On 5 Oct 2001, Mayhem & Chaos Coordinator wrote:
Hi!
> On Thu, 2001-10-04 at 07:39, [EMAIL PROTECTED] wrote:
> > I mention this because although the obsd server seems to work
> > fine, clients experience some glitches (due to VBR?):
>
> Its probably not VBR -- just a guess. If the player stops playing for a
> second and has to buffer up, then maybe its a VBR problem. However, I
> suspect that you're getting the famous 'gurgling' sounds that an mp3
> decoder makes when a portion of its stream is missing.
That's it! gurgling defines it perfectly.
> Try this: Set up icecast and listen to the streams using
> winamp/freeamp/xmms, whatever. Do the problems go away?
I'll try... but I cannot use that here... our network, although
"powerful" (100_Base_TX Full switched, full duplex) is already very
busy. I can't afford about 20 or so people using bandwith with that. But I
see your point...
> If so, then you have a multicast problem. Multicast is carried over UDP,
Yeah, that make sense...
> which is not error corrected. When there is a collision on the network
> or for some reason one of the mp3 packets gets lost, the decoder is
> missing crucial information, and its decoder gets screwed up and outputs
> gurgling garbage. Also, does the problem get worse with a higher load on
> your network?
mmm... I also recall from reading some driver notes from Donald
Becker in the past that most ethernet cards multicast implementation, if
available, were flawed, and usually a poor hardware design couldn't be
solved by the driver implementation.
Tha fact is the obs server sits on the worst and unique hub
available in the network (the rest goes to either a Intel EtherExpress
520T or a 410T switch, I'd bet the problem are not due to them).
Another fact is the obs server is the only one with a Intel PCI
EtherExpress Pro100 82557. The rest all have one of the best (if not the
One) 100BaseTX cards available. Kingston's KNE100TX (tulip based, former
DEC and now Intel's 21x4x chip).
I'll try first swapping the card, if it doesn't behave better,
then I'll try hooking the server directly to one of the switches, I'll
keep the list informed.
> > Question: may this problems be due to VBR encoding? may the use of
> > reencoding solve the matter? Is reencoding integrated on the 0.4.0 branch?
>
> No. Reencoding will only make matters worse. Trust me. :-)
Ok!.
> > Secondly, I notice that there are very audible delay's/mismatch on
> > the music listened from different machines. Is this normal? It would be
> > great to have all be in synch, so that users near to others listening (at
> > moderate volumes) music from their own speakers hear the same, dunno if
> > this is impossible though.
>
> This one is tough, if not impossible to solve.
>
> The fundamental problem is that most sound cards to not take exactly one
> second to play one second of audio. Some cards do it in 1.000001 seconds
> and some do it in .9999999 seconds. Two identical cards from one
> manufacturer will even drift differently.
>
> So, one computer may be playing a bit faster than another computer. To
> the human the sound will be right because we cannot detect such minute
> changes. But two computers next to each other will drift apart from
> another over time, and there is little you can do.
hmm... well, I guess that the future multichannel will aid on
solving that: after all, if two persons are near, they could listen to the
same channel on one of the PC's and raise the volume a little; if they're
listening to different ones, they'll have to use headphones :)
> I hope this all makes sense.
Sure!
> P.S. I'd love to see a Spanish version of Obs. I'll do the German
> translation. :-)
count with it as soon as either the "segmentation" of the
HTML text from the cgi code is done, or a cgi version non prone to upgrade
soon is released!
thanks for everything,
*****---(*)---**********************************************---------->
Francisco J. Montilla - System & Network admin - Seville - Spain
pacopepe>AT>insflug.org - irc: pukka - DJB fan
DoQmail: qmail en castellano es.qmail.org - OpenSlink: insflug.org/slink
_______________________________________________
Obs-dev mailing list
[EMAIL PROTECTED]
http://www.freeamp.org/mailman/listinfo/obs-dev
