Center for New Music and Audio Technologies |
Next: Proof
Up: Controlling the Oscillator
Previous: Phase
Each sinusoid needs to be weighted appropriately to its magnitude
within the spectrum. The normal way to do this is to use a multiplier
controlled by the parameter amplitude. At the beginning of this
chapter we mentioned the expense of multiplication in digital hardware
and found a way around by using CORDIC for computing the sine
function. At this point a simple replacement of multiplication by the
CORDIC algorithm would be possible but is not necessarily a gain of
resources. However, as described by equation 1.2.1 the
CORDIC algorithm contains a gain due to the K-factor. We compensated
this gain by initializing the rotation with a vector of length ,
() respectively, so that the final vector was the unary
vector. Hence, initializing with different 's would result in a
vector with length which is not one. This leads to the idea of
initializing the rotation with the control parameter for
amplitude and performing multiplication and computation of sine
using the same resources. Thus, it is to be proved that the gain
inherent in the CORDIC algorithm is linear and indepent of varying
conditions, e.g., input angle, sequence of directions, initial values.
Subsections
Next: Proof
Up: Controlling the Oscillator
Previous: Phase
Home
Norbert Lindlbauer
2000-01-19