Re: [Discuss-gnuradio] Polyphase Clock Synchronization with Matched Filter

[Su Li]

Hi Sylvain and Tom,

Thanks a lot. It works.

So the "Taps" should contain all the matched filter coefficients for Filter_Size (32) filters instead of just one filter. Then each filter takes SPS (4) out of them to generate a phase-shifted shaping filter. Hence, we have a filterbank with different phases. The optimal sample timing could be estimated as described in Tutorial 7.

Best regards,

Su Li

2015-04-27 15:58 GMT+02:00 Tom Rondeau <address@hidden>:

On Mon, Apr 27, 2015 at 9:34 AM, Sylvain Munaut <address@hidden> wrote:

Hi,

I am trying to use Polyphase Clock Synchronization (PCS) block to conduct timing synchronization for QPSK signal with half-sine pulse shaping. In the doxygen document of PCS, it says that the "taps" parameter could be the matched filter. Thus, I put the matched filter of half-sine pulse here, [0,sin(pi/4),1,sin(3*pi/4)]. However, the output is all 0. 

Is there anyone knows how to generate the "taps" parameter here?  Any suggestion would be appreciated. Thanks in advance.

I'm not an expert in all the details, but what I know is that the taps there need to be generated at "Filter Size" * "Sample Rate".

So I'd try something like [math.sin(math.pi*i/(32*4)) for i in range(32*4)]

Cheers,

   Sylvain

What Sylvain said should work, I believe. But you should study the manual page on this block:

http://gnuradio.org/doc/doxygen/classgr_1_1digital_1_1pfb__clock__sync__ccf.html

Hopefully understanding the way it works will help you understand why the taps need to be this way.

Tom