[Discuss-gnuradio] Re: Seg Sync

[Eric Blossom]

On Fri, Mar 15, 2002 at 01:26:04PM -0800, Matt Ettus wrote:
> > I've got one too.  I'll compare the versions we came up with.
>
> It's no fun duplicating effort.

No problem.  I just used my commercial filter generator to build a
couple of different filters, so there was no duplication of effort.
I'll compare the output of the two to ensure that they have the same
frequency response (more or less).

> I try to make sure that I check in my changes every day, unless they
> break the build or cause things which previously worked to stop
> working.  That way everyone gets to see/fix/extend my stuff as soon
> as it's written.

Sounds good.

> I'm going to work on a data slicer, an ATSC generator, and a
> multipath channel model, unless you'd like me to work on something
> else next.

Those sound good.  Also, if you could put together one of those
daughtercards and send it to me that would also help.

In looking at our various sources of live ATSC data, it appears that
my reception of ch12 is probably the best by far.  The average fft of
the base band data is just about flat all across the band (with the
exception of the pilot, of course).  The data that I have from you
(563Mhz and 731MHz) as well as the two samples from John's are a long
way from flat.  I guess that's where the equalizer comes in.

For now, I resampled 200MB of my ch12 data up to 21.52 using the
bandlimited resampler code at CCRMA
http://www-ccrma.stanford.edu/~jos/resample/resample.html

My strategy for the segment syncs is based on the observation that
they occur at 10.76e6 / 832 = 12932 Hz


                                                  +-----> equalizer etc
                                                  |
GrAtscFPLL ----> RRC filter ----> fractional    --+--> slice on --> (A)
                                  interpolator         sign
                                  +/- k/(8*Fs)
                                    ^
                                    |
                                    +--------------------------------+
                                                                     |
                                                                     |
                            +----> * ---> lpf ---+                   |
                            |      ^             |                   |
       highly decimating    |      |             |                   |
(A) -->bandpass centered -->+   12932 Hz      atan (q, i) --> loop  -+
       at 12932             |      |             |            filter
                            |      v             |
                            +----> * ---> lpf ---+


With your daughter card I think that we just use the control output
to tweak the VCXO and that takes the place of the fractional
interpolator.

Do you think that this will work?

I figure that the "slice on sign" block actually subtracts the running
mean (DC from pilot) then makes it decision.

Eric

A scanner is looking pretty good about now...