CF, Thom, re: bandwidth requirements for regular voice
The short Answer:
Circuit-switched voice requires from 32 to 64 kb/s, with 64 kb/s still the most common. And compressed, packetized voice usually requires between 6 and 20 kb/s.
-----
The long-[er, but by no means complete] answer:
As recently as three years ago one could generalize about this with some certainty, but no longer. The following addresses the international voice space, since that venue seems to be the area of interest here.
Circuit-switched PSTN and GSTN voice services usually require a full "digital signal at level zero" (or, DS-0) worth of bandwidth, or 64 kb/s. Increasingly, switched voice is using compressed forms of delivery (this is particularly prevalent by discount carriers) reducing the demand to 32 kb/s or 24 kb/s, or even lower, depending on the gateway carriers' choices of line and voice compression protocols.
In any event, these reduced rates are only taking place on the mid-link spans, and not on the end sections. At the tail sections, such as the local access and transport facilities supported by the xLECS, voice continues for the most part to be rated at a full DS-0 (64 kb/s). Even here, however, when DSL is used for voice VoDSL, for example, reduced bit rates are becoming common.
Now let's consider the the emerging packetized voice forms. The amount of equivalent bandwidth (bit rate) required is usually a function of voice compression algorithm used, [ii] the signaling environment and modality, and [iii] the size of the pipe used.
Voice over IP, or VoIP.. and in the near future, voice telephony over ATM, or VTOA... will soon be delivered at commercially acceptable levels, sometimes with increasing sacrifices to quality and latency at lower rates than PSTN, down to approximately 4 kb/s at the lower end, depending on the compression algorithm in use.
[I should also note that with improvements in DSP technology, we will also see higher fidelity and multidirectional sound in the future at the same rates as today's PSTN, or 64 kb/s. While I am it, I will also ask a question, and that is, Why are we breaking our backs compressing voice and conserving bandwidth, if a bandwidth glut is as imminent as some pundits claim?]
It appears that most VoIP applications going into the next several years will take advantage of the efficiencies offered by 8 or 13 kb/s compressed voice rates (the G.72x series), and will require approximately 16 to 24 kb/s per talker, after signaling overhead and 'headroom' - i.e., collision avoidance measures - are accounted for). Again, depending on the protocols used.
These are only some rule-of-thumb estimates. The compressed packetized rates above assume that we are talking about lower capacity pipes, like fractional T1s or T1s/E1s. With increasing pipe sizing, and depending on the fill rates/ratios of voice versus other forms of data, the relative demands on bit rate actually decreases (per talker), as you begin to utilize larger pipe sizes. These relationships are neither constant nor arithmetically linear, and depend on a host of parameters and other traffic management factors.
As a closing set of guidelines, it's safe to assume that regular PSTN switched services will continue to require from 32 to 64 kb/s, and that the emerging forms of VoIP will demand considerably less, usually under 16 to 20 kb/s, all factors considered.
Hope this helped, Frank Coluccio [more]
ps - I just noticed Robert S.'s comments. Thanks for the vote of confidence. The overall costs of transiting services continues to go down. But the larger piece of the pie, from a costing perpsective, is the amortized port costs at the end points (on the public network), and the costs associated with domestic switching, which are still largely dependent on class 5 central office infrastructures.
Mid-span costs are approaching zero, true, but until internet telephony truly approaches the functionality and surety on the 'net that can be trusted and supported, the same way we have grown accustomed to today's POTs, then we will continue to pay the piper for service.
Maybe the pricing floor will be that which the ILEC charges for monthly loop charges and per minute switching charges. It's interesting to note here that while the Interexchange carriers are in the midst of another pricing war, the ILECs service charges haven't decreased nearly as fast, in some cases not at all.
But competition is coming to the ILECs, too. Competition, and self-cannibalization on some levels, as they duck and weave, I might add. And so it goes... |
