[opendtv] Re: Analog v Digital TV
- From: Tom Barry <trbarry@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Sun, 14 Jan 2007 09:19:42 -0500
Bert -
I think we are just talking around each other here, with maybe no
disagreement. I said error correction never "provides" new information
but meaning that sufficient bits of extra information have to be sent
("provided") by the sender, not mathematically created somehow at the
receiving end. The sent extra information is the redundancy that all
schemes include so you can calculate the values of any missing bits.
Reed-Solomon is a good example, say a simple erasure code. There, for
each group of so many symbols you can add a few extra ones, with
carefully calculated redundancy. If you drop a few symbols, say n, in
delivery you can make up the complete group as long as you have sent an
equal number n of redundant symbols (not just duplicates). But you
still have to plan ahead to send a sufficient number (at least n) of
extra ones to match the max number the receiver might miss.
After that number you are into error concealment, or sometimes resend
requests.
In any event, I suspect we are just debating the use of the word "provided".
- Tom
Albert Manfredi wrote:
Tom Barry wrote:
Error correction never really provides any extra information. It just
averages errors over a longer period making them more predictable
and able to be handled. But it always requires enough redundancy
to make up for the amount of errors over that long period.
Depends what kind of error correction you're talking about.
I'm talking about real error correction schemes here, like the
convolutional Viterbi or turbo codes, and the block codes like RS. They
do provide more information. They provide redundant information in the
bit stream to allow missed blocks of bytes to be reconstructed
completely. Within certain limits.
Beyond that limit, then you fall back to error *concealment* schemes,
which sounds like what you're referring to.
Nothing is free.
Indeed. So, for example, the Viterbi codes used in ATSC assign 1/3 of
the bit rate to redundancy. The Reed Solomon block code assigns almost
10 percent of the bit rate to redundancy.
What makes these schemes more effective than the accidental redundancy
available in NTSC is that they are specifically designed for the
purpose. So, while in NTSC you can still see an image when there's snow
and ghost, just because of the extra image info that isn't really
essential, you still do see the errors. Instead with DTV, within the
range of effectiveness, the errors are completely corrected, and then
less perfectly concealed beyond that range, and then all quits.
The Shannon limit for 19.39 Mb/s in a 5.38 MHz channel is 10.47 dB of
SNR. As the erorr correction is accomplished in ATSC receivers today,
about 15 dB of SNR is needed. So the question is, what does an NTSC
quality image look like if the signal is received with only 10.47 or 15
dB of SNR? And remember, that ain't HDTV. A more fair comparison would
be 4.5 Mb/s SDTV, against NTSC, and then the SNR needed for DTV is under
2 dB (A-VSB, diversity antenna). How would that look for NTSC??
Bert
_________________________________________________________________
Type your favorite song. Get a customized station. Try MSN Radio
powered by Pandora. http://radio.msn.com/?icid=T002MSN03A07001
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
--
Tom Barry trbarry@xxxxxxxxxxx
Find my resume and video filters at www.trbarry.com
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
