[Discuss-gnuradio] Re: gr_noise_source_X( ) only negative seeds?

[Tim Meehan]

I have submitted a patch to  patch-gnuradio@  for gr_noise_source and qa code

The patch is to gr_noise_source_X rather than gr_random::ran1()
I am not sure why but most implementations from ran1() set all positive
values of seeds to 1.  Rather then mess with what may be convention I
thought it might be better to fix in gr_noise_source.

Index: gnuradio-core/src/lib/gengen/gr_noise_source_X.cc.t
===================================================================
--- gnuradio-core/src/lib/gengen/gr_noise_source_X.cc.t (revision 4763)
+++ gnuradio-core/src/lib/gengen/gr_noise_source_X.cc.t (working copy)
@@ -43,7 +43,8 @@
      gr_make_io_signature (1, 1, sizeof (@TYPE@))),
    d_type (type),
    d_ampl (ampl),
-    d_rng (seed)
+    // for ran1() positive values for seed all return same sequence
+    d_rng ((seed < 0)?seed:-seed)
{
}


On 3/8/07, Tim Meehan <address@hidden> wrote:
> Hi All
>
> When seeded with a positive number gr_noise_source_X seems to always
> return the same sequence.
>
> i.e.
>
> call gr_noise_source_f(gr.GAUSSIAN,1,Z) where Z > 0 the sequence is
> always the same.  Called with a negative seed (Z < 0 ) it behaves as I
> would expect.
>
> Am I missing something?
>
> I think this is from lines below in gr_random, but the comment
> made me scared to touch it :-)
>
> /*
>  * This looks like it returns a uniform random deviate between 0.0 and 1.0
>  * It looks similar to code from "Numerical Recipes in C".
>  */
> float gr_random::ran1()
> {
>   int j;
>   long k;
>   float temp;
>
>   if (d_seed <= 0 || !d_iy)  {
>     if (-d_seed < 1)                    <<<<<<<<<<  HERE
>       d_seed=1;                         <<<<<<<<<<  and HERE
>     else
>       d_seed = -d_seed;
>     for (j=NTAB+7;j>=0;j--) {
>       k=d_seed/IQ;
> <snip>