Here’s some pictures for Friday - I though these were interesting,
the output of the 0.9 atsc equalizer showing the 8 levels getting
more distinct with every field:
http://webpages.charter.net/cswiger/g09_eq_data-1.jpg
Zooming in:
http://webpages.charter.net/cswiger/g09_eq_data-2.jpg
shows the field sync which is largely made up of only +5/-5,
and zooming in on the beginning:
http://webpages.charter.net/cswiger/g09_eq_data-3.jpg
you can make out the +5 -5 -5 +5 segment sync, followed
by the start of the pn511 pattern (7 -5’s, +5, -5, then
8 +5’s).
The 2.0 port isn’t doing as well - a few eyes show up here
and there:
http://webpages.charter.net/cswiger/g2_eq_data-1.jpg
and that’s with the NOP equalizer. The LMS eq is all noise
all the time.
The 2.0 port isn’t doing as well - a few eyes show up here
and there:
http://webpages.charter.net/cswiger/g2_eq_data-1.jpg
and that’s with the NOP equalizer. The LMS eq is all noise
all the time.
To me it looks like the symbol timing tracker is not tracking properly.
As the sample time shifts relative to the correct sample time you come
in and out of alignment. As I recall, the equalizer comes after the
timing sync, so it won’t work. It needs proper symbol timing, and
without it will actually make things worse.
Matt
On Fri, 2006-05-05 at 14:45 -0700, Matt E. wrote:
in and out of alignment. As I recall, the equalizer comes after the
timing sync, so it won’t work. It needs proper symbol timing, and
without it will actually make things worse.
Yep - looking at the 0.9 bit-timing-loop output:
http://webpages.charter.net/cswiger/g09_btl_data-1.jpg
vs a 2.x port of it shows something going wrong, using
the same input data:
http://webpages.charter.net/cswiger/g2_btl_data-1.jpg
could be something to do with differences in buffering,
scheduling, etc between gr0.9 and gr2.x.
On Sat, May 06, 2006 at 10:16:15AM -0400, Chuck Swiger wrote:
vs a 2.x port of it shows something going wrong, using
the same input data:
http://webpages.charter.net/cswiger/g2_btl_data-1.jpg
could be something to do with differences in buffering,
scheduling, etc between gr0.9 and gr2.x.
I think it’s some kind of a porting error
Sorry, I can’t get to it right now, but please ping me early next week
if you haven’t got it sorted out.
Eric