You are on page 1of 12

PROBLEM SOLVING

COEFFICIENT QUANTIZATION ERROR


Co-efficient Quantization Error

   In this case of a digital IIR filter, 2nd order transfer function.

 The filter co-efficient in this case are and respectively. These are fractions of value less than 1. the
binary equivalent of these filter co-efficient are 0.1100 and 0.0001 respectively.
 But they are quantized the freq. response of the actual filter desired form. They are stored in finite
length register. There are several problems associated with finiteness in the length of registers used
for signal processing.
 Both IIR and FIR filters are affected by these problems. If the poles of the desired filter are close to
the unit circle, then those of the filter with quantized co-efficient may lie just outside the unit circle.
Problem


1.   Consider a 2nd order IIR filter with Find the effect on quantization of pole locations of the given s/y
fun. in direct form and in cascade form. Take B=4 bits.
Solution: Given
Cascade form:
Cont’d

  

Structure,
Cont’d
  
Direct Form-I:

Co-efficient are and

0 1


Cont’d

   4-bit truncation,
After

Structure⇒
Example:

1.  

⇒ factorization (-0.3 & -0.45)
2.
Overflow Error

 Overflow of bits during binary arithmetic addition causes some error called overflow error.
Ex: 3+6 ⇒ bits of the word length is 3 bit.
3 → 011 +
6 → 110
9 → 1001 ⇒ 4 bit
⇒ if the length of the binary register is 3 bit, it cannot store 4 bit o/p which makes
overflow of 4th bit over remaining 3 bits which cause error.
(i.e.) instead of 1001 → 011 ⇒ 101 ⇒ 000
↑ ↳ overlap of 4th bit ‘1’ over remaining bits
cause error, original o/p is not obtained

You might also like