Filter Design Toolbox

unitquantize

Quantize all numbers in a data set except numbers within eps of 1

Syntax

• unitquantize(q,...)
  

Description

unitquantize(q,...) works the same as quantize except that numbers within eps(q) of 1 are made exactly equal to 1 (see quantize for a description of the parameters).

This function is especially useful for quantizing fixed-point coefficients.

Examples

• [b,a] = ellip(4,3,20,.6); 
  m = tf2sos(b,a) 
  m = 
  
      0.2758    0.3883    0.2758    1.0000   -0.0548    0.4731 
      1.0000    0.7197    1.0000    1.0000    0.5884    0.9574 
  
  m==1 
  
  ans = 
  
       0     0     0     1     0     0 
       1     0     0     1     0     0 
  
  q=quantizer;
  m > realmax(q) 
  
  ans = 
  
       0     0     0     1     0     0 
       1     0     1     1     0     0 
  

It appears that there are four elements that are exactly equal to 1. In fact, there are only three. Element m(2,3) is greater than realmax(q), but less than 1. Ordinarily, m(2,3) would be counted as an overflow and be set to realmax(q). However, the desired behavior would be to force m(2,3) to be equal to 1 without recording an overflow. This is what unitquantize does, as shown in the following example.

• m = unitquantize(q,m) 
  m = 
  
      0.2758    0.3882    0.2758    1.0000   -0.0548    0.4731 
      1.0000    0.7197    1.0000    1.0000    0.5884    0.9574 
  m==1 
  
  ans = 
  
          0     0     0     1     0     0
          1     0     1     1     0     0
  

By forcing values between eps and 1 to be equal to 1, signal processing algorithms can avoid multiplication operations that involve these numbers, saving processing steps and time.

See Also
qfft, quantize

twiddles

unitquantizer