[PD] signal continuity and filtering with block~ 2*N 2

Charles Henry czhenry at gmail.com
Tue Feb 7 19:59:37 CET 2006


I have been designing some externals for subband adaptive filtering,
and I realized there was something wrong with my approach.

I needed the block~ 2*N 2 inside a subpatch, to get a block of N new
samples, while holding the N old samples for filtering.  The actual
filtering operation works on only the N new samples, leaving the N old
samples the same.

So, the thing is, I want to circumvent the addition of overlapping
blocks, and have the output based only on the samples which have been
filtered.  I came up with one method, called reformat~, which does
this:

signal : |  N old samples  | N new samples |
output: |  N new samples |  zeros--------------|

It just moves N samples from the front to the back.  This is resulting
in clicking and brief cuts in the output.

So far, I have coded all necessary parts, except the adaptation of
filter coefficients.  There may still be something wrong with the
convolution routine, but (when not using reformat~) the signal shows
good continuity and no clicks.

Any thoughts?  Suggestions?

Chuck




More information about the Pd-list mailing list