To Push or Pull in 2014

[hollus]

My understanding of suspensions is quite limited, so in this department I only have questions.

After seeing thise videos:

It seems that most designers this year payed attention to this guy in class.

Enrique Scalabroni explains the 2014 suspension intricacies:

http://www.youtube.com/watch?v=izRhrcI7S7c

http://www.youtube.com/watch?v=9YvV2Qvp75M

If I understand it right, when compressed by a bump (but also by downforce), the suspension will react, partly, by increasing the track, the distance between the wheels, and more so in the case of a pull rod.

Questions:
a) Wouldn't this make the car transiently illegal as it becomes too wide? I remember when Michelin had to change the tires midway through the season as they were bulging too much at speed.
b) Would this wider track be beneficial for performance? I am thinking that bumps are too transient, but downforce isn't, and it would make the car a bit wider, which counteracts roll. Would this be useful if only at one end of the car?
Also, a perverse effect of this in 2014 is to make the out-wash wing an even more difficult trick to pull. Or maybe not?

[thisisatest]

the first question in particular is intriguing.
first, though, strictly speaking it's not a case of whether the suspension is pushing or pulling, but rather the wishbone inclination when viewed head-on. a pull-rod becomes more viable if the wishbone slopes downward from the chassis to the wheels, and it is this wishbone orientation that is prone to track width variances in its travel.
that said, im going to bet that it is less of an issue than what is immediately apparent. looking at the Ferrari front end, the upper wishbone slopes to a greater degree than the lower wishbone. this give camber loss in compression. without this compromise, the front roll center would be quite high, jacking forces would be high in a corner, and the suspension would be jerky and ineffective.
with this setup, the overall track width does not grow nearly as much as it would if the wishbones were parallel or diverging towards the wheels.
the rear of the car is completely different, there is always camber gain in compression, but also the criteria for pullrod viability is not the same as for the front end.

[Neno]

To Push or Pull in 2014?
Why not have both

[mep]

I think they do it to get a small aerodynamic gain regardless of mechanical drawbacks. However not every team might have experienced an aerodynamic gain in it.

[munudeges]

Pull-rod is just plain wrong.

[Lycoming]

Pull-rod is just plain wrong.

Care to elaborate?

[thisisatest]

thank you, dynatune, for validating my observations.

[dynatune]

Pull-rod is just plain wrong.

Care to elaborate?

I do not agree that pull rod is plain wrong. It all depends on the front view installation angle which defines more or less the motion ratio and as we have seen in the video the forces in the pull rod. If one would look at it carefully in details one can come to the conclusion that, those angles are fundamental for the installed stiffness of the system. Lets assume that the pull-rod (or push-rod) exists out of 5 parts: Bracket on wishbone, rose-joint bracket, carbon rod, rosejoint bracket and fixture on rocker. Now these rose joints are usually diameter 7 to 8 mm and assuming a stiffness of 25 kN/mm for those and the brackets (this means that those tiny titanium parts deflect only 1 mm when lifting 2,5 ton of steel) and assuming a carbon rod stiffness of 100kN/mm (to be fancy) the total stiffness of the whole system becomes: 1/(1/25 + 1/25 + 1/100 + 1/25 + 1/25) = 5,88 kN/mm. Even if we double all the involved stiffness the system will face difficulties in real life to arrive above 10 kN/mm. And this is only the "push/pull rod" part. So there is a significant amount of deflection going on and here comes in the angle. If the pullrod angle is very unfavorable (just like on the Ferrari) the forces increase dramatically if running very stiff springs a significant part of that rate can be lost causing the car not to react to setup changes. Imagine a horizontal pull rod ... Now the same is obviously valid for a push-rod but due to the package their angles is usually a lot better. The main advantage of the pull rod is that it's load are in tension and there is no risk of buckling.

Cheers,
dynatune, http://www.dynatune-xl.com

[dynatune]

thank you, dynatune, for validating my observations.

good observation !

dynatune

[munudeges]

Pull-rod is just plain wrong.

Care to elaborate?

Errrr, try watching the actual fricking video?

You're sacrificing a heck of a lot of mechanical certainty and stability for an extremely hypothetical and tenuous aerodynamic gain based on nothing more than 'freeing up some space'. Scalabroni lays it right out in his scepticism and you don't have to be much of a genius to pick that up.

What's puzzling with the Ferrari is they've kept this suspension arrangement and then haven't given themselves a high enough nose to maximise the volume of air available.

[Jersey Tom]

Pull-rod is just plain wrong.

Care to elaborate?

Errrr, try watching the actual fricking video?

The video is kinda BS y'know. As is how it goes with "things on the internet."

Push-rod, pull-rod, all the same stuff fundamentally. Crazy how worked up people get over it.

[Tim.Wright]

I'd agree... I don't see what the push/pul rod has to do with the location of the geometric roll centre. I'm also not convinced that his calcs on the push/pull rod forces were done right.

In your experience, do you know/think teams might be doing this regardless of it giving camber loss in bump?

It actually seems like a resonable apprach to me since front ride height control is probably more importantant than a few hundredths of a degree of camber. Especially so given the short stroke of modern suspensions and the possibility of just dialing more static or caster induced camber to compensate...

I would suspect every team on the grid to try to get the Front Roll Center Height as low as possible. When I was designing F1 suspensions the front kinematic rollcenter was below ground level (using the same "trick") but due to the lateral camber compliance (which will always move the "force based" roll center up, and that's the one that counts) we were between 0 and 50 mm. But lately a lot of teams have worked a lot on upright compliance reduction, one can see why . All other negative effects can more or less be compensated by static camber setting or caster induced camber with steering.

Cheers,
Dynatune, http://www.dynatune-xl.com

Interesting... Whats the mechanism which causes camber compliance to move the force based roll centre up? Is it a geometric thing where the compliances causing forces to point in directions transverse to the links? Or perhaps the restoring Mx moment is causing an increase in the contact patch Fz?

Tim

[dynatune]

The roll center can as you certainly know be calculated "geometrically". However I would recommend to forget the classic approach of drawing lines through links because that does only work well for 2D suspensions (Experts divide suspensions in 3 types - 2D, 2,5D and 3D or "plane", "spherical" & "3 Dimensional") and not so well for multi-link suspensions.
In my opinion it is better to determine the Roll Center Height geometrically via the line that starts out of the "normal" vector to the "lateral contact patch displacement curve over wheel travel" (=trackwidth change @ contact patch over wheel travel) in any given vertical position.

Now the roll center defines as we also know the momentary point of rotation of the front suspension (left & right) which means from a mechanical point of view, that the resultant of all contact patch forces must point into that direction (sum of moments = 0 --> Resulting Force hits point of rotation). Based on this fact one can establish the "force based roll center" out of delta increase in vertical load and delta increase in lateral load . On a K&C rig (kinematics and compliance rig) one usually executes both measurements but since for a pure vertical motion the loads in the suspension links are in general lower than in a combined vertical & lateral load case the effect of "camber compliances" on the "contact patch lateral displacement curve" is hardly affecting the results of the geometric measurement but will affect the results of the force measurement.

In the many K & C test that I have executed I have always found the force based roll center higher than geometric roll center and the stiffer the suspension was in lateral camber compliance the less the difference in roll center height was between the two methods. Beyond that, even when the geometric roll centers were below ground the force based were always above. I am sure one could investigate more on this topic ....

I would certainly agree with you on your initial thoughts and see the mechanism as a combination of "forced" geometric changes to the lateral contact patch curve (including of course all effects of link deflections) combined with the effect of the restoring MX moment on the contact patch and the corresponding change in vertical force. I am pretty sure however, that the camber compliance factor can be seen as a factor that the "driving" car will see/affect whilst with respect to MX effects of the tire I have some serious doubts on the actual behavior of the tire under rolling conditions in real life vs. a static rig test. Food for thought indeed

Sorry for being a bit long, but it might give other people also a chance to follow the topic.

Cheers,
dynatune, http://www.dynatune-xl.com

[Tim.Wright]

Ok interesting, thanks for the reply. Apologies in advance for the long winded reply...

So, if I understand correct, the way you are defining the RC is using a combination wheel travel + lateral load test and finding the intersection of the L+R contact patch normal vectors? Or are you looking at the dFz/dFy slope to imply the normal vector?

the first method is basically the method that Adams and the MTS K&C rigs (both of which I know you are familiar with) are using to define the roll centre height, but typically people are only looking the RC calculations during pure wheel travel tests without the lateral load.

In this case it seems you would get a lower roll centre with a combined WT + Fy test, because (for the outside wheel) the lateral load is reducing the lateral contact patch movement which lowers its respective trajectory normal (n-line) to a more horizontal slope and therefore the RC height follows.

I have always been a bit sceptical of this method (using contact patch n-lines) of implying a roll centre because I'm not sure it's valid to imply normal lines in a system that is not purely kinematic.

The method I have been using lately (not just for roll centres but also anti-X) is to look at the change in Fz due to an input Fy (or Fx). The change in Fz is the jacking force/geometric load transfer which is going through the links. Since these forces act on the chassis the the direction opposite to the roll moment, the can be considered an anti roll force

Since, in my eyes, the primary point of the roll centre is to determine the split of rolling (through the springs) and non-rolling (through the links) components of the contact patch vertical force, this method gives a more "direct" representation of that effect using forces only and no geometry (and, incidentally no single "roll centre" point either).

Consider the case of a pure horizontal trailing link suspension. Kinematically, you have the roll centre on the ground because the contact patch has no lateral displacement. If you add a decent amount of camber compliance (say with a compliant wheel bearing), you get contact patch displacement towards the inside of the car and the n-line and roll centre goes under the ground. This implies that you have a negative jacking force, but in my opinion this isnt possible if the only link in the suspension is transverse and horizontal to the ground. So in this case, the trajectory of the contact patch is implying that the IC of the suspension is in a different place to where it physically is. In my opinion, its implication that the roll centre is below the ground is wrong...

When I've got a bit of spare time I would like to compare the 2 methods in Adams because I'm sure they will give different results.

As to which one is correct...

[dynatune]

You might have partly misunderstood me. I have also used in the past the vertical motion test and found just like you the dFz/dFy method to provide different results. The results with the latter method were generally providing higher roll centers than the geometric method. Just like you I tend to myself to prefer the dFz/dFy method for the reasons you also mentioned, but I can imagine that combined effects could also play a role.
Whether a "combined" method would be better/preferable or even bullet proof would have to be investigated ... but as I mentioned before, since we do not know for sure what the rolling tire (and the contact patch) does do or even where the CP accurately is I wonder whether - with what we can identify today - this is going to be anyway accurate enough. And then even more complex ... how to filter out of ADAMS the root cause ? I remember having worked in the mid 90s on quite an extensive research project to "visualize" the roll-axis in ADAMS out of all those data. We did learn a lot but we were not able to create a 100% bullit proof "rule" on how and why.

Ok interesting, thanks for the reply. Apologies in advance for the long winded reply...

So, if I understand correct, the way you are defining the RC is using a combination wheel travel + lateral load test and finding the intersection of the L+R contact patch normal vectors? Or are you looking at the dFz/dFy slope to imply the normal vector?

the first method is basically the method that Adams and the MTS K&C rigs (both of which I know you are familiar with) are using to define the roll centre height, but typically people are only looking the RC calculations during pure wheel travel tests without the lateral load.

In this case it seems you would get a lower roll centre with a combined WT + Fy test, because (for the outside wheel) the lateral load is reducing the lateral contact patch movement which lowers its respective trajectory normal (n-line) to a more horizontal slope and therefore the RC height follows.

I have always been a bit sceptical of this method (using contact patch n-lines) of implying a roll centre because I'm not sure it's valid to imply normal lines in a system that is not purely kinematic.

The method I have been using lately (not just for roll centres but also anti-X) is to look at the change in Fz due to an input Fy (or Fx). The change in Fz is the jacking force/geometric load transfer which is going through the links. Since these forces act on the chassis the the direction opposite to the roll moment, the can be considered an anti roll force

Since, in my eyes, the primary point of the roll centre is to determine the split of rolling (through the springs) and non-rolling (through the links) components of the contact patch vertical force, this method gives a more "direct" representation of that effect using forces only and no geometry (and, incidentally no single "roll centre" point either).

Consider the case of a pure horizontal trailing link suspension. Kinematically, you have the roll centre on the ground because the contact patch has no lateral displacement. If you add a decent amount of camber compliance (say with a compliant wheel bearing), you get contact patch displacement towards the inside of the car and the n-line and roll centre goes under the ground. This implies that you have a negative jacking force, but in my opinion this isnt possible if the only link in the suspension is transverse and horizontal to the ground. So in this case, the trajectory of the contact patch is implying that the IC of the suspension is in a different place to where it physically is. In my opinion, its implication that the roll centre is below the ground is wrong...

When I've got a bit of spare time I would like to compare the 2 methods in Adams because I'm sure they will give different results.

As to which one is correct...