Re: [Discuss-gnuradio] Hardware sampling rates

[Johnathan Corgan]

On Mon, Mar 22, 2010 at 01:49, Charles Brain <address@hidden> wrote:

> I have tried the fractional interpolator and it just sucks too many CPU 
> cycles.

There is one more thing you can try--use the polyphase resampler.
This is a new block by Tom Rondeau based on the Harris book.  It
hasn't gotten a lot of use yet, but it is working so far and is very
efficient.

gr.pfb_arb_resampler_ccf(resample_rate, taps, size)

'resample_rate' is your desired interpolation rate
'taps' is the vector of FIR taps to use as the interpolation filter
'size' is the number of phases to use in resampler (default=32)

You can use gr.firdes.* to design a filter to create 'taps'; but the
sample rates to use require special calculation.  See:

http://gnuradio.org/redmine/repositories/entry/gnuradio/gnuradio-examples/python/pfb/interpolate.py

...for an example using both the polyphase integral interpolator and
the polyphase arbitrary resampler.

In your case, if I recall correctly, your final output filter is a
root-raised-cosine filter.  So you'd use the arbitrary resampler to
upsample to your final output sample rate, and use the RRC tap
generator to create the taps for it.

Johnathan