Intel 2920 Design Handbook page 75

ADVANCED TECHNIQUES
Xl-------___.._ .....
.......---WT
B 1 A - - t - - - -
Figure 6-2. Cascade Realization
This range for BI suggests that a control variable
representing a fraction of BI could be used (as any con-
trol variable c is limited to the range
-1~c<1.0)
or BI
could be represented as the sum of a constant and
variable part, e.g., BI = 0.7546. As c varies from 0.0601
to 0.8167, the filter sweeps through the desired range.
By limiting c to positive values, i.e.,
O~c~
1.0, some
additional range is provided, and a simpler multiply
alogrithm is usable.
Therefore a block diagram such as is shown in Figure
6-3 results where the filter input is X, its output is Y,
and
O<c~
1 is the control parameter.
The maximum gain of the filter can be found from
For the coefficient values given, the maximum gain is
21.9, when c=1.0.
Note that the gain varies with the value of c, reaching a
minimum when c=O.O of 17 .9, corresponding to a 1.76
dB gain variation over the setting range. If such a varia-
tion is unacceptable, c may be used to weight the input
X to compensate. For example, if the weighting of X is
of the form
X(2c-ll)1256
overflow is prevented, and gain variation is on the order
of ±0.03 dB.
This weighting of the input can be achieved by condi-
tionally adding X to YO using the value of c in the DAR,
followed by three terms to subtract IIX1256 from YO.
Note, however, that the polarity of X has effectively
been reversed by this procedure.
Table 6-la shows the basic variable filter program,
while Table 6-1 b shows the gain compensation.
x---------------------~.
__
I - - - - - Y O
c
Figure 6-3. Second Order Stage With Variable
Center Frequency
6-2