Re: net questions for radio broadcast

John D. Lauer (johnlaue@umich.edu)
Thu, 20 Jul 1995 12:07:14 -0400


I have done a lot of thinking about this issue in the past. For broadcast
situations, where you expect hundreds or thousands of users there's nothing
like multicast and the MBone because you send one stream to all listeners
instead of a separate stream to each listener.

When people connect to reflectors with CU-SeeMe they each get their own
unicast stream. The load of 40 participants (the max for a reflector)
receiving 32 kbps DVI audio is 1280 kbps or 1.3 MBps, almost the bandwidth
of a T1. Sure the load for audio is less relative to video, but its still a
lot of data to be pushing around.

You could potentially set up a group of reflectors receiving feeds from
eachother across the MBone, and then reflecting the feeds in a unicast
manner to clients. But you're still using up huge amounts of bandwidth b/w
the reflector and clients. This would be your most efficient method though
with reflectors and CU-SeeMe.

Finally, CU-SeeMe simply does not have good broadcast quality audio. Just
connect to internetRADIO 141.214.138.248 and try to listen for more than 5
minutes without getting annoyed by the s/n ratio and distortion. The best
sound you'll get per kilobit is MPEG level III. You can have almost mono
broadcast quality at 64 kbps. CD-Quality at 128 kbps. But for real-time
compression you need expensive hardware and for decompression you'll need at
least a Pentium class machine. This is when I see broadcast digital radio
across the net a reality.

At 10:27 AM 7/20/95 -0400, david a. schlussel wrote:
>
>I haven't taken any networking classes as of yet, but for any gurus
>out there:
>
>Wouldn't the fact that CU-SeeMe does not "broadcast" but people
>simply connect to reflectors, have a much kinder effect on the net?
>
>Also, if radio were done audio recieve only connections as, I think its
>called CYBERNET, (is that it John?) does, there should be a much lower load
>on the net as a whole? Or is the packet structure the same regardless of
>whether the video contains anything? And if it does contain a STATIC
>picture, (which requires only minimal updates) The load shouldn't be
>too heavy.
>
>Which has a heavier cost on the net, audio or video?
>
>Two final thoughts, semi-queries, whatever, is that the constant audio
>transmissions required by a radio type service would add to the load
>
>Thanks much,
>
> +=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
> + David Schlussel +
> + dschluss@umich.edu +
> + MCIT-Special Projects +
> + http://www.umich.edu/~dschluss/ +
> +=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
>
>
>