Subject: Re: [linux-audio-user] jaming over the internet?
From: Robert Jonsson (robert.jonsson_AT_dataductus.se)
Date: Mon Sep 29 2003 - 22:33:45 EEST
måndagen den 29 september 2003 19.56 skrev Rob:
> On Monday 29 September 2003 12:50, Robert Jonsson wrote:
> > Anybody know of an application that allows streaming of midi
> > and/or audio over the net for the purpose of allowing several
> > people to jam together?
>
> I seem to remember something like that for Windows too, but
> remember that latency that would be more than acceptable for
> gaming (30-40ms) could make it impossible to jam as you're
> envisioning.
Well, 30-40ms might be stretching it a bit too far. But theoretically I think
it would be possible. Probably not with ADSL or Cable though...
First, ping is a roundtrip measurement. I would argue that only the time for
the data to travel one way is important, you wouldn't listen to yourself
playing over the internet, on the other end the stream would be listened to
by the jam-peer ( oh, if there isn't such an entity, can I invent it? can I?)
the peer would play her part back over the wire to complete the roundtrip.
Let's imagine a theoretical example using audio. Naturally the buffers and
soundcard would add to the latency so the buffers would have to be kept
short. With buffers at 2x128 the input latency would be about 5.3ms with
48khz sampling frequency, the soundcard would probably add another 2ms.
If ping times of 10 ms is possible, which is no big deal if the computers are
on the same DMZ, then the net would add half that, 5 ms.
Adding these values together, the peer would receive the audio and output it
over the speaker after about 5.3 + 2 + 5 + 2 + 5.3 ms = 19.6 ms.
Now... the packing of the audio data isn't added. I have no idea how
much/little that would add. I imagine the data would be sent with some
compressed format and packed in frames, like mp3...
5 ms is a nice round number that might be applicable as the time to pack and
schedule new audio data to the network card, the same number might also be
applicable for receiving and unpacking it at the jam-peer (tm).
This would mean that one way transportation would approximately take 30 ms.
Comparing this number to the actual time it takes for sound to travel through
air, we find that in 30 ms sound has travelled some whooping 10.3 m.
So... the scenario might be comparable to if you and your mate (sorry,
jam-peer!) would be standing some 10 meters apart. Probably not ideal, but I
think it would be workable.
With midi pretty much all the packing/unpacking would be omitted leaving even
better possibilities of making it realizable.
But then again, who knows until one has tried :)
I wonder if ecasound/ogg/shoutcast would be a workable combination if one
wishes to try something stupid.... :) ?
/Robert
This archive was generated by hypermail 2b28 : Mon Sep 29 2003 - 22:50:57 EEST