Scribd
Upload
120 views

Derivation of A Discrete-Time Lowpass Filter: Finn Haugen Finn@techteach - No

This document derives a discrete-time lowpass filter from the continuous-time transfer function of a first order lowpass filter using the Backward Differentiation discretization method. It shows that the discrete-time filter output y(tk) is calculated as a weighted average of the previous output y(tk-1) and the present input u(tk), where the weights are determined by the filter time constant Tf and time step h. The derivation results in an equation that defines the exponentially weighted moving average (EWMA) filter.

Uploaded by

triplbingazi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
120 views

Derivation of A Discrete-Time Lowpass Filter: Finn Haugen Finn@techteach - No

This document derives a discrete-time lowpass filter from the continuous-time transfer function of a first order lowpass filter using the Backward Differentiation discretization method. It shows that the discrete-time filter output y(tk) is calculated as a weighted average of the previous output y(tk-1) and the present input u(tk), where the weights are determined by the filter time constant Tf and time step h. The derivation results in an equation that defines the exponentially weighted moving average (EWMA) filter.

Uploaded by

triplbingazi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Derivation of a Discrete-Time Lowpass Filter

Finn Haugen
[email protected]
March 21, 2008

A lowpass filter is used to smooth out high frequent or random noise in a


measurement signal. A very common lowpass filter in computer-based
control systems is the discretized first order or time-constant filter.
You can derive such a filter by discretizing the Laplace transfer function of
the filter. A common discretization method in control applications is the
(Euler) Backward Dierentiation method. We will now derive a
discrete-time filter using the Backward Dierentiation method.
The Laplace transform transfer function also denoted the
continuous-time transfer function of a first order lowpass filter is
H(s) =

1
y(s)
=
u(s)
Tf s + 1

(1)

where Tf [s] is the time-constant. u is filter input, and y is filter output.


Cross-multiplying in (1) gives
(Tf s + 1) y(s) = u(s)

(2)

Resolving the parenthesis gives


Tf sy(s) + y(s) = u(s)

(3)

Taking the inverse Laplace transform of both sides of this equation gives
the following dierential equation (because multiplying by s means
time-dierentiation in the time-domain):
+ y(t) = u(t)
Tf y(t)

(4)

Let us use tk to represent the present point of time or discrete time:


Tf y(t
k ) + y(tk ) = u(tk )
1

(5)

Exact
slope,

y(t k)
y(tk )
y(t k )
y(tk-1 )
Slope with
Eulers
Backward
method

tk-1

tk

Figure 1: The (Euler) Backward Dierentiation approximation to the timederivative


Here we will substitute the Backward Dierentiation approximation for the
time derivative. This approximation is
y(t
k)

y(tk ) y(tk1 )
h

(6)

It is illustrated in Figure 1.
Tf

y(tk ) y(tk1 )
+ y(tk ) = u(tk )
h

(7)

Solving for y(tk ) gives


y(tk ) =

Tf
h
y(tk1 ) +
u(tk )
Tf + h
Tf + h

(8)

which is commonly written as


y(tk ) = (1 a) y(tk1 ) + au(tk )

(9)

with filter parameter


a=

h
Tf +h

(10)

which has a given value once you have specified the filter time-constant Tf
and the time-step h is given. (9) is the formula for the filter output. It is
ready for being programmed. This filter is frequently denoted the
exponentially weighted moving average (EWMA) filter, but we can just
denote it a first order lowpass filter.
According to (9) the present filter output y(tk ) is a function of the present
filter input u(tk ) and the filter output at the previous discrete time,
y(tk1 ). Therefore, the filter output must be stored in the program so that
it is available at the following execution if the filter algorithm.
It is important that the filter time-step h is considerably smaller than the
filter time-constant, otherwise the filter may behave quite dierently from
the original continuous-time filter (1) from which it is derived. h should be
selected as
Tf
(11)
h
5

You might also like