I think I solved a bug in gri_iir.h


#1

Hi all,
I think I solved a bug in gri_iir.h
I think the index numbers are wrong when you decrement the latest_n and
latest_m after you set prev_output and prev_input.
I created a simple standalone program to test the index numbers.
The patch and the test program are at:
http://www.olifantasia.com/pub/projects/gnuradio/mdvh/mypatches/gri_iir/

Because these things are tricky to understand I would appreciate some
feedback if the algorithm is correct now.
If it is, let me know and I can apply the patch.

I think gri_iir is only used in gr_iir_filter_ffd which is only used in
the emphasis filter of the wfm modulator.
And there it isn’t even used properly because the filter taps are not
set.

Greetings,
Martin

Index: gri_iir.h

— gri_iir.h (revision 5766)
+++ gri_iir.h (working copy)
@@ -143,12 +143,6 @@
for (i = 1; i < m; i ++)
acc += (d_fbtaps[i] * d_prev_output[latest_m + i]);

  • // store the values twice to avoid having to handle wrap-around in
    the loop
  • d_prev_output[latest_m] = acc;
  • d_prev_output[latest_m+m] = acc;
  • d_prev_input[latest_n] = input;
  • d_prev_input[latest_n+n] = input;
  • latest_n–;
    latest_m–;
    if (latest_n < 0)
    @@ -156,6 +150,13 @@
    if (latest_m < 0)
    latest_m += m;
  • // store the values twice to avoid having to handle wrap-around in
    the loop
  • d_prev_output[latest_m] = acc;
  • d_prev_output[latest_m+m] = acc;
  • d_prev_input[latest_n] = input;
  • d_prev_input[latest_n+n] = input;
  • d_latest_m = latest_m;
    d_latest_n = latest_n;
    return (o_type) acc;