FreeLists / opendtv / [opendtv] Re: The numerology doesn't go away...

["Albert Manfredi" <bert22306@xxxxxxxxxxx>]

Olivier Houot wrote:

> Some news back from NAB 2007 about a Thomson H.264 encoder that
> compresses 1080i down to 6 or even 4 Mbps. Hype has it that this is
> done without sacrificing quality, but rather by making full use of all the
> tricks in the AVC toolbox with an optimized chipset :
Some time ago, one of ther CE giants claimed they could get 1080i down to 9 
Mb/s using MPEG-2 (H.262). At the time, I figured why not? If the content is 
slow, like content that started life as 24p, then that should in principle 
be feasible. If 720/24p can be transmitted with an average 9 to 10 Mb/s with 
H.262, in principle at least, the same content over 1080i should come close. 
Maybe not quite as good, but close.
> I can't help telling to myself that if it is acceptable to broadcast 720p
> in 14 Mbps today, and considering that 1080p compresses as well or
> perhaps better than 1080i, which apparently could tolerate 6Mbps,
> surely there must be a way to transmit 1080p in that same 6Mhz
> channel without suffering too adverse economic effects (at least
> from the broadcasting point of view).
I don't see any problem with that, as a recent infinite thread shows.

Correct me if I'm wrong, but isn't much of live HDTV captured at 1080i or 
720/60p today? If yes, then it is not oversampled, and yet it's been 
transmitted that way for 9 years, over 6 MHz channels, using H.262 
compression.
So it stands to reason that a properly designed 1080/60p camera should be 
able to create a bit stream with proportionally no more noise or other 
glitches than a non-oversampled 720/60p or 1080i stream. Which means it will 
compress down just as efficiently.
Or saying this another way, spurious noise has essentially infinite 
bandwidth. It is the initial LPF, before the sampling step, that is 
responsible for removing all components beyond the Nyquist frequency (or 
more than that, as Jeroen points out). So this initial LPF should be able to 
filter the noise bandwidth down to be proportionally no worse at 1080p than 
it is at 720p. Proportionally, I say again, in a correctly designed 1080p 
acquisition system.
And if oversampling makes things better, even that isn't out of the 
question, as you pointed out.
The bandwidth comparison is 2.25X more for 1080p than the equivalent 720p. I 
don't see any reason to assume going in that the 1080p must compress less 
efficiently than 720p. So, with a well developed H.264 algorithm, assuming 
the hype to be accurate, 1080/60p should fit in the 6 MHz channel.
> Or are we saying that there is an irreductible information core in a
> 1080p signal, and that better encoders can only get closer to that
> core, and not reduce it?
Why would anyone say that? I'm going in assuming that the actual scene has 
infinite detail. Hence, both 720p and 1080p encoding processes must cope 
with infinite excess bandwidth in the subject matter.
Bert

_________________________________________________________________
Play games, earn tickets, get cool prizes. Play now?it's FREE! 
http://club.live.com/home.aspx?icid=CLUB_hotmailtextlink1


----------------------------------------------------------------------
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.