[opendtv] Re: (S) DTV adaptors available for less than 50 euros in France

The (207, 187) Reed-Solomon codes used in 8VSB DTV were designed to correct
10 bytes out of the 207 in each data segment.  That is, 10 bytes in the
187-byte MPEG-2 packet presuming the erroneous bytes are not in the R-S
parity bytes.

If one locates byte errors independently, 20 erroneous bytes can be
corrected in each data segment using an alternative RS decoding algorithm.
Erroneous bytes can be located pretty well by soft trellis decoding, but the
exact techniques are proprietary.  The methods are similar to those used for
correcting erasures in digital tape recording.  Linear block codes that
"halve" code rate can be used for error location, a (15, 8) code being
preferred.

It does appear to be true that there is not much further to be gained with
additional R-S coding.  The scuttlebutt is that the additional R-S coding in
Zenith's E4VSB is not very effective, although one can question how well the
measurements have been done.  Some time ago I had proposed transversal R-S
coding in addition to the lateral R-S coding in the 1995 DTV broadcast
standard.  The gains there were much less than originally hoped, which
supports the observations allegedly made concerning additional lateral RS
coding.  I suspect this is because burst error escalates rapidly once
signal-to-noise falls below a certain level.

The TRS coding scheme is described in published U.S. patent application
20040237024 for anyone interested.  TRS per se was first described in U. S.
patent No. 6,430,159 titled "Forward error correction at MPEG-2 transport
stream layer" and issued August 6, 2002 to Xiang Wan and Marc H. Morin.
They forward error correction coded using TRS across fields of MPEG-2
packets that had no lateral R-S coding or trellis coding.
Interestingly, the type of shortened RS coding used can identify the type of
material in a packet.  Thusfar nobody has used this to improve code
efficiency insofar as I am aware.

Al Limberg

----- Original Message ----- 
From: "Kon Wilms" <kon@xxxxxxxxxxxx>
To: <opendtv@xxxxxxxxxxxxx>
Sent: Monday, September 26, 2005 10:50 AM
Subject: [opendtv] Re: (S) DTV adaptors available for less than 50 euros in
France


> > There are practical limits on how many errors can be corrected using
> > Reed-Solomon codes.  IIRC, with ATSC it's around 10 per second.  On my
DTV
> > tuner cards, it's either a low number of uncorrected packets (or none)
or
> > it's no usable signal at all.
> >
> > Codecs are at least one layer away from this matter.
>
> You can also put error correction at the data level, and do it better
> yet than the bit-error-correction of RS by using packet erasure
> correction. You could recover a burst of lost packets using this approach.
>
> Codecs like windows media already have EC in their transport (I won't go
> any further than that, I am not a supporter of proprietary codecs). For
> ones that don't, you can always wrapper them. Ofcourse this is an
> additional hit on your decoder (leaky bucket delay), CPU and memory,
> depending on how you post-process these blocks and how far you spread
> the erasure coding.
>
> Cheers
> Kon
>
>
>
> ----------------------------------------------------------------------
> 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.
>
> ---
> [This E-mail scanned for viruses by Declude Virus]
>
>

---
[This E-mail scanned for viruses by Declude Virus]

 
 
----------------------------------------------------------------------
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: