"stephen" <stephen_hope@xyzworld.com> hath wroth:
>> In that case, you'll do fine with your existing setup. The limiting
>> factor is the outgoing bandwidth and compress scheme selected. The
>> worst is G.711 PCM which hogs 64Kbits/sec and is NOT compressed.
>
>Jeff - this is the voice bandwdith, but doesnt include the IP packet over
>head.
Sorry. I made some assumptions that I didn't bother detailing. Throw
in UDP instead of TCP, silence suppression, and raw protocols. See
below for numbers. A better chart would be:
<http://www.voip-info.org/wiki-Bandwidth+consumption>
Codec BR NEB
G.711 64 Kbps 87.2 Kbps
G.729 8 Kbps 31.2 Kbps
G.723.1 6.4 Kbps 21.9 Kbps
G.723.1 5.3 Kbps 20.8 Kbps
G.726 32 Kbps 55.2 Kbps
G.726 24 Kbps 47.2 Kbps
G.728 16 Kbps 31.5 Kbps
iLBC 15 Kbps 27.7 Kbps
BR = Bit rate
NEB = Nominal Ethernet Bandwidth (one direction)
>G.711 is 80 to 84 Kbps. (as used at work on Cisco call manager).
I'm currently doing battle with a 7690 IP Phone. The phone is
winning. I'm trying to convert it to a SIP phone and TFTP update
hangs at 80% complete. It's probably broken.
Yep. I goofed on G.711.
>when you get to G.729A or B with 8 Kbps of voice samples, you have 24 Kbps
>of actual traffic.
>http://www.erlang.com/calculator/lipb/
Yep. I thought the chart looked a bit too simple. I was in a hurry,
didn't read the associated text, wasn't paying attention, etc.
Anyway, this is the online calculator I prefer to use:
<http://www.bandcalc.com/>
Your 24Kbits/sec is for TCP. More common is UDP requires only
16Kbits/sec for G.729b. If I tack on silence suppression, the average
bandwidth utilization approached 9Kbits/sec while the peak bandwidth
remains at about 20Kbits/sec.
I think (not sure) that the appropriate model for wireless is 802.3
emulation (under the link window pull down). With G.729ab, 802.3, and
silence suppression, I get 9.6Kbits/sec average and 19.2Kbits/sec
peak. That's not my claimed 8Kbit/sec bandwidth, but it's certainly
less than your 24Kbits/sec.
Another calculator (that doesn't have silence suppression) is at:
<http://site.asteriskguide.com/bandcalc/bandcalc.php>
However, it includes the note:
"The results obtained from the calculator should be considered
only in one direction. In half-duplex networks, like 802.11b(Wifi),
you should double the estimates."
which means that I also ignored the half/full duplex problem and that
all bandwidth numbers I stated should be approximately doubled for
wi-fi. In the "L2" pulldown menu, if I select 802.11 wi-fi, the
bandwidth numbers double. Oh-oh.
>Note - the overhead varies depending on the physical link - AFAIR this is
>for Ethernet.
Yep. 802.11a/b/g wireless is encapsulated ethernet.
>On DSL it will depend on the protocol stack used.
Yep. My at&t DSL is PPP over ATM (RFC1483/2684). The "ATM tax" is
fairly heavy. Estimating from my download speed, I get 1250 bytes
through for every 1500 I send or about 17% overhead. Ouch. That
doesn't show in any of the VoIP bandwidth calculators and needs to be
added.
Thanks for the correction and sorry about my screwup(s).
--
Jeff Liebermann
jeffl@comix.santa-cruz.ca.us
150 Felker St #D
http://www.LearnByDestroying.com
Santa Cruz CA 95060
http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558