[Discuss-gnuradio] tv-reception

[Martin Dvh]

Hi all,
Prateek Dayal wrote:
> have u had any success with tv reception on gnuradio .... I am all
> caught up in a new job here and still fighting out to take out some
> time to work on this .. even my net connection has not been installed
> yet at home :(
John Gilmore wrote:
> Thanks for your ongoing work on this.  Television, here we come!  The
> next part, I think, will be looking for the sync pulse, throwing away
> the samples before it.  Eric has written such code before for HDTV, he
> can suggest where to look or what to do.  (He's improved the core
> buffer/sched code since then too, and may recommend using a new
> capability.)
Eric, can you give a hint?
>
> Copyright notice on the new code should be 2005, e.g. in gr_float_to_uchar.
> (There's always *some* nit that could be improved!)
I will patch this nasty bug ;-)


I am working on it right now.
There is a preliminary tv-reception example in cvs now.
gnuradio-examples\python\usrp\tvrx_tv_rcv.py
I got some images that way.

I am working on improved code right now.
(Got some better images)
First thing was to convert to unsigned char in stead of char.

Next thing is an agc (I have a somewhat working version of an agc, using 
standard blocks in python which is quite computationally intensive)
I also made an gr_agc_ff block based on gr_agc.h which should work 
better/faster.

Using an agc you can get rid of the scale and offset parameter in the 
tvrx_tv_rcv.py example.

What I am doing now is building a block which does am detection using a 
synchronous am_detector in stead of complex_to_mag
(this block contains agc, pll and complex multiplier. This is all in one block because the different signals for phase-detection and agc are 
re-used and interact with each-other)
This is mainly based on the FPLL which is in the old atsc code.

This block is also quite computationally intensive but should give us a cleaner 
video signal.
I haven't decoded video with it yet, but is seams to lock to a generated sinus 
if the initial frequency is close enough.

I want to have a clean and stable videosignal before we try to detect the 
sync-pulses.

I found that the right setting for rfgain and ifgain in the tvrx frontend is 
quite important.
This should be included in the agc loop but I don't know yet how.
Is there a way to send and get signals to other blocks in the flow-graph?
(This will probably be usrp tvrx specific so this code probably can't go into 
gnuradio-core)

Greetings,
Martin