Guest
Hello
I'm developing a CORDIC-pipeline to calculate twiddle-factors for FFT-
calculations. The required precision isn't known at the moment so I
want all word-lengths to be set by generics. In order to allow target
angles around the whole unit circle I have to check the angle and
choose a suitable 'start-vector' in the first stage. This is because
the area of convergence for the simplest implementation of CORDIC only
is -pi/2 - pi/2. Now to my problem.
Since the word-lengths aren't known until component instantiation, how
can I make sure that the constants are fed to the first pipeline stage
in the same precision as the internal signals? I have tables with the
constants in different word lengths, but the compiler complains loudly
about the data types not being correct when I try to use them.
Sincerely
Anton Andersson
I'm developing a CORDIC-pipeline to calculate twiddle-factors for FFT-
calculations. The required precision isn't known at the moment so I
want all word-lengths to be set by generics. In order to allow target
angles around the whole unit circle I have to check the angle and
choose a suitable 'start-vector' in the first stage. This is because
the area of convergence for the simplest implementation of CORDIC only
is -pi/2 - pi/2. Now to my problem.
Since the word-lengths aren't known until component instantiation, how
can I make sure that the constants are fed to the first pipeline stage
in the same precision as the internal signals? I have tables with the
constants in different word lengths, but the compiler complains loudly
about the data types not being correct when I try to use them.
Sincerely
Anton Andersson