The redundancy is in the 2/3 trellis coding of MPEG-2 data packets, which makes 8VSB performance on all-white Gaussian noise slightly better than 4VSB, and in the (207, 187) Reed-Solomon coding, which adds 20 parity bytes to the 187-byte MPEG-2 data packet to improve burst error performance. A-VSB uses turbo coding to add further redundancy. ----- Original Message ----- From: "Craig Birkmaier" <craig@xxxxxxxxx> To: <opendtv@xxxxxxxxxxxxx> Sent: Monday, January 15, 2007 8:57 AM Subject: [opendtv] Re: Analog v Digital TV > At 6:30 PM -0500 1/14/07, Tom Barry wrote: > >Albert Manfredi wrote: > >> Agreed, Tom. I was taking issue with your "no extra info is sent." > >> > >>You send redundant info, so that even though MPEG encoding deleted much > >> of this redundancy info from the image itself, your FEC schemes put it > >> back in. > > > >Yes, darn it, we agree you send some redundant info. (I am doing it > >now ;-) ) I was both originally, and still, saying and agreeing > >that no extra info is magically conjured up ("provided") on the > >receiving end that was not sent, at least in aggregate. Glad we > >have that cleared up. > > Obviously the issue is not cleared up. Once again Bert is smokin some > good stuff. > > The MPEG encoded bit are what they are. There is no redundancy or > error correction in the encoded bits. The MPEG encoder produces four > main components of the image that the decoder uses to reconstruct the > image: > > 1. I - frame DCT coefficients - These can stand on their own allowing > the decoder to reconstruct the anchor frame in the GOP. The encoder > quantizes the DCT coefficients and sends what is left, nothing more. > Think of this as a data file. THERE IS NO WAY TO MAGICALLY SEND > ADDITIONAL COEFFICIENTS AS PART OF ANY ERROR CORRECTION SCHEME. Any > downstream processing can only use the bits in what's left of that > "file," in an effort to reproduce it as accurately as possible. > > 2. Motion vectors - these are used by the decoder to reconstruct > other frames. These too are "data files" which carry the motion > vector information. Again, there is no error correction in these > files. Any downstream processing can only use the bits in that > "file," in an effort to reproduce it as accurately as possible. > > 3. DCT coefficients carrying the differences between the I frame and > the P frames that are included in the GOP. These coefficients may > also be quantized by the encoder. These too are "data files" which > carry DCT coefficient information. Again, there is no error > correction in these files. Any downstream processing can only use the > bits in that "file," in an effort to reproduce it as accurately as > possible. > > 4. DCT coefficients carrying the differences between the actual B > frames and the predictions of the B frames produced by the decoder. > These coefficients may also be quantized by the encoder. These too > are "data files" which carry DCT coefficient information. Again, > there is no error correction in these files. Any downstream > processing can only use the bits in that "file," in an effort to > reproduce it as accurately as possible. > > Obviously there is additional information about the encoded frames > that is delivered (size, aspect ratio, frame rate etc.). In other > words, more data. All of these bits are wrapped up in an MPEG > transport stream. Within THAT stream there may be redundancy bits, > pad bits to fill out the packets etc. BUT THERE IS NO ADDITIONAL > INFORMATION THAT CAN IMPROVE THE PICTURE. > > Anything downstream in an ATSC transmission is error protection with > the sole purpose of trying to get the data in the files I described > above to the decoder in the receiver with the lowest number of errors > possible. IF you can receive the ATSC transmitted bits perfectly, you > can reconstruct the MPEG video stream at the same level of accuracy, > as a decoder that is connected to the output of the encoder ( i.e. NO > CHANNEL ERRORS). > > IN AN ATSC TRANSMISSION THE VIDEO QUALITY CANNOT BE ANY BETTER THAN > WHAT WAS ORIGINALLY PRODUCED BY THE ENCODER, BUT IT CAN BE > SIGNIFICANTLY WORSE. > > Error Protection is there to deal with the fragile nature of the > transmission channel. And this error protection actually reduces the > delivered image quality. > > WHY? > > Because any overhead for error protection that is used in the > transmission channel reduces the payload available to the encoder for > image data. If you use 2 Mbps for error protection that is 2 Mbps > that cannot be used for image data. All of the cool stuff in A-VSB, > is there for only two purposes: > > 1. To help the equalizer in the receiver lock to the stream better. > 2. To protect/correct channel errors. > > The result is that the payload for the image data is further reduced. > > Regards > Craig > > > ---------------------------------------------------------------------- > 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. > ---------------------------------------------------------------------- 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.