Re: [Discuss-gnuradio] GMSK code

[Joshua Lackey]

I wrote the original gmsk code (which used the old lame correlator) and
then Eric, Matt and Bob implemented the most excellent M&M synchronizer
and used that (with the rational resampler) to construct the gmsk2 code.
The comment is mine but doesn't make much sense with the changes made in
gmsk2.

The GSM documents specify that you should have up to three bits in the
Gaussian filter at once.  I took that to mean that you could have three
bits "ringing" in the filter at one time but not necessarily all three
bits completely in the filter.

The original code in gmsk had:

        ntaps = 2 * sps                 # up to 3 bits in filter at once

Where sps is "samples per symbol."  So if the number of filter taps is
twice the number of samples per bit, you can see that most of the time
the values from three bits are in the filter.  (Now and then there are
only two -- hence the "up to" part of the comment.)

Make sense?


Josh

ps: I've experimented and it doesn't seem to make all that much
difference.


Quoting Dawei Shen (address@hidden):
> Hi, Guys
> 
> In gmsk2.py, when we try to generate the Gaussian filter taps, the length of
> the filter is given by:
> 
>     ntaps = 4 * spb            # up to 3 bits in filter at once
> 
> Can anyone help explain why the comment said up to "3" bits? Eric?
> 
> Thanks
> 
> Dawei

> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/discuss-gnuradio


-- 
Joshua Lackey, PhD. -- address@hidden