Structures for LTI system functions

under the form H(z) = B(z)/A(z)

EE5630 Digital Signal Processing

Updated May 17, 2021
Prof. Yi-Wen Liu
National Tsing Hua University
 Basic topics
• Direct form I and II
• Cascade form
• Parallel form
• Transpose form
• “Ladder” structure for FIR
• Linear phase
Figure 6.6 _________________

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.7 ________________________

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.10 (a) Block diagram representation of a 1st-order digital filter. (b) Structure of the signal flow graph
corresponding to the block diagram in (a).

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
A simplified plot

Figure 6.11 Signal flow graph of Figure 6.10(b) with the delay branch indicated by z−1.

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.18 Cascade structure for a 6th-order system with a direct
form II realization of each 2nd-order subsystem.

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.20 Parallel form structure for 6th-order system.

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.29 Direct form realization of an FIR system.

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.32 Direct form structure for an FIR linear-phase system when M is an even integer.

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
Figure 6.25 Figure 6.26
A direct form II structure. Transposed direct form II

Discrete-Time Signal Processing, Third Edition Copyright ©2010, ©1999, ©1989 by Pearson Education, Inc.
Alan V. Oppenheim • Ronald W. Schafer All rights reserved.
 Advanced topics
• Quantization effects

• Stability issues
– Cascade helps simplifying the matter

• Workaround: the “coupled form”

THE BIGGEST PRECISION ISSUES:
STABILITY NON-PRESERVATION AFTER
COEFFICIENT QUANTIZATION
Figure 6.47 IIR coefficient quantization example.
(a) Log magnitude for unquantized elliptic bandpass filter.
(b) Magnitude in passband for unquantized (solid line) and 16-bit quantized cascade form
(dashed line).
Table 6.1 UNQUANTIZED DIRECT-FORM COEFFICIENTS FOR A 12TH-ORDER ELLIPTIC FILTER
Figure 6.48 IIR coefficient quantization example.
(a) Poles and zeros of H(z) for unquantized coefficients.
(b) Poles and zeros for 16-bit quantization of the direct form coefficients.
 Here is an advice:
Use cascade or parallel structures

Quantize the coefficients in this building block, and stability will preserve.
Figure 6.50
Pole-locations for the
2nd-order IIR direct form
system of Figure 6.49.

(a) 4-bit quantization of

coefficients.

(b) 7-bit quantization.

1
H(z) = 1
1 (2r cos ✓)z + r2 z 2
A structure (“coupled form”) that guarantees uniform
distribution of possible pole locations
Comparison: precise pole locations at the
expense of computation load?

direct form

“coupled form”