Analog Devices VISUALDSP++ 3.5 Manual page 496

DSP Run-Time Library Reference
The twiddle table is passed in the argument
twiddle factors. The function
¾n
ize the array. If the twiddle table contains more factors than needed for a
particular call on
appropriately; otherwise it should be 1.
The argument
output to avoid overflow. If no scaling is selected by setting
to zero, then the input signal should be sufficiently conditioned to avoid
overflow. The
The function will perform static scaling if
static scaling, the function will scale intermediate results to prevent over-
flow. The final output will be scaled by
set to .
log2(n)
If
scale_method
Under dynamic scaling, the function will inspect the intermediate results
and will only scale to avoid overflow. Dynamic scaling therefore mini-
mizes loss of precision but at the cost of slightly reduced performance.
The
block_exponent
no scaling was performed) and
Algorithm
−
N 1
1
∑
=
x n ( ) X k W
N
=
k 0
The implementation uses core FFT functions implemented as direct
radix-4 algorithm. To get the inverse effect, it first swaps the real and
imaginary parts of the input, performs the direct radix-4 transformation
and finally swaps the real and imaginary parts of the output.
Domain
Input sequence length
4-80
, then the stride factor has to be set
ifftrad4_fr16
controls how the function should scale the
scale_method
argument will be set to zero.
block_exponent
is set to 2, then the function will select dynamic scaling.
will be set to a value between 0 (which indicates that
log2(n)
−
nk
( )
N
must be a power of four and at least 16.
n
VisualDSP++ 3.5 C/C++ Compiler and Library Manual
, which must contain at least
w
twidfftrad4_fr16
scale_method
, and
1/n block_exponent
(if static scaling was performed).
may be used to initial-
scale_method
is set to 1. For
will be
for ADSP-219x DSPs