Race car Damping – Some food for thought

The first article I ever wrote for Racecar Engineering was how to specify dampers using a dual rate
damper model. This was an approach that my colleagues and I had applied with considerable
success. However at the time I wrote the article I said this was definitely not the last word on race
car damping. This article will carry on this discussion.

This will be the first in a series of articles on race car damping. I’m doing this for a number of
reasons. Firstly to try and cover the whole panacea of race car damping in one article is lunacy. It is
simply too big. The second reason is that at the time writing I don’t have all the answers either.
However what I do want to do with this article is set some food for thought and really encourage a
good public discussion and hopefully bring a few people out of the wood work.

On a more philosophical note one of the greatest tragedies I see in our line of work is that it is very
rare to have a genuine technical discussion. This is as a direct result is that we are either so focused
on the business of winning we actually forget what we can learn from others. It’s also a direct result
that some times we come up with a problem at the race track that we can’t solve straight away so
we have to fudge a solution. Consider this the start of one of these genuine technical discussions
that can really delve into what is going on. Also be warned this will not be one of these discussions
that are going to be this is the problem and this is the answer.

To kick this off Let’s consider our quarter car model. I’ve stated on a number of different occasions
that this is the basic building block of car suspension analysis. It’s also really simple to quantify.
The quarter car model is presented in Fig-1,

Fig-1 – Quarter car model

Before I dive into the equations of motion let me just be clear about the nomenclature,

xb = Movement of the sprung mass (m)

xw = Movement of the wheel (m)
AR = Road amplitude (m)
f = Frequency of road input (Hz)

We are going to subject our quarter car model to a sinusoidal input of amplitude AR and frequency
f. I am choosing this for two reasons. The first is for simplicity. I want to layout the obvious case
before going into further detail. Also thanks to our good friend Joseph Fourier repeating signals can
be summed as a bunch of sine waves. Consequently this forms a very good starting point.

The equations of motion for our case study here are the following,

m B ⋅ &x&B = − K B ⋅ ( x B − x w ) − C B ⋅ ( x& B − x& w )

(1)
m w ⋅ &x&w = K B ⋅ ( x B − x w ) + C B ⋅ ( x& B − x& w ) − K T ⋅ ( x w − AR ⋅ sin (2πf ⋅ t ))

Just to make this discussion simpler I’m considering the system from it’s equilibrium state. That is
the zero state is with the quarter car model on the ground and at rest. Or to use data engineer’s
parlance we have zeroed it on the ground. It’s the reason I haven’t included the mass terms in
equation (1). Bare with me it will become clear why I have done this in a few paragraphs.

Let’s consider the case what should the damping be if we want the body to remain at the rest
position. That is what should we do to try and duplicate active suspension. This has some rather
interesting ramifications. These are summarised in the following equations,

&x&B = 0 (2)
x B − x w = AR ⋅ sin (2πf ⋅ t )
x& B − x& w = 2πf ⋅ AR ⋅ cos(2πf ⋅ t ) (3)
( )
&x&B − &x&w = − 4π f ⋅ AR ⋅ sin (2πf ⋅ t )
2 2

Equation (3) represents the situation that the measured damper movement is effectively the
suspension system riding the bump profile (here a sine wave) perfectly.

So using equation (1) to (3) we can know solve for what we would like the damping to be. Plugging
in the numbers we can show the damping rate as a function of time should look something like this,
K
c B = − B ⋅ tan (2πf ) (4)
2πf

Hooray! We can have the perfectly damped car. There’s just one small problem, which the
mathematicians amongst will instantly recognise (See trigonometry and trig functions weren’t just
devised as creative intellectual water torture). This small problem I’m going to illustrate
graphically. This is illustrated in Figure 2.
 Fig-2 – Illustration of damper position and the damping rate required for perfect body
control.

What’s happening here that as the bump amplitude hit’s it’s maximum and the damper velocity hits
it’s 0 the damping rate goes infinite. This represents the fact that we are tyring to get the damper to
do what effectively isn’t a damper job. This arises because we are tying to get the damper to hold
position when it has no influence to do so.

However all is not completely lost. We do have some options. Option 1 is to pray to what ever deity
your worship for an infinite damping rate. Some how this might be a prayer that goes unanswered.
Option 2 is to approximate this with a max damping rate and blow it off very quickly. The best way
to approximate this is to plot damping rate vs damping velocity and integrate the damping rate to
get the damper force. (Hint motorsport engineering students and young data engineers this is a job
for you). You should get something that looks like Fig-3,
 Fig-3 – Idealised damping curve.

This could have half a chance there is just one small problem, the high initial damping rate is
effectively spring pre load. As we have discussed in previous articles this can potentially have
disastrous consequences as the car loads up in braking and cornering.

The only pay off of this is it gives a way off looking at a frequency analysis of the data from track
to track and piecemealing damping rates for various damper positions for each particular circuit. It’s
far from perfect but it’s an idea worth investigating.

The other approach we might like to canvas is dispensing with the damper entirely and replace it
with an inerter. Let’s define the inertance as a value in b in kg and place the inerter between the
wheel and the body. In this case the governing equation at the body reduces to,

m B ⋅ &x&B = − K B ⋅ ( x B − x w ) − b ⋅ (&x&B − &x&w ) (5)

Using equation (3) and asking for the body not to move it can be seen that,

KB
b= (6)
(2πf )2
These terms cancel out because the sine term is both in position and acceleration. What is
physically happening is that the inerter is anticipating what is coming and can make the necessary
adjustments in the force.
At face value it would appear we have reached the promised land. However as always the devil is in
the detail. First things first at low frequencies the value of the inerter is going to be rather large,
almost to the point where it is impractical to use. The second thing is that this will work really well
for the specified frequency. There is no guarantee at what happens outside this specified frequency.

What all this illustrates is that passive elements don’t truly replace a full active suspension system. I
personally think the banning of active suspension in most level of motorsports is one of the greatest
tragedies to be visited on our sport. What I would love to be writing about is how we can use active
suspension to truly optimise what a car is capable of. However the powers at be have seen fit to
stick us with what we have.

This also illustrates why race engineers are always playing with damper adjustments. As can be
seen very clearly what works for one condition will not necessarily work for another. This is why
race teams spend money on simulation and other tools so they can get some idea of what to do
when they go from track to track. Also as illustrated by this example sometimes it can be a
challenge to see the forest through the trees.

To put this discussion in it’s proper context, let’s actually review the dual rate approach I discussed
in my first article. This approach actually had its origin in control system analysis, and analysing the
fundamental equations of motion to illicit a desired response. That is heavily damped to control the
body motion and lightly damped to allow the bumps to blow off.

The heart of this approach was to evaluate the natural frequency of the quarter car model and then
use this to evaluate the damping ratio. For the readers convenience I’ll present the two equations
below, which are the heart of the technique,

KB
ω0 = (7)
mB
C B = 2 ⋅ ω 0 ⋅ mB ⋅ ζ (8)

This is the first point of the analysis. Equation (7) specifies the natural frequency in rad/s of the
system, and equation (8) means a damper rate based on the damping ratio that is desired can be
specified. Please note the spring rate is specified in N/m and the mass is specified in kg. What this
means is that we can now specify a damping rate based on what we want the system to do as
opposed to guessing. Also because of the form of equation (8) this can be readily applied for both
the front and the rear. How to choose the damping ratio is illustrated in Fig-4,
 Fig-4: Damping ratio selection guide

For the derivation of this I will refer the reader to my article on Specifying Dampers.

Now that we have reminded ourselves of this method how does this stack up in controlling our
quarter car model rolling down the road. To illustrate this graphically consider Fig-5,

Fig-5 – Comparison of Calculated damping vs using a bilinear model.

For ease of comparison I’ve illustrated the bilinear rate model using a dashed line and the
calculated line as solid. It can be clearly seen that using the bi linear approach while we don’t get
the exact response where after we can get pretty close.

Now before some of you start howling in protest there is a precedent for my little party trick. For
subsonic aircraft the optimum wing planform is an ellipse. The most marked example of this is the
Supermarine Spitfire of World War II fame. However a really easy approximation of this is to do a
straight chord for half the wing span and then at the tips the chord is half the length of the centre of
the wing. To better illustrate this look at Figure 6,

Fig-6: Straight-line approximation of an ellipse

This gets you pretty close performance wise with something that is really easy to manufacture. This
configuration is seen on thousands of light aircraft the world over. The moral of the tale is don’t
underrate a few strategically placed straight lines.

The other advantage of the damping ratio approach is it actually makes it really easy to quantify
what is going on with the car. In my article on advanced aerodynamic analysis I walked the reader
through on State space analysis and the use of eigenvalues and eigenvectors. This involved this
horrible mathematical monster called complex numbers (Well it’s actually not all that horrible it is
actually a mathematician’s fudge to the conundrum of i2 = -1, but the fudge actually worked quite
well) but the use of these complex numbers actually well tell you what the car is going to do.

In closing let’s just reflect on a number of things. First off it’s impossible to get a passive
suspension system to do what an active system can. You can get away with it in certain cases, but in
all honesty you wouldn’t bet your house on it. Secondly our old war horse of using a linear
approximation based on damping ratios can be stretched. Yes it’s not perfect, but sometimes getting
something in the ball park is sometimes good enough. But more importantly, for the time being I’m
turning the discussion over to you.

As I stated in the beginning of this article this will be the first of many articles on this matter. This
article has set the tone by outlying the challenge we face by considering a rather simple example
that illustrates really well what we are up against. Until the next instalment I turn this over to you
the reader to reflect on this and see what you can come up with. If this is the only result that comes
of this then I have succeeded.
Race car damping 2

A number of issues ago I wrote an article on exploring approaches to specify a race car damper.
This article is the second in that series and we shall be following on from where we left off. In
particular in this article we will be exploring the use of damping ratios and what they tells us and
more importantly how we can use it.

Before we get into this let’s recap where we are at with the first article. In the first article I
presented the example of controlling a quarter car mass being subjected to a sinusoidal road input.
In that article we explored what damping would have to be used to control this and what we could
do via an inerter. The conclusion of this was that damping alone wouldn’t get the job done and an
inerter would only work at a specified frequency. However we did leave the article on the note that
we could approximate what we wanted to do by approximating the job with damping ratios. This is
where we will take up the discussion.

As a precursor to what we will discuss let me go through the two schools of damping I have
encountered over my career. The first school dictates all the damper does is control the oscillation
of the wheel and that’s it. The second school regards the damper as an essential element of the setup
of the race car and in addition to controlling the oscillation it is used as a vital tool to control chassis
attitude and control tyre loads. During my career I have varied between both schools of thought and
I trust what we will discuss here might provide some answers to this conundrum.

Let’s talk about who we can characterise dampers through damping ratios. As before we will be
using the ¼ car model approximation to draw our line in the sand. I realise that a real car functions
with a chassis and four contact patches. However bare with me this is a very useful approximation.
To refresh our memories Let’s look at the quarter car model,

Fig-1 quarter car model.

Looking at this we have a system with 4 degrees of freedom. Body movement and velocity and
wheel movement and velocity. To solve this properly we need to start solving eigenvalues and
eigenvectors and this doesn’t lend itself to nice hand calculations.

At this point we need to make some approximations. The approximation that we are going to make
is,

K B << KT
mt << mB (1)

Please note all the rates we are about to quote are going to be in wheel rates. The astute reader will
instantly recognise that this is perfectly valid for a road car, but we could run into trouble with an
open wheeler applications where the spring rate of the chassis are of the same magnitude as the tyre,
or in a V8 Supercar or NASCAR where the unsprung masses are quite high. However bare with me.
What we are constructing here is a rough rule of thumb. You’ll see its power as we carry on.

In this case the governing equations of the sprung mass reduce to,

″ ′
mB ⋅ x B = − K B ⋅ xB − C B ⋅ x B (2)

What this means in layman’s terms is that the acceleration of the sprung mass is the sum of the
spring rate of the damper times the damper movement and the damping rate times the damper
velocity. The damping rate is the slope of the damping curve at that particular damping velocity. Let
me illustrate this in Figure 2,

Fig-2 – Evaluating a damper slope from a typical force vs damping curve

The beautiful thing about equation (2) is that we can now do a laplace transform on this. The end
results are shown below,

CB K
0 = s2 + ⋅s + B (3)
mB mB

If we compare this to the ideal second order system with natural frequency ω 0 and damping ratio
ζ it is seen that,

0 = s 2 + 2ζ ⋅ ω0 ⋅ s + ω0 (4)

When equations (3) to (4) are compared we know have the tools to specify our desired damping rate
in terms of damping ratios and natural frequency,

KB
ω0 = (5)
mB
CB = 2 ⋅ ω0 ⋅ mB ⋅ ζ
CB
ζ = (6)
2 ⋅ ω 0 ⋅ mB

This is the first point of the analysis. Equation (5) specifies the natural frequency in rad/s of the
system (To convert to Hz divide by approximately 6.283185), and equation (6) means a damper rate
based on the damping ratio that is desired can be specified. It’s at this point when calculating you
need to be really precise with your units. Everything here needs to be in strict SI units. That is
masses in kg, Spring rate in N/m and damping rate in N/m/s. Also all rates are in Wheel rates not at
the damper. I make absolutely no apologies for this. Besides as far as I am concerned any
measurement that has the slug as it’s measure of density is intellectually flawed.

Before we get into this in further detail why don’t we do a worked example of this. For this
particular case let’s outline the following variables,
 Table-1 – Values for ¼ car example

Item Value
Motion Ratio (Damper/Wheel) 1
Spring Rate 175000 N/m (1000lbf/in)
Damping Rate 21333 N/m/s
¼ car mass 157 kg

The damping rate was actually calculated from Fig-2 and I would invite the interested reader to

repeat the calculation. Let’s now work through the calculation of the natural frequency and the

damping ratio,

MR 2 ⋅ K B 12 ⋅ 175000
ω= = = 33.386rad / s
mB 157
(7)
MR ⋅ C B
2
1 ⋅ 21333
2
ζ = = = 2.03
2 ⋅ mb ⋅ ω 0 2 ⋅ 157 ⋅ 33.386

As far as I am concerned the real power of this is that it gives us a powerful way of non
dimensionalising what is going on with a damper. If you characterise a damper this way the
moment you do a spring change you can specify a damper, which in theory has identical
characteristics. I should also add this applies both in bump and rebound. Obviously when back
calculating for rebound when re working your numbers remember to multiply the slope by –1, if
using a standard force vs damper curve, that shows bump positive and rebound negative.

Now that we have gone to all this trouble to calculate damping ratio what does all this actually
mean. To get to the heart of the matter we need to return to what our quarter car is going to do when
we apply a step input to it. The damping ratios will actually tell you what the car is going to do.
This is illustrated in Fig-3,
 Fig-3 – Second order system response to a step input.

Reviewing this graph is very interesting. At low damping ratios the ¼ car model is decidedly
underdamped and it takes a long time to return to an equilibrium state. However think about what
this is ideal for. If we are hitting a bumpy bit of the circuit we actually want the car to do this. When
the damping ratios hit about 0.5 – 0.7 the car’s oscillation is remarkable decreased. Not so great
with dealing with a bumps but ideal for controlling the body when we are pitching and rolling.
Damping ratios greater than this are ideally placed to force the car to almost act like a spring as we
discussed in our previous articles. However be warned there are some consequences for over doing
it.

On the back of this I came up with a rough damper guide that I actually outlined in my first article.
Now that we have discussed this in further detail let me outline it in further detail.

Table – 2 – Rough outline to damping ratios

Damping Ratio Range What this applies to

0.3 – 0.4 Ideal for filtering out bumps
0.5 – 1.0 This deals with body control.
1.0 + This deals with extreme body control/driving
temperature into the tyres.

Armed with this knowledge let’s look at the first damper curve we saw in Fig-2. Assuming the
same numbers from Table – 1, I’ll work out the damping ratios for both bump and rebound. This is
presented in Table 3
 Table-3 – Damping ratios for Damper presented in Fig-2

Velocity (mm/s) Damping ratio in bump Damping ratio in rebound

0 1.24 0.95
13 2.03 0.6
25 0.616 0.707
38 0.175 0.31
50 0.167 0.286
63 0.174 0.31

Table 3 presents some enlightening insights into what this damper is trying to do. First things first
the damping ratios from 0 tell me immediately this is a high downforce car. The high damping
ratios are tell tale signs this is a high downforce car. The high damping ratios immediately suggest
that body control is paramount. Looking at the bump at 13mm/s the damping ratio jumps to 2.03.
This indicates the damper engineer is trying to give some feel to the car as well as load the tyres.
Beyond this range the dampers blow off to a low ratio to allow the car to ride the bumps. In rebound
from 13 – 25 mm/s the damping ratio is 0.7. This tells me body control is still paramount. Beyond
the damping ratios blow off to 0.3. This tells me this is designed for the bumps.

I think the reader is fast starting to get the idea that damping ratio is not just a useful parameter but
it can tell you an awful lot about what you want both the chassis and the tyre to do. As a rough rule
of thumb the higher the damping ratio the better it is for both controlling the chassis and putting
temperature into the tyre. The lower it is the better it is for riding bumps.

However the question has to be asked how do we determine what we want and need. One is
experience and a bit of informed intuition. Let me illustrate with an example say you have a low
downforce car going over a circuit with a lot of undulations. In general terms for a car without a lot
of downforce you would actually like lowish damping numbers in rebound – Say damping ratios in
the order of 0.3 – 0.5. However let’s just say your rounding out a turn with an undulation. It could
be desirable to actually have a higher damping ratio in rebound to keep the car connected with the
ground? I’m not saying this is what you need to do I’m just asking the question so you can think
about it.

The second method is to use simulation software to aid us in determining this. Indeed it was this
very question of evaluating the appropriate dampers that lead to the creation of ChassisSim. Indeed
when I was evaluating what lap time algorithm to use it was this reason that I dismissed the pseudo
static lap time approach. To illustrate what I am taking about let me evaluate too totally different
types of dampers for the front of an F3 car. The results of the simulation is shown below in Fig – 4,
 Fig-4 – Simulation results for two totally different dampers

This is a bit of an obvious example because this was a very bumpy circuit and I set the second
damper very stiff in both bump and rebound, but this none the less shows how you can use
simulation to look at different configurations and see the impact they can have on vehicle
performance. For example in this trace we can see very quickly the impact this has had on loads,
corner speed and temperatures (The red is the baseline and the black is the damper change). I would
wager the loss in temperature is due to the fact the apex corner speed has dropped by 2km/hr.

In closing then while we haven’t achieved the grand unified theory of race car damping we
nonetheless have delved into how useful the damping ratio can be a very effective tool. Not only
does it non dimensionalise the way we can look at damping but we can use it too force the
behaviour of the race car to get it too what we wanted too. Is this a magic bullet? No it isn’t
however if properly utilised it is a very powerful tool to investigate what is going on with the car. It
will also give us some good for thought for the next instalment in this series on race car damping.
Damper Analysis – Using State Space Analysis to really get the most out of your
dampers – PART – 1

In the last article in my damper series I discussed the use of damping ratios to gain insight into what
to do with our dampers. In summary we used a very simple quarter car model as a tool to deduce
damping ratios. We then used this knowledge of damping ratios to construct some rough rules of
thumbs as to where to go with our damping.

In this article what we are going to be doing is introducing state space analysis to take these
approximations to the next level. What is state space analysis you might ask? I actually introduced
this concept in my article on how to analyse aero maps about 2 years ago. What state space analysis
is that it is a tool that represents the equations of motion in matrix form and then we look at
eigenvalues and eigenvectors to divine the results.

The reason we are going to be going to all this trouble is that while damping ratios and the quarter
car model gives us a great start point ultimately they run of steam. They run out of steam due to the
following reasons,
• Ultimately a ¼ car model gives a picture of a car going up and down. In reality a car is a 3D
object, which pitches, rolls and yaws and has these 4 tyres connected to it.
• With the quarter car model introducing aero effects can be somewhat cumbersome.
• Now that inerters have turned up on the scene, we really need to do a through analysis and
see what really affects what.

Also what I’m about to introduce is the bread and butter of what control engineers do for a living
and ultimately that’s what we do as race and data engineers. Furthermore this concept was
introduced to me be one of my mentors in vehicle dynamics, Dr Gregory Chamitoff. Greg went on
to work on the international space station and become an astronaut. However the techniques he
taught me was like someone turning the light switch on when it came to vehicle dynamics and I
trust you’ll get something out of it.

Does this render what we have done with damping ratios and the quarter car model obsolete?
Absolutely not. The thing that I love about the quarter car model approximation and damping ratios
is that it gives you a hands on feel for what the numbers look like and how this fits in. In that
regards it’s the first stepping stone to what we are about to do. Let me put this way. Trying to do
state space analysis without doing the quarter car approximation is a little bit like getting a Phd in
mathematics without having the ability to add up.

Let me give you a warning straight up that this is a high level article with an awful lot of complex
maths involved and to cover everything I might have to miss a few steps. I will do my best to
explain it as well as I can, but if you struggle with this that’s OK. In my last year of my science
degree I just missed out on a High Distinction in pure maths, and when I was first introduced to
complex numbers when I was a teenager it took me two bits of the cherry to get it. So if you get a
little lost with the maths your in very good company.
Also to give the reader the heads up I’m going to break this up into two articles. Trying to show you
theory and application in one article would be way too ambitious. Also with what we are about to
discuss it is better to let the concepts and ideas digest for a month anyway.

To introduce state space analysis we are going to return to the ¼ car model. We are starting on this
because we only have 4 variables to worry about and it forms a good introduction to introduce the
ideas of state space analysis. To start this discussion let’s consider a ¼ car with the following
properties,

Table-1 – ¼ car modelling parameters and values

Parameter Symbol Unit Value

Sprung Mass ms kg 300
Spring rate Kb N/m 60 000
Damping rate Cb N/(m/s) 5000
Tyre Spring rate Kw N/m 300 000
Tyre mass mt kg 20
Sprung mass displacement Xb m
Sprung mass velocity x’b m/s
Sprung mass displacement Xw m
Sprung mass velocity x’w m/s

Let me make a few housekeeping observations. First things first all spring and damping rates are in
wheel rates. This is mr*mr*spring or damping rate squared. Also if you’re going to have any hope
with this you have to use SI units. Don’t even bother using imperial units you won’t stand a snow
flakes chance in hell. To keep the discussion simple I’m going to keep all spring and damper rates
strictly linear.

Now that we have defined our nomenclature what do the equations of motion look like? This is
outlined in equation (1),

m B ⋅ &x&B = − K B ⋅ ( x B − x w ) − C B ⋅ ( x& B − x& w )

(1)
m w ⋅ &x&w = K B ⋅ ( x B − x w ) + C B ⋅ ( x& B − x& w ) − K T ⋅ x w

The derivation of this I presented in my first article on race car damping. The only difference I’ve
presented here is I have removed the road input. So at this point you might be thinking so what?
This is where we pull our first rabbit out of the hat. The trick comes from looking at the
mathematically obvious,

∂x& b
&x&b = (2)
∂t
What this says is the acceleration of the body is the time derivative of the velocity. Ditto for the
unsprung mass.

So to make our life a bit easier let’s down the following vector x as,

⎡ x& b ⎤ ⎡ x1 ⎤
⎢ x ⎥ ⎢x ⎥
x = ⎢ b ⎥ = ⎢ 2⎥ (3)
⎢ x& w ⎥ ⎢ x3 ⎥
⎢ ⎥ ⎢ ⎥
⎣ xw ⎦ ⎣ x4 ⎦

We can now express equation (1) as the following,

m B ⋅ x&1 = − K B ⋅ ( x 2 − x 4 ) − C B ⋅ (x1 − x3 )
x& 2 = x1
(4)
m w ⋅ x& 3 = K B ⋅ ( x 2 − x 4 ) + C B ⋅ ( x1 − x3 ) − K T ⋅ x 4
x& 4 = x3

The subscripts refer to the individual elements of the vector x. Remember what I said about keeping
the spring and damper rates strictly linear. What this means is that we can represent equation (4) as
a matrix. This matrix looks like this,
⎡ CB KB CB KB ⎤
−
⎢ m − ⎥
mB mB mB
⎢ B
⎥
′ ⎢ 1 0 0 0 ⎥
x = C
⎢ B KB CB K B + KT ⎥ ⋅ x (5)
− −
⎢ mw mw mw mw ⎥
⎢ 0 ⎥
⎣ 0 1 0 ⎦
= A⋅x

Equation (5) is what is referred to as the state space matrix and the construction of this matrix is the
first step in using state space analysis.

The next step is to determine the eigenvalues and eigenvectors of the state space matrix because it
will tell you everything you’ll ever need to know about the way this system responds to the world.
The eigenvalues determine the time response of the system, and the eigenvectors determine in what
ratios the various states move in relation to each other and these are associated with the
corresponding eigenvalue. Mathematically eigenvalues are determined by,

det(A − λI ) = 0 (6)
In this case we are finding the solution for a big polynomial that is equal to 0. The number of
eigenvalues is dictated too by the size of the state space matrix. This reflects the fact that if we have
4 variables that we are modelling, then that will represent 4 time responses. The eigenvectors are
given by the equation,

(A − λi I ) ⋅ x i
=0 (7)
You have two options to solve this. You can do it by hand, which beyond 4 variables can get quite
tricky, or you can use programs such as Matlab, Maple or mathematica.

I realise that we are now pushing into some pretty advanced stuff, but we are doing this because this
is the most effective way we have of seeing what our system is actually up to. I do realise here we
are talking strong language, violence and that most dreaded thing of all complex numbers. However
we are going to all this trouble because every eigenvalue will tell you how the system responds to
an input and the corresponding eigenvector will show what this will look like in the real world. This
is absolutely vital in nailing down what the system is going to do. Without these techniques the F-
16 would never fly, and we are about to see it’s going to reveal a wealth of information about what
the car is going to do.

To further quantify this let’s just say for the sake of the argument that with our state space model
we have eigenvalues of values -σ1, -σ2 and a complex conjugate pair σ +/- ωdi . Then depending on
the eigenvectors the response of say the sprung mass to a step input (such as body acceleration etc)
can be represented by,

⋅ sin(ω d ⋅ t + φ )
−σ ⋅t
y = A1e −σ1 ⋅t + A2 e −σ 2 ⋅t + A3 e (8)

To put this into practice let’s see what our ¼ car model is doing. Using equation (5), A the state
space matrix is given by,
⎡− 16.7 − 200 16.7 200 ⎤
⎢ 1 0 0 0 ⎥⎥
A=⎢ (9)
⎢ 250 3000 − 250 − 18000⎥
⎢ ⎥
⎣ 0 0 1 0 ⎦

The associated eigenvalues are given by,

⎡ − 164.7 ⎤
⎢ − 88.4 ⎥
λ=⎢ ⎥ (10)
⎢− 6.8 + 12.6 ⋅ i ⎥
⎢ ⎥
⎣ − 6.8 − 12.6 ⋅ i ⎦

For convenience I’ll write the eigenvectors in matrix form, with column 1 associated with
eigenvector 1 and so on. This is shown below,
 ⎡- 0.103 - 0.191 0.976 0.976 ⎤
⎢ 0
⎢ 0.002 - 0.032 − 0.0599 ⋅ i - 0.032 + 0.0599 ⋅ i ⎥⎥
EigenVector = (11)
⎢ 0.995 0.981 0.110 + 0.170 ⋅ i 0.110 − 0.170 ⋅ i ⎥
⎢ ⎥
⎣- 0.006 - 0.011 0.007 − 0.012 ⋅ i 0.007 + 0.012 ⋅ i ⎦

So what does all this mean? What we are seeing is all the fundamental modes this system will
respond to any input applied to it. The eigenvalues, when they are real give you the time constant or
period the mode will act over and when it is complex it will tell you the natural frequency of that
particular response. The eigenvector tells you what modes are affected. So in this particular
example, the first eigenvalue acts primarily on the unsprung wheel velocity. This is just as well,
because with a value of 164.7, this dies of fast. In the second eigenvalue, it acts primarily on the
wheel velocity and a little bit of the sprung mass. The final complex modes act primarily on the
sprung mass velocity. Also due to the fact they are much smaller in magnitude this will dominate
the response.

The final sprung mass mode is what is most comparable with our approximation from the last
damper article. Recall that we ignored the unsprung mass, and from this we started our
approximation. Let’s see how this stack up from what we have seen above. Before we do this, we
need to introduce another mathematical party trick. For a complex mode, it’s natural frequency and
damping ratio is given by,

σ ± i ⋅ ω d = ζω n ± i ⋅ ω n ⋅ 1 − ω n 2
ωn = σ 2 + ωd 2 (12)
2
⎛ω ⎞
ζ = 1 − ⎜⎜ d ⎟⎟
⎝ ωn ⎠
I would refer the reader to my second damper article for the approximation formula. However what
is very interesting is comparing the two results. This is shown in Table 2.

Table – 2 – Comparing Eigenvalue to ¼ car approximation

Value ¼ car approximation Eigenvalue

ωn (Natural frequency) 14.14 rad/s 14.3 rad/s
ζ (Damping ratio) 0.589 0.47

As can be seen our ¼ car approximation got the Natural frequency really close. However we have
overestimated the damping ratio. This is not to be unexpected, because with the spring rate is still
less than the tyre spring rate, the tyre will still be moving. This is not to be unexpected. However
before we all lose too much sleep over this just remember the ¼ car technique was a quick and dirty
approximation to get us in the ball park. While the damping ratio estimation isn’t ideal at least it put
us in the ball park. This is what it is designed too.
To continue this discussion let’s now consider some numbers for an F3 car. The variables we are
considering are shown in Table – 3,

Table – 3 – Typical F3 quarter car model values

Parameter Symbol Unit Value

Sprung Mass ms kg 100
Spring rate Kb N/m 120 000
Damping rate Cb N/(m/s) 12000/5000
Tyre Spring rate Kw N/m 240000
Tyre mass mt kg 20

You will notice I’ve shown two values, the lower value corresponds to high speed damping, and the
other corresponds to low speed damping. The eigenvalues for both of these cases is shown in Table
4 below,

Table – 4 – Eigenvalues for the low speed and high speed damping case.

Case Low Speed Damping High Speed Damping

Eigenvalue ωn /ζ Eigenvalue ωn /ζ
1 -695.3 695.3 -225.3 225.3/inf
2 -6.86 + 42.9i 43.4/0.157 -52.5 52.5/inf
3 -6.86 - 42.9i 43.4/0.157 -11.1 + 33.1i 34.9/0.32
4 -11 11 -11.1 - 33.1i 34.9/0.32
¼ car approximation N/a 34.6/1.73 N/a 34.6/0.72

Table – 4 makes for some fascinating reading. Remember we discussed about the ¼ car
approximation, that it was valid when the spring rate was much less than the tyre spring rate. The
F3 case is where the ¼ car approximation runs out of steam, and on a first look at the results in
Table 4 you would definitely have to say so. However with all things the devil is in the detail.
Remember the higher the damping ratio the more it emphasis body control.

Let’s review the high speed damping results. On a first cut comparing to the ¼ car approximation
we did a good job of picking the natural frequency but the damping ratio leaves a little bit to be
desired. Or does it? Remember as a rough rule of thumb the mode with the lowest natural frequency
will dominate the time response. However looking at table 4, we have an exponential mode at a
frequency of 52.5 rad/s. This is right next door to our complex mode. What this means is that this
mode will effect what we are going to see with the complex mode. Yes there is absolutely no way
we’ll hit the damping response of say 0.72 but it certainly won’t be totally underdamped. To
illustrate this consider Fig-1 that is our simulated quarter car subjected to a step input. What is
presented here is the high speed case. The y axis is body position in m, the x axis is time.
 Fig-1 - Time response of high speed damping to a step input.

As can be seen the initial peak is greater than what we would expect, but it does die down very fast.

Let’s now move on to the low speed result. The ¼ car approximation said we should have a natural
frequency of 34.6 rad/s and a damping ratio of 1.73. So effectively we are forcing this into a very
damped response. As we discussed last time this is great for body control and/or forcing
temperature into the tyres.

The eigenvalues for the low speed result make for interesting reading. The dominant mode is a
straight exponential response at 11 rad/s. On the surface we seemed to have done our job. However
this has come at a price. Note the next mode is at a frequency of 43.4 rad/s and it is chronically
underdamped. What this means is that there is going to be some wobble to our results. Let me
illustrate what this would look like in Fig-2,
 Fig-2 – Time response of low speed damping to a step input.

As before the y-axis is body position in m, the x-axis is time. Also the last mode has now been
blown out from a frequency of 225.3 rad/s to nearly 695.3 rad/s. This is going to make this rather
interesting when we go over bumps. The moral of the tale is this; you can use high damping rates to
give the driver good feel, and almost approximate spring rate. However it comes at a cost.

The astute reader will recognise up to know we’ve been considering linear springs and damper, but
how do we incorporate bump rubbers and non-linear dampers? Recall in my last article I showed
you how to derive damping rates from a non-linear bump curve. Let me remind you of the
technique in Fig-3,
 Fig-3 – Deducing instantaneous damping rates

The technique for deducing instantaneous spring rates is identical to the damping rate. You simply
plot spring force vs deflection and the rates that you calculate is what you put into your state space
matrix as illustrated in equation (5)

I’m hoping your starting to appreciate why state space analysis is such a powerful tool and why it’s
the next logical step from our damping ratio calculations. In the next article we are going to take
this discussion from the quarter car model to the bicycle model, and we’ll see the effects that aero
and inerters have on the car and how we can use our knowledge of state space matricies and
eigenvalues to determine where to go in our damper selections. Remember at the end of the day this
is a thinking man’s business.
Racecar Damping 4 – Applying State Space analysis to the Race car – PART 2

In our last instalment of race car damping we introduced and discussed the concept of using state
space analysis in choosing damping for the race car. We discussed how to formulate the state space
matrix, how it compared to the ¼ car approximation and we played with the ¼ car model to see how
things compared.

In this instalment we will be taking what we learnt in part 1 and applying it to a bicycle model of a
current spec F3 car. Were considering the case of the bicycle model of the F3 car so we can see the
effects that aerodynamics has on the dynamics of the race car. It’s also a very effective way of
introducing inerters as well. The F3 car is a very interesting case because it marks the crossover
where aerodynamics has a significant impact on vehicle performance. Consequently for our
discussion it is perfect.

So that we are crystal clear let us outline the parameters of our F3 car. This is summarised in Table
1

Table – 1 – Parameters for an F3 car.

Parameter Symbol Value Unit

Total mass Mt 550 Kg
Total unsprung mass at the front Mtf 20 Kg
Total unsprung mass at the rear Mtr 30 Kg
Front spring rate Kf 120000 N/m
Front Damper rate Cf 12000/5000 N/m/s
Front tyre spring rate Ktf 240000 N/m
Rear spring rate Kr 100000 N/m
Rear Damper rate Cr 12000/5000 N/m/s
Rear tyre spring rate Ktr 260000 N/m
Wheelbase Wb 2.7 M
Front weight distribution Wdf 41%
Distance from front axle to c.g a 1.593 m
Distance from rear axle to c.g b 1.107 m
Approximate CLA 2.5

To further clarify our discussion we are considering a twin spring car front and rear, and I have just
quoted the rates at the corner. In reality in the bicycle model I’m doubling these numbers, but for
clarity I’ve quoted at corner values. These numbers are also in wheel rates.

To save ourselves a bit of time I’m going to quote the linearised state space matrix. This is
summarised in equation (1),
 (1)

The reader should recognise equation (1) from my articles on how to analyse aeromaps that was
published about 2 years ago. I leave the derivation of equation (1) to the reader ( In reality it
actually isn’t that hard, it’s just an extension of what we did in part 1).

However one area I do want to go over in a little detail is the derivation of the aero derivatives. The
reader will recall I laid out the definition of the aero derivatives. I will recap these in equation (2),

δFz ∂
= (C L A) ⋅ 1 2 ⋅ ρ ⋅ V 2
δz ∂z
(2)
δFz ∂
= (C L A) ⋅ 1 2 ⋅ ρ ⋅ V 2
δθ ∂θ
δM ⎡ δ δ ⎤
= wb ⋅ 1 ⋅ ρ ⋅ V 2 ⋅ ⎢(wdf − ab ) ⋅ (C L A) − (ab ) ⋅ C L A⎥
δz 2 ⎣ δz δz ⎦
δM δ
⎡
= wb ⋅ 1 ⋅ ρ ⋅ V 2 ⋅ ⎢(wdf − ab ) ⋅ (C L A) − δ (ab ) ⋅ C L A⎤⎥
δθ 2 ⎣ δθ δθ ⎦

At the time I mentioned you had to convert aeromaps from functions of front and rear ride height to
functions of heave (z) and pitch angle (θ). Well in the process of researching this article I found a
shortcut, and I’m kicking myself I didn’t think of it 2 years ago.

The shortcut uses the chain rule of differentiation of several variables. The ace in the hole is the
following,

rh f = rh f 0 − ( z − a ⋅ θ )
rhr = rhr 0 − ( z + b ⋅ θ )
(3)
a = (1 − wdf ) ⋅ wb
b = wdf ⋅ wb
So what we need to do is to construct partial derivatives of front and rear ride height as functions of
heave and pitch angle. Using the chain rule of several variables it can be shown,
 ∂ ⎛ ⎞
(C L A) = −⎜⎜ ∂ (C L A) + ∂ (C L A)⎟⎟
∂z ⎝ ∂rh f ∂rhr ⎠
∂
(C L A) = a ⋅ ∂ (C L A) − b ⋅ ∂ (C L A)
∂θ ∂rh f ∂rhr
(4)
∂ ⎛ ⎞
(ab ) = −⎜⎜ ∂ (ab ) + ∂ (ab )⎟⎟
∂z ⎝ ∂rh f ∂rhr ⎠
∂
(ab ) = a ⋅ ∂ (ab ) − b ⋅ ∂ (ab )
∂θ ∂rh f ∂rhr

The power of equation (4) is that we can use our existing aeromaps to readily construct our aero
derivatives. That’s pretty cool and I leave the derivation of equation (4) to the interested reader.

Armed with this knowledge I constructed state space matricies for the high speed damping case.
This included the case without and with downforce. The results were very interesting. The non
downforce results are shown in Table –2. For brevity I have rolled in the complex conjugate pairs,

Table – 2 Eigenvalues and eigenvectors for the non downforce case.

Mode 1 2 3 4 5 6
Eigenvalue -503.7 -302.7 -55.3 -15.5+46i -35.1 -11+26i
Eigenvectors
Z’ 0.04 -0.08 -0.46 0.09-0.09i -0.01 0.829
Z 0 0 0.01 0 0 -0.01-0.027i
θ’ -0.07 -0.065 -0.09 -0.45+0.36i -0.85 0.21-0.06i
θ 0 0 0 0.01+0.01i 0.02 0
ztf’ -0.99 0.1527 0.44 0.721 -0.38 0.14+0.18i
ztf’ 0 0 -0.01 0 0.01 0
ztr’ 0.06 0.98 0.76 -0.36 – 0.05i 0.35 0.32+0.33i
ztr’ 0 0 -0.01 0 -0.01 -0.01i

As we can see there are six primary modes. Modes 1 and 2 act primarily on the velocities of the
unsprung mass. These die off very quickly which is a good thing. Modes 3 and 4 have very similar
frequencies (approximately 55.3 rad/s) and are combined modes that effect heave, pitch angle and
the unsprung mass. Modes 5 and 6 dominate the time response and have frequencies approximately
35 rad/s, and act primarily on the sprung mass modes of heave and pitch. The first thing to point out
is we have taken up where we left off with our ¼ car example from part 1. We saw very similar
trends, but now we have added more detail.

The downforce results are presented in Table 3. The eigenvectors didn’t change that much.
However the eigenvalues did and the results are shown in Table 3
 Table – 3 Eigenvalues for the downforce case @ 200 km/h.

Mode 1 2 3 4 5 6
Eigenvalue -503.7 -302.7 -55.6 -16.9+44i -32.4 -10.8+27.2i
Eigenvectors
Z’ -0.04 -0.08 -0.46 0.075-0.09i 0.03 0.827
Z 0 0 0.01 0 0 -0.01-0.027i
θ’ 0.07 -0.065 -0.09 -0.44+0.38i -0.91 0.19-0.06i
θ 0 0 0 0.01+0.01i 0.03 0
ztf’ 0.99 0.1527 0.44 0.712 -0.30 0.15+0.19i
ztf’ 0 0 -0.01 0 0.01 0
ztr’ -0.06 0.98 0.76 -0.36 – 0.05i 0.285 0.33+0.32i
ztr’ 0 0 -0.01 0 -0.01 -0.01i

Apart from the eigenvector for mode 1 the rest of them are very similar to the non downforce case.
The reader will note that mode 1 is inversed to what it was in Table 2. However the magnitudes are
almost identical. This means this mode is still acting primarily on the same way, which is primarily
on the front unsprung mass, it’s just acting in the opposite way. However this is all a bit academic
as this mode dies off very fast.

In terms of the eigenvalues as we can see modes 1 – 3 didn’t change that much. Mode 4 became a
bit more damped, the frequency of mode 6 dropped a bit but mode 5 dropped by nearly 3 rad/s.
What this indicates is that when ever a mode drops a natural frequency it indicates the system has
been destabilised. This is exactly what we would expect a ride height sensitive aero car to do. Albeit
in this case while this is something we have to deal with, it’s certainly not something you’d be using
sleep over.

However what Tables 2 and 3 tell us some general trends, that the more aero sensitive a car is the
more you have to stabilise it. This situation is illustrated in this pole zero diagram as illustrated in
Fig 1.
 Fig-1 – Pole zero diagram as more downforce is applied.

As you can see the more downforce that is applied the less stable the more dominant mode is. This
is the reason sports prototypes can’t be run below a minimum ride height. They experience
something called proposing where the chassis rocks back and forth, and this is the extreme case
illustrated in Fig 1.

At this point you might be thinking this is rather entertaining but so what? Remember the
eigenvalues returned by the state space give you the natural frequency and damping ratio. For the
reader’s convenience let me remind you of the formulation,

σ ± i ⋅ ω d = ζω n ± i ⋅ ω n ⋅ 1 − ω n 2
ωn = σ 2 + ωd 2 (5)
2
⎛ω ⎞
ζ = 1 − ⎜⎜ d ⎟⎟
⎝ ωn ⎠

Remember what I said about the ¼ car model. When it comes to aero and inerters it runs out of
steam. Yet using the bicycle model, we have captured exactly what is going and we can see what it
is up to. Once we know the effects the aero is having on the car for a given setup, then we can
change the spring and damper settings to deal with it. This is where equation (5) is such a powerful
tool because we can nail this down with far greater clarity then what we could do with the ¼ car
approximation.
The next real powerful application of state space techniques is considering the effect of inerters.
Some of you may recall an article I did about a year and a half ago where I used state space analysis
to analyse and play the devils advocate on inerters. I have no desire to repeat word for word what I
said in that article, but to help us with inerters I used the concept of an inertance matrix. The
formulation was,

x& = (I _ inert ) A ⋅ x
−1
(6)

Here A was the typical state space equation as outlined in equation (1), and by choosing values of
inerters we could construct the inertance matrix. Once we have equation (6) we can solve for the
eigenvalues and eigenvectors and away we go. This is something we simply couldn’t have done
with the ¼ car approximation.

With all of this fancy mathematical footwork I have presented the next question that has to be asked
is where does this all fit in. To that end I have the following guide on what to do to specify dampers
for the race car.
• Start by using the ¼ car approximation technique we discussed in Part 2 of this series. Use
the damper guide but just remember you can go a bit over. Remember the powerful thing
about this is it is a quick hand calculation, which you can readily put into a setup sheet.
• Once you have you start points as specified by the ¼ car approximation, then you can move
on to state space analysis. If you need something quick and dirty use the quarter car. If you
need to incorporate aero and inerters, use the bicycle model, and if you need to incorporate
roll use the full beam pogo stick model. Remember don’t be a hero, use Matlab, maple etc to
do the eigenvalue, eigenvector stuff for you.
• Once you have refined your choices use simulation software like ChassisSim extensively to
test the results. This is the major reason I used a transient simulation engine for ChassisSim
so make good use of it.
• Finally always validate at the track.

Remember treat this as a guide and not as something in the same league as the Ten Commandments.
Last time I looked I didn’t have a direct line to any religious deity. However if you go through this
process I guarantee it will get you pretty close.

In closing then the subject of race car damping is incredibly vast and complex and this article is
certainly not the last word on the subject. I can tell you right now anybody who thinks they now it
all on race car damping should be shot on site for their own safety. However I trust what I have
presented here with the use of the ¼ car approximation and the use of eigenvalues and eigenvectors
provides a path through the forest. If that is all I have done then I have succeeded.