Design Procedure

Steady-State Design
The steady-state requirements depend on the system type (the number of poles at the origin of
the Loop Transfer Function L(s) = C(s)G(s)) and on the gain Kg = lim s → 0sg L(s). Since the
Plant Transfer function G(s) is fixed, those requirements can be translated on suitable choices of
the values KC and gc . In order to proceed systematically with the design of C(s) we can express
it as C(s) = CSS (s)CT (s) where CSS (s) = sKgCC is the steady-state controller and CT (s) =
′ ′
NC (s) N (s)
′
DC (s)
, lim s → 0 DC′ (s) = 1 is the transient controller. To properly meet the requirements we need
C
to shape CSS to:

1. Add a suitable number of poles at the origin to get the required value of g (system type)
2. Tune the value of KC in order to get the required value of Kg .
How do we choose the sign of KC to ensure system stability?

• If the plant transfer function G(s) does not include poles and/or zeros with strictly positive
real part, then the sign of Kc is chosen so that Kg = KG · KC > 0
• If the plant transfer function G(s) includes poles and/or zeros with strictly positive real part,
then the sign of KC is chosen according to a conceptual procedure.
Conceptual procedure:

′
1. Draw the Nyquist diagram of the L (s) = CSS (s)G(s) using the sign of KC so that Kg =
KG · KC > 0
′
2. If L (s) leads to a stable feedback system, then this is the correct sign choice, provided that
′
CT (s) will be designed to avoid significant modifications of the frequency response of L (s)
near the critical point
′
3. Else if L (s) leads to an unstable feedback system, discuss whether a suitable choice of CT (s)
may be able to stabilize the feedback system (e.g. through a phase lead action) around the
critical point
4. If the previous steps fail, repeat the procedure changing the sign of KC to verify that it’s the
correct choice

Transient Design
Transient requirements of a feedback system, are defined considering the controlled output response
y(t) when the reference signal r(t) is a step function, because it introduces a sudden change in the
desired behavior of the controlled output causing critical solicitations during the transient phase.
In this case, the following indices are used to define the transient performance of a feedback control
system for the time response

• Maximum overshoot ŝ → accuracy

• Rise time tr → trigger off quickness.
• Settling time ts,α% → extinction quickness.

These can be translated into Frequency Response requirements

• Resonant Peak of the complementary sensitivity function TP

• Resonant Peak of the sensitivity function SP
• Crossover frequency ωc

1
Using a graphical procedure we can obtain the minimum dampening coefficient ζ from the
requirement on the maximum value of the overshoot ŝ. From the dampening coefficient, using the
formulas or the graphical approach we can get the desired values of TP and SP .

Remember that the obtained values of TP and SP are not in dB and need to be converted
before using the functions SGrid and TGrid for the plots of the Constant Magnitude
Loci

The rise time tr and settling time ts,α% requirements can be translated in a crossover frequency
ωc requirement. Using the formulas or the graphical procedure, we get two different values
for ωc . The desired crossover frequency is the maximum between the two found values ωc,des =
max(ωc,tr , ωc,ts,α% )

The requirements have been obtained assuming that T (s) is exactly described by second
order prototype model. However, this is hardly the case, and we will need to evaluate if the
requirements have been met through simulation

Transient Controller Design

After obtaining the desired values of TP ,SP and ωc but before the design of the transient controller
CT we must do this preliminary steps:

′
1. Consider the Loop transfer function L (s) = CSS (s)G(s) obtained during the steady-state
′
design. Plot the frequency response L (jω) on the Nichols Plane
2. Mark the point corresponding to ωc,des
3. Plot the constant magnitude loci TP and SP
′
We need to shape L (s) in order for the point corresponding to ωc,des to be tangent to the constant
magnitude loci and to be on the x-axis (y = 0dB). We have multiple ways to approach this

Lead Network
1+ ωs
CD (s) = 1+ m
D
s , ωD > 0, mD > 1 The lead network introduces a desired phase lead and a
D ωD
possibly undesired magnitude increase. The greater mD the greater the phase lead. Basic
guidelines for design:

• Quantify the amount of the phase lead ∆ϕ needed at ωc,des in order to shift the value of
′
̸ L (jωc,des ) outside the ”influence” of the constant magnitude loci
• mD is is choosen on the basis of the required value of ∆ϕ
• ωD is fixed to obtain that the phase lead ∆ϕ occurs exactly at ωc,des
We use the Universal Lead Network Diagrams. It is strongly suggested to use multiple lead
networks if the required phase lead is greater than 60°

Lag Network
1+ m sω
CI (s) = I
1+ ωs
I
, ωI > 0, mI > 1 The Lag Network introduces a desired magnitude decrease
I
and a possible undesired phase lag. The greater mI the greater the magnitude decrease. Basic
guidelines for design:

2
 • Quantify the amount of the magnitude decrease ∆ϕ needed at ωc,des in order to shift the value
′
of ̸ L (jωc,des ) outside the ”influence” of the constant magnitude loci
• mI is is choosen on the basis of the required value of ∆ϕ
• ωI is fixed to obtain that the magnitude decrease ∆ϕ occurs exactly at ωc,des

We can use two different methods to decide the values of ωI , mI :

1. Empyrical method: To limit the phase lag at ωc,des the zero of the lag network mI ωI should
ωc,des
be sufficiently far from ωc,des . We have ωc,des ≈ αmI ωI , α >> 1 → ωI = αm I
. As a rule
of thumb, start with α = 10 and try to go down with its values. Increasing α decreases the
phase lag introduced at ωc,des
ωc,des
2. More precise method: using the Lead Universal Phase Diagrams we get ωI = ωnorm

Real Negative Zero

CZ (s) = 1 + ωsZ , ωZ > 0 The maximum phase lead introduced by CZ (s) is 90°. If we need more
than that multiple real negative zeros can be employed.

Issues: Since the transfer function of a real negative zero is not proper, it is not guaranteed, in
general, that the final overall controller C(s) = CSS (s)CZ (s) is proper. If C(s) is not proper, we
need to add closure poles that have to be placed at higher frequencies with respect to highest
KC (1+ ω s )(1+ ω s )
Z1 Z2
frequency of the controller zeros. C(s) = s(1+ ωs ) , ωP >> max(ωZ1 , ωZ2 )
P

Sensor Noise: Steady State Response to Sinusoidal Disturbances

At steady state, in the presence of sinusoidal dt = δt sin(ωt ) we have yss = δt |T (jω)| sin(ωt +
̸ (−T (jω)). The steady-state output error is |y ∞ | = max(|yss (t)|)) = δt |T (jω)|. The steady-state
dt
output error is required to be bounded by a given constant |yd∞t | ≤ ρt . This constraint can be
translated into a Constant Magnitude Locus MTHF = ρδtt , and this introduces a constraint on
the value of ωc such that ωc,des ≤ 0.1ωt

Output Disturbance: Steady State Response to Sinusoidal Disturbances

At steady state, in the presence of sinusoidal dy = δy sin(ωy ) we have yss = δy |S(jω)| sin(ωt+̸ S(jω).
The steady-state output error is |yd∞y | = max(|yss (t)|)) = δy |S(jω)|. The steady-state output error
is required to be bounded by a given constant |yd∞y | ≤ ρy . This constraint can be translated into a
ρ
Constant Magnitude Locus MSLF = δyy , and this introduces a constraint on the value of ωc
such that ωc,des ≥ 10ωy

Actuator Saturation
Physical limitations actuator devices impose hard constraints on the control input u(t). The above
described actuator limitations, impose to the control input u(t) a saturation constraint of the
form |u(t)| ≤ uM , ∀t ≥ 0. The saturation constraints can be described as a nonlinear static function
of the control input. To satisfy this requirement the only approach is ”a posteriori”. We need to
simulate the behaviour of the system (usually with a step input) and check if the requirement is met.
If it’s not met a common procedure is to reduce the value of the actual crossover frequency
ωc (keep in mind that this may cause unsatisfaction of transient requirements)