Approximation of Derivatives
Spring 2012
© Ammar Abu-Hudrouss -Islamic ١ University Gaza
Introduction
IIR filter design techniques depend mainly on converting analogue filter to a digital one.
Analogue filter can be described by three methods 1) Its system function. M k sk k 0 H a (s) N k sk k 0 2) By its impulse response which is related to the system function by
st H a (s) ha (t)e dt
٢ Digital Signal Processing ٢ Slide
١ Introduction
3) By linear constant-coefficient differential equation d k y(t) M d k x(t) k k k k dt k 0 dt
Each of these three characterizations leads to alternative method for converting the analog filter to its digital equivalents
٣ Digital Signal Processing ٣ Slide
Introduction
The system function Ha(s) is stable if all its poles lie in the left half of the s-plane. Hence, the conversion is effective if
1. The j axis in the s-plane map into the unit circle in the z-plane. 2. The left-half plane LHP of the s-plane map into the inside of a unit circle in the z-plane (i.e stable analogue filter will be converted to stable digital filter.
٤ Digital Signal Processing ٤ Slide
٢ Approximation of derivatives
The simplest methods to convert an analogue filter to a digital filter is to convert the differential equation into equivalent difference equation.
For the first derivative dy(t) y(nT) y(nT T) y(n) y(n 1) dt tnT T T
y(t) H (s) s dy(t) dt
1 z 1 y(n) H (z) y(n) y(n 1) T T
٥ Digital Signal Processing ٥ Slide
Approximation of derivatives
For the analog derivative, the system function is given by H (s) s
Whereas, the digital system function is H (z) (1 z 1 ) /T Which leads to 1 z 1 s T
٦ Digital Signal Processing ٦ Slide
٣ Approximation of derivatives
The 2nd derivative is replaced by the second difference
d 2 y(t) d dy(t) [y(nT ) y(nt T)]/ T [y(nT T ) y(nT 2T)]/T 2 dt dt dt t nT T y(n) 2y(n 1) y(n 2) T 2 In frequency domain this is equivalent to
2 1 2z 1 z 2 1 z 1 s 2 2 T T
Then the kth derivative is equivalent frequency-domain relationship
1 k k 1 z s T ٧ Digital Signal Processing ٧ Slide
Approximation of derivatives
The system function for IIR filter obtained as a result of approximation of the derivatives is H (z) H (s) a s(1z 1 ) /T Let us investigate the implication of the mapping from the s-plane to the z-plane 1 z 1 sT Then the j axis is mapped into 1 z 1 jT 1 T j 1 2T 2 1 2T 2
٨ Digital Signal Processing ٨ Slide
٤ Approximation of derivatives
The y -axis in the s -plane is mapped into a circle of radius ½ and with centre z = ½. Any point in the left hand plane will be mapped to points inside that circle. So the resultant filter is stable
٩ Digital Signal Processing ٩ Slide
Approximation of derivatives
The possible locations of poles of the digital filter are confined to small frequencies.
As consequence, the mapping is restricted to low-pass and band-pass filters with relatively small resonant frequencies.
It is not possible to transform a high-pass analog filter into a corresponding high-pass digital filter.
In attempt to overcome this limitations, more complex substitutions for the derivatives have been proposed such as
dy(t) 1 L y(nT kT) y(nT kT) tnT k dt T k 1 T
١٠ Digital Signal Processing ١٠ Slide
٥ Approximation of derivatives
The resulting mapping between the s-plane and the z-plane
is now L 1 k k s k z z T k0
When z = ej j2 L s k sink j T k1
Which is pure imaginary, which means by carefully selecting ks, The j axis can be transformed into the unity circle. This can be done by using optimization
١١ Digital Signal Processing ١١ Slide
Approximation of derivatives
Example: Convert the analogue bandpass filter with system function 1 H s a (s 0.1)2 9 Into a digital IIR filter by substituting for the derivatives method 1 H (z) ((1 z 1 ) / T 0.1) 2 9
T 2 /(1 0.2T 9.01T 2 ) H (z) 2(1 0.1T ) 1 1 z 1 z 2 1 0.2T 9.01T 2 1 0.2T 9.01T 2
T should be less than or equal 0.1 for poles near the unity circle
١٢ Digital Signal Processing ١٢ Slide
٦ Approximation of derivatives
Example: Convert the analogue bandpass filter in the previous example by use of the mapping 1 s z z 1 T Solution: by substituting for s in H(s), we obtain
1 H (z) ((z z 1 ) /T 0.1) 2 9
z 2T 2 H (z) z 4 0.2Tz 3 (2 9.01T 2 )z 2 0.2Tz 1
H (z) has four poles while H (s) has two poles which means that the conversion has led to more complex system
١٣ Digital Signal Processing ١٣ Slide
٧