Jeff McAdams on 14 Dec 2004 14:18:54 -0000 |
Doug Crompton wrote: > Anyone using commercial VOip? I was looking at the AT&T site. They claim > they need 90Kb Up/down BW/line/call. That really sounds excessive. What is > this super high fidelity phone connections! I would think they could get > by with a lot less then that even with overhead. Given a 3K voice channel > you should be able to sample at 12-16K, Add overhead and control.. so we > double that... 32K ... what the heck is the rest? I think verizon breaks a > T1 down into 23 voice channels / 1 control, each 48K. That always seemed > like a waste of BW to me. Someone already touched on this a bit...and I fear this may be more information than you wanted to know...feel free to skip this if you're not interested in the tech specs. :) POTS service uses a bandwidth range of about 400hz to 4000hz, giving a range of about 3.5khz. Shannon says that you have to sample at a bit more than twice the rate to get the data...double 3.5 is 7, so bump that up a bit to a nice round (in binary terms) 8khz, 8000 samples a second. Each sample is 8 bits (giving 256 quantization levels), so 8 bits, 8000 times per second is 64kbps, which is, as someone else, pointed out, the exact channel size of a T1 channel (preserved in ISDN). He also mentioned g.711 being a 64kbps codec, which is true, but even more significantly, its the exact some encoding of audio as is used in T1/PRI channels. Its also called PCM (Pulse Code Modulation). This gives VoIP the nice feature of not having to do any sort of signal processing if its connecting a digital channel to a VoIP call...just take an 8 bit sample off the channel, stuff it into a packet with all of the various headers and send it off, no manipulation of the audio stream whatsoever. Now, it was also mentioned that Cisco uses a very low ratio of audio data to header overhead, which is generally true (this value can usually be tweaked, but it defaults to a pretty small frame size). The reason for this is that the amount of audio data for a specific timeframe is fixed, so your trade-off is to buffer that audio data up to get more of them in a packet to build a bigger packet. The flip side of that trade-off is greater latency in the call. You have to send off these packets at a certain rate, or the other end will run out of audio data and there will be skips in the conversation...a problem worse than even just latency, so most of this VoIP equipment will default to smaller frame sizes and accept the header overhead. So, your audio data is 64kbps, period...that doesn't change, regardless of how many samples you pack into a packet...plus RTP, UDP, IP header, and probably some RTP payload overhead, you get up around 80kbps or so...maybe slightly over. If the packet sizes are kept small, that number creeps up a bit more. Most people that I've talked to will quote 90kbps in the interest of being conservative and making sure someone doesn't try a VoIP call on a link that's exactly 80kbps (maybe a GPRS phone connection? something along those lines) and being disappointed in the call quality. You could go with a different codec, but if one or the other sides is connecting to digital telco trunks, now its having to do more DSP manipulation of the actual audio data, rather than just repackaging the samples that are already there for it. Despite that, most VoIP gear will let you specify a set of codecs to use. G.711 is required to be supported, but others are frequently supported that offer lower bandwidths, with a similar reduction in audio quality (this is an area where a *LOT* of research has been done over the years, and it continues), and the VoIP call control protocols (SIP, H.323, etc.) do support the negotiation of codecs between end-points, so you can configure your phone or software to use a lower bandwidth codec, but fall back to g.711 is that's the only codec supported in common between the end-points. You'll find that most gateway services only support g.711 because of the ability to do the gateway function without any signal processing. > Anyhow it might make sense to change a DSL lines analog voice number to > absolute minimum service and get one of these all inclusive digital voice > packages. You can get everything under the sun for $29 and there are many > less expensive packages. The difference is that there are no hidden > charges like wireline or wireless where they hit you for darn near 20% of > your bill in hidden costs. A friend of mine here in Louisville, KY (my brother is in Philly, and I have visited, but I live in Louisville) is in the process of switching his DSL service to speakeasy, which offers "naked" DSL (no POTS service), and he's going to port his land-line number to Vonage (his job is currently to do a VoIP roll-out at the University of Louisville, so he's very comfortable with the technology). So, you might want to check with Speakeasy if they offer service up there. -- Jeff McAdams "They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -- Benjamin Franklin Attachment:
signature.asc ___________________________________________________________________________ Philadelphia Linux Users Group -- http://www.phillylinux.org Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce General Discussion -- http://lists.phillylinux.org/mailman/listinfo/plug
|
|