John Shutt wrote: > I seemed to have taken away from the R-S presentation given to our SBE chapter last fall that AVSB monkeyed around with the Trellis encoder and R/S error correction to create "pseudo" training sequences. > As that requires that the data interleaver and a deterministic R/S reset be under the control of the modulator, AVSB would "break" all existing encoders anyway. A-VSB really is a composite of two proposals. One part of the proposal is to put prescribed reference sequences (SRS) of up to twenty bytes following a 6-byte deterministic trellis reset (DTR), all put into an adaptation field following a 2-byte adaptation field header in accordance with MPEG-2 protocol. The other part is to use turbo coding of a serial form that includes the 2/3 trellis coding as one of the component convolutional codes. This allows turbo coding to be done at half the code rate of ordinary 8VSB, rather than one third that code rate. This is per se a damned good thought IMO, but Samsung also puts the turbo coding into the adaption fields of MPEG-2 compliant data packets. I think the Samsung serial turbo code will be used, but will be packed into data segments differently than currently proposed. As I think John Wilkie observed, PSIP is not fully MPEG-2 compliant; and some other utility signals do not provide for adaptation fields. SRS is used to help auto-regressive adaptive equalization under dynamic multipath conditions, providing Wiener type feedback to help out Kalman type feedback. It is probably better to use Doug McDonald's DFT methods for adaptive equalization, which rely on self correlation of blocks of a few thousand 8 VSB symbols and skip the SRS. This also boosts effective code rate. Al Limberg ---------------------------------------------------------------------- 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.