Jay Beavers jbeavers at microsoft.com
Fri Aug 30 19:40:27 CDT 2002

I'll throw in that ConferenceXP is capable of doing 640x480p with high
quality compressors.  It uses considerably less bandwidth than MPEG-1 or
MPEG-2 at equivalent sizes (MSVideo v7 is ~2 Mb/s CBR for 640x480p vs.
6-12 Mb/s VBR for MPEG-2) and is less CPU intensive so it can be encoded
on desktop computers instead of requiring a dedicated hardware encoder.
It also has ~300-400 ms latency.

One of the nice features of the Video Mixing Renderer is a smoothed
stretch of the video window.  With this, we generally demo our 'high
end' conferencing using a starting 320x240x30 fps stream but stretch it
to 640x480 on the rendering device.  At 750 kb/s, it's very hard to tell
the difference between this result and a DVD source signal.

We can also easily switch to full screen -- take a look at the Windows
Media comparisons found at
http://www.microsoft.com/windows/windowsmedia/WM8/video.asp to get a
feeling for the quality, and be sure to try playing them at 200% zoom
and full screen.

Do note two differences between these samples and ConferenceXP -- we're
using Windows Media v7, which consumes less CPU but is about 15-25% more
bandwidth than Windows Media v8 and some of these samples were probably
done using dual pass encoding which will improve the quality during
scene transitions since the encoder takes the time to smooth out the bit
rates over time.

We'll be investigating moving to the Windows Media v9 codecs which
improve quality yet again after they release on September 9th.  See
http://www.microsoft.com/windows/windowsmedia/9series/default.asp for
information on them.  These codecs support 5.1 audio encoding, but the
problem will remain of how to get the multichannel sound to start with.
Gentner devices reduce the microphone signals to mono and I've spoken
with our signal processing researchers and it's going to be quite
complex to combine the required echo cancellation, noise reduction, and
gain control algorithms with a multichannel output.

For those looking to play with codecs, I'll invite you to check out the
source code to ConferenceXP.  In the past, I've compared PICVideo's
M-JPEG and DivX's 4.02 and 5.01 codecs against Windows Media v7 for AG
scenarios.  My findings are that Windows Media Video v7 is a far better
codec, but you don't have to take my word for it ;-)  You'll find it
easier to add new codecs into this source since we're based on
DirectShow which many commercial and open source codecs work with out of
the box.  Heck, I think I have a configuration file setting that lets
you change audio and video codecs without recompiling ConferenceXP, but
it's been over a year since I used it so it's long gone from my head and
we didn't test this for release, so it may be broken.

Please note that when looking for the source code, I just completed an
edit pass on the docs, so it will be a few hours before it gets placed
on our site.

 - jcb

-----Original Message-----
From: Sascha Ignjatovic [mailto:sascha at chello.at] 
Sent: Friday, August 30, 2002 4:44 PM
To: ag-tech at mcs.anl.gov
Subject: Re: [AG-TECH] HDAG

am 30.08.2002 19:15 Uhr schrieb Jon Zeeff unter jzeeff at internet2.edu:

> Full NTSC quality video is readily doable.   But we should be working
> true HD - say 1280x720p.  30 frames/sec for video and
> remote screen control and perhaps .5 frame/sec for screen shots.

Transport of HDTV over IP networks


Internet HDTV


are just view of many initiatives in this directions..

> Processing this resolution at 30 frames/sec requires a lot of CPU and
> hardware assist.  Data rates could vary from 1.5 gbps (no
compression), 100
> mbits (DCT) to 19 mbps (mpeg2) to 5 mbps (mpeg4).

the compressed transfer of HDTV signal is delayed by half of a second
john shalf and other people told me..


More information about the ag-tech mailing list