kind of posts. Thumbs Up.

RR98ITR

I think you too late...it all over...less somebody starts arguing...or we get some serious input from the Eighth Dimension...

Scott, who knows it's not over...that it will never be over...but there's only so much to be done about that...

TunerN00b

Well, if you want to argue some more, we can always discuss how the McLaren MP4-12C doesn't have, or need, swaybars...

RR98ITR

Yeah but it's got the monkey business interconnected/controlled dampers. The F1 GTR had swaybars.

Scott, who's just gonna extrapolate a "Yes, swaybarz, of course" from Gordon Murray...

VTECIntegra9

Wait, what's done?

Egezzy

iTrader: (2)

bring up your shopping kart dilemma. that should stur the pot.

RR98ITR

There's always the old "They just use them to provide a cockpit adjustment".

But there's still the little matter of the physical scientific reality. Remember Warp? You could describe a warp event as concurrent deflections on a diagonal - the combination of simultaneous bumps at right front and left rear for example. Based on the line of argument in this thread, a given roll stiffness acheived using only springs would have greater warp stiffness than had that same roll stiffness been acheived using some bar*. Now we're talking about More mechanical grip resulting from the use of bars.

* I'm pretty sure I got this wrong a couple of years ago when I believed that bars and springs gave the same roll stiffness per unit of wheel rate. I'll go back and read Erik Z's articles in RCE and see if I missed any of his pearls.

Scott, who has shopping kart trouble too...

RR98ITR

Just some tidying up etc...

Claude makes a nice point about keeping the formula generic enough to accomodate different units.

If you choose to use english units then the two formulae above with the constant divisors are suitable and correct. These are taken directly from Olley except that by using the conventional single wheel rate for the bar, I omitted the bar motion ratio and lever arm from it. They are right there on the page waiting to be found and understood. You can hardly find a higher authority.

And if numbers aren't your thing, then my description of the bar in roll, whereby you see the bar as two springs with rates calc'd using half the bar length is enough to get you there.

Then, going back and rereading Ortiz from RCE V7N7 and V7N8 and Zapletal's response from RCE V10N2 I am reminded of what nice writers they both are and how without dealing explicitly with the aspect of sway bars under discussion here provide fuel for discussion.

Ortiz:

"To generalize, though, for most applications we'll get the best combination of roadholding and sprung mass control with a suspension that is stiffest in roll, also fairly stiff in pitch, softer in heave, and very soft in warp....The devices we call anti-roll bars, and all other non-rigid connections of wheel pairs, are tools we can use to get closer to the optimum."

"One sometimes hears the view expressed that an independent suspension with anti-roll bar is not truly independent. This is a mistaken notion, since an anti-roll bar is non-rigid. In fact all cars non-rigidly interconnect their wheels. Individual wheel springs and a frame constitute a non-rigid interconnection, even with no additional measures. What we're doing when we add other interconnections is to stiffen the system selectively in specific modes."

Ortiz provides a table that tells us that an anti-roll bar stiffens roll and warp modes. Duh, you say, we know that. Yes, but I want to stress the point that while true, in the sense that if you took a car with whatever spring rates, and added an anti-roll bar, you'd get higher roll and warp stiffness. But if the question is to bar or not to bar, then the answer is that for a given roll stiffness, achieved using a bar, you get lower warp stiffness. The pitfall to avoid is thinking that what is true for "adding" a bar is also true for "involving" a bar.

Zapletal doesn't say anything of direct relevance to my bar hobby-horse, but he does say something else I just have to quote, if you don't mind my switching hobby-horses in mid-stream:

"The idea is to reduce the grip at one end of the car to bring it into balance with the maximum available, but lower grip at the other end of the car. If the car does not have this even balance of grip at each end, then it is very difficult to drive at the limit."

Thank You Erik for speaking up for killing the grip at the end of the car that Is working. A front wheel drive car usually has one end that tends to have the most grip - the Rear. After you've done everything you can to get the most out of the front end, it really is ok to kill off what you need to at the rear to make the car work.

That's it. Seriously. Done. And I apologize for my part in leading anyone off in the wrong direction over the years as I too slowly found my way to the truth about sway bars. And the front of your front wheel drive race car? It should have a bar too.

Scott, who is able to understand at least one valid rationale for going barless: you're too lazy to package them and you're winning championships anyway...but that's not a workable philosophy for everyman...