[opendtv] Re: SFN considerations (was Doug is Missing the Point)
- From: "Allen Le Roy Limberg" <allimberg@xxxxxxxxxxxx>
- To: <opendtv@xxxxxxxxxxxxx>
- Date: Fri, 17 Feb 2006 08:50:17 -0500
Doug is familiar with the equalizer problem. Before he takes you to task,
let me explain that pre-echoes cannot be suppressed by recursive
(infinite-impulse-response or IIR) digital filtering.
Finite-impulse-response (FIR) filtering has to be used. A pre-echo is
suppressed in the FIR filter by combining two differentially delayed
signals, the principal signal in the less delayed signal adjusted to cancel
the pre-echo in the more delayed signal. But this causes a smaller new
pre-echo in the filter response that in turn may have to be cancelled. This
pre-echo is twice as advanced as the original echo. The strongest pre-echo
can usually be reduced to acceptably small size in seven or eight steps.
So, the kernel of the FIR filter should be seven or eight times (not twice)
as long as the range over which pre-echoes are expected to be strong.
Pre-echoes usually diminish in strength the further advanced they are, of
course. Substantial pre-echoes advanced as much as 30 microseconds have
been reported; substantial post-echoes delayed as much as 60 microseconds
have been reported. SFNs may up the echo range.
Al Limberg
----- Original Message -----
From: "Richard Hollandsworth" <holl_ands@xxxxxxxxx>
To: <opendtv@xxxxxxxxxxxxx>
Sent: Friday, February 17, 2006 2:25 AM
Subject: [opendtv] Re: SFN considerations (was Doug is Missing the Point)
> Looking forward to practicing (pretending) to read en Francais....
> So does anyone have a copy of the LINX report????
>
> BTW: I believe that the big breakthrough in Adaptive Equalizer
implementation (whether for HF communications or DTV) was the realization
that you needed a Double Length Equalizer, i.e. the equalizer length needed
to be at least TWICE the maximum "expected" delay spread between multipath
components.
>
> Sometimes (esp in urban environment), the first arriving signal can be in
a signal fading condition and the equalizer can lock onto a strong delayed
component (if there are any). In a classical "short" equalizer, it would be
initialized so that this strong signal would be towards the beginning of the
equalizer. Eventually, the early arriving signal would come out of the
fade, resulting in a very large "pre-echo". The Pre-Echo capability would
therefore need to be equal to the Post-Echo capability.
>
> You might think that you could detect when the excessive Pre-Echo came out
of a fade and very carefully right shift everything in the equalizer....but
that would require an special correlation detector, because the equalizer
would only know that things are messed up and not know exactly what to do
about it.
>
> Multiple resync retries could be attempted, but they would occur after the
equalizer has already declared a valid solution on the late arriving
multipath component, making it difficult to determine when to stick with the
current setting and when to go fish.
>
> <holl_ands>
>
> ===================================
> "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx> wrote:By the way,
here's some interesting stuff. It turns out that this French
> DTT (TNT) document
>
> http://www.csa.fr/pdf/Rapport-GT2-Aspects_Radiofrequence_de_la_TNT.pdf
>
> covers lots of SFN design issue quite well, both for receiver and
> transmitter considerations. Much like A/111 does. They discuss both
> true, synchronized SFNs, as well as on-channel repeaters. The majority
> of their discussion points apply equally to COFDM and 8-VSB. But a
> couple that are really almost humorous are (starting p. 22):
>
> 1. After defining what a pre-echo is, they caution that the presence of
> pre-echo can disupt reception on receivers that are so-called "1st or
> 2nd generation." That sounds familiar.
>
> 2. They mention that various synchronization strategies are possible for
> receivers, but they don't all favor pre-echo tolerance. As long as all
> echoes arrive within the GI, a strategy for pre-echo tolerance devised
> by Philips is to sync up on the first received signal, and treat all
> other echoes as trailing echoes. The interesting point here being that
> this was the essence of what Linx did, in their 4th generation demod.
> Linx discussed this in a white paper that disappeared along with their
> web site. Linx filtered the incoming signal in such a way that the
> equalizer was presented with a strong initial pulse followed by a long
> series of lower level trailing echoes, which equalizers can handle well.
>
>
>
> ---------------------------------
>
> What are the most popular cars? Find out at Yahoo! Autos
>
>
>
> ----------------------------------------------------------------------
> 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.
Other related posts: