UNIT 5 Session 5
UNIT 5 Session 5
UNIT 5 Session 5
Digital Control
Chapter 12 was concerned with building models for systems acting
under digital control.
We next turn to the question of control itself.
Topics to be covered include:
• why one cannot simply treat digital control as if it were
exactly the same as continuous control, and
• how to carry out designs for digital control systems so that
the at-sample response is exactly treated.
Having the controller implemented in digital form introduces several
constraints into the problem:
(a) the controller sees the output response only at the
sample points,
(b) an anti-aliasing filter will usually be needed prior to the
output sampling process to avoid folding of high
frequency signals (such as noise) onto lower frequencies
where they will be misinterpreted; and
(c) the continuous plant input bears a simple relationship to
the (sampled) digital controller output, e.g. via a zero
order hold device.
A key idea is that if one is only interested in the at-sample response,
these samples can be described by discrete time models in either the
shift or delta operator. For example, consider the sampled data
control loop shown below
These can be used and understood in essentially the same way as they
are used in the continuous time case.
Are there special features of digital control
models?
Many ideas carry directly over to the discrete case. For example, one
can easily do discrete pole assignment. Of course, one needs to
remember that the discrete stability domain is different from the
continuous stability domain. However, this simply means that the
desirable region for closed loop poles is different in the discrete case.
We are led to ask if there are any real conceptual differences between
continuous and discrete.
Zeros of Sampled Data Systems
We have seen earlier that open loop zeros of a system have a profound
impact on achievable closed loop performance. The importance of an
understanding of the zeros in discrete time models is therefore not
surprising. It turns out that there exist some subtle issues here as we
now investigate.
If we use shift operator models, then it is difficult to see the connection
between continuous and discrete time models. However, if we use the
equivalent delta domain description, then it is clear that discrete transfer
Functions converge to the underlying continuous time descriptions. In
particular, the relationship between continuous and discrete (delta
domain) poles is as follows (See Chapter 12):
p i , p i
The relationship between continuous and discrete zeros is more
complex. Perhaps surprisingly, all discrete time systems turn out to
have relative degree 1 irrespective of the relative degree of the original
continuous system.
Hence, if the continuous system has n poles and m(< n) zeros then the
corresponding discrete system will have n poles and (n-1) zeros. Thus,
we have n-m+1 extra discrete zeros. We therefore (somewhat
artificially) divide the discrete zeros into two sets.
1. System zeros: Having the property
z1
,..., z m
z 0q
We see that (in the delta form), the discrete system has a pole at =0
and a pole at =-0.9516. These are consistent with the continuous
time poles at s=0 and s=-1.
Note, however, that the continuous system has relative degree 2,
whereas the discrete system has relative degree 1 and a sampling zero
at -19.67 (in the delta formulation).
The next slide shows a plot of the sampling zero as a function of
sampling period.
Figure 13.2: Location of sampling zero with
different sampling periods. Example 13.1
In the control of discrete time systems special care needs to be taken
with the sampling zeros. For example, these zeros can be non-
minimum phase even if the original continuous system is minimum
phase. Consider, for instance, the minimum phase, continuous time
system with transfer function given by
For this system, the shift domain zeros of [G0Gh0]q(z) for two different
sampling periods are
Note that = 0.5[s], the pulse transfer function has a zero outside the
stability region.
Thus, one needs to be particularly careful of sampling zeros when
designing a digital control system.
Is a Dedicated Digital Theory Really
Necessary?
We could well ask if it is necessary to have a separate theory of
digital control or could one simply map over a continuous design
to the discrete case. Three possible design options are:
1) Design the controller in continuous time, discretize the result
for implementation and ensure that the sampling constraints do
not significantly affect the final performance.
2) Work in discrete time by doing an exact analysis of the at-sample
response and ensure that the intersample response is not too
surprising, or
3) carry out an exact design by optimizing the continuous response
with respect to the (constrained) digital controller.
We will analyze and discuss these 3 possibilities below.
1. Approximate Continuous Designs
where the factor s2 + 10s + 25 has been added to ensure that the
degree of Acl(s) is 4, which is the minimum degree required for an
arbitrarily chosen Acl(s).
On solving the pole assignment equation we obtain P(s) = 88s2 + 100s
+ 100 and This leads to the following PID controller
L ( s ) s 15.
For the above example, we see that method 1.1 (i.e. simply replace s
by ) has led to an entirely satisfactory digital control law. However,
this isn’t always the case as we show by the next example.
Example 13.3
The basic idea in this control design strategy is to achieve zero error at
the sample points in the minimum number of sampling periods, for
step references and step output disturbances (with zero initial
conditions). This implies that the complementary sensitivity must be
of the form
Case 1:
The plant sampled transfer function, G0q(z) is assumed to have all its
poles and zeros strictly inside the stability region. Then the controller
can cancel the numerator and the denominator of G0q(z) and the pole
assignment equation becomes
where
Simplifying, we obtain
Consider the servo system of Example 3.4. Recall that its transfer
function is given by
C q ( z ) 208.33 zz00..967
905
T0 q ( z ) 1z
Figure 13.6: Plant output for a unit step reference
and a minimal prototype digital control.
Plant with integration.
Note that the above results are essentially identical to the simulation
results presented for the motivational example given in the slides for
Chapter 12.
Minimum Time Dead-Beat Control
C q ( z ) G 0 q ( z ) B 0 q ( z )
T (z)
1 C q ( z )G0 q ( z ) zn
Example
Synthesize the digital control which achieves zero steady state at-
sample errors.
We observe that the reference generating polynomial qr(z), is given
by z10 - 1. Thus the IMP leads to the following controller structure
Notice that, after an initial transient, the output tracks the desired
periodic reference exactly at the sample points. Unfortunately, we
notice that tracking is only guaranteed at the sample points for this form
of control law.
Robustness of Repetitive Controllers