[opendtv] Re: Backward Compatibility of Robust DTV Transmissions

  • From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx>
  • To: <opendtv@xxxxxxxxxxxxx>
  • Date: Tue, 7 Nov 2006 15:41:08 -0500

John Shutt wrote:

> Once a receiver is built, how much can be done ex post facto to
> improve the performance? With so many receivers designed around
> "blind equalization" exclusively, I don't understand what sort
> of "training wheels" can be added to ATSC to help them.
> I can see how one could create pseudo-training sequences ala
> A-VSB that could help new receivers improve their performance
> without upsetting existing hardware, but I think the best one
> can hope for with legacy receivers is similar to the Hippocratic
> oath, "first do no harm."

Hmmm. Since blind equalization works best with a strong signal, why
wouldn't it help these legacy receivers if the 8-VSB transmissions
included some variable number of "bogus" segments, each of which is, for
example, always modulated at 100 percent (i.e. all large amplitude

-------Quote from A/54 p. 96---------
Tracking dynamic echoes requires tap adjustments more often than the
training sequence is transmitted. Therefore, the prototype Grand
Alliance receiver was designed so that once equalization is achieved,
the equalizer switches to a decision-directed equalization mode that
bases adaptation on data symbols throughout the frame. In this
decision-directed equalization mode, reception errors are estimated by
slicing the data with an 8-level slicer and subtracting it from the
equalizer response.

For fast dynamic echoes (e.g., airplane flutter) the prototype receiver
used a blind equalization mode to aid in acquisition of the signal.
Blind equalization models the multi-level signal as binary data signal
plus noise, and the equalizer produces the error estimate by detecting
the sign of the output signal and subtracting a (scaled) binary signal
from the output to generate the error estimate.

To perform the LMS algorithm, the error estimate (produced using the
training sequence, 8-level slicer, or the binary slicer) is multiplied
by delayed copies of the signal. The delay depends upon which tap of the
filter is being updated. This multiplication produces a
cross-correlation between the error signal and the data signal. The size
of the correlation corresponds to the amplitude of the residual echo
present at the output of the equalizer and indicates how to adjust the
tap to reduce the error at the output.

I've noticed, for example, that in certain cases, DTT channels with a
lot of echo take longer to show up on the screen. It seems to me that
there must be some particular sequence of data symbols that makes the
blind equalization process easier, i.e. more robust. And once synched
up, the receiver tends to stay locked in, so this happy sequence of
symbols wouldn't need to be repeated extremely frequently.

Of course, it takes up some bandwidth.

You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: