Mythbusting HD Voice: Frequency Response vs Data Rate

snom_820_links_hoch_perspektive_200pxOne common misconception that keeps coming up is the assertion that the higher quality audio available through the use of wideband telephony (aka HDVoice) requires more bandwidth on the network. This is simply not true.

The terminology gets confusing for some folks. On the one hand we’re talking about frequency response of the audio channel being much greater, 50 Hz to >7 KHz for most wideband codecs, as compared to 300 Hz to 3.4KHz for the reference standard G.711. All that extra information has to go somewhere, right?

Essentially all of the codecs deployed on the PSTN are quite old. Some like G.711 are, in technology terms, extremely old. At the time of their design the world of technology was very different. CPU and DSP chips offered much less processing power so they could not undertake the complex compression schemes found in more current codecs.

Consider G.729, the PSTN reference standard reduced bitrate codec. G.729 is itself quite vintage. I once had a SIP phone that could handle several G.711 calls at the same time, but only one G.729 encoded call. Why? The DSP inside the device simply didn’t have the power to encode & decode multiple call streams at once. In that case the phone only supported three-way conferencing when all the calls were G.711.

Leap ahead a few years to today. Even the relatively old G.722 codec only requires 64 kbps before packet overhead. G.722 was introduced in the 1980’s at about the time that ISDN was being touted as “the next big thing” so it had to fit into the same 64k B-channel as a normal analog call. That’s exactly the same data rate as G.711, yet with a 16 KHz sample rate it conveys  the full 50Hz – 7Khz frequency response.

More modern codecs & processor technology allow for higher compression rates without incurring additional processing latency. So extended frequency response does not necessarily mean increased data rate. If you look at the following table of codecs you’ll see that only two of the current crop actually exceed the data rate of the plain old G.711.

The codecs are indicated as narrowband or wideband by color code.

AMR 4.75 – 12.2 kbps
AMR-WB (aka G.722.2) 6.6 – 23.85 kbps
AMR-WB+ 5.2 – 48 kbps
G.711 64 kbps
G.711.1 64 – 96 kbps
G.719 32 – 128 kbps
G.722 48 – 64 kbps
G.722.1 24 – 32 kbps
G.722.1 Annex C 32 kbps
G.726 16 – 40 kbps
G.729 8 kbps
G.729.1 8 – 32 kbps

Some of the concern about codec selection and bandwidth use revolves around the fact that some ISPs are imposing monthly transfer caps, and charging their customers for overages. Given that reality consider the following:

  • G.722 or G.711 = 64 kbps before IP overhead, or about 80 kbps with IP overhead.
  • 80 kbps = 10 kB/s
  • 10 kB/s = 600 kB/minute
  • 600 kB/minute = 3,600 kB/hour = 3.6 MB/hour
  • 3,600 kB/hour = 864,000 kB/day = 864 MB/day
  • 864 MB/day = 25.92 GB/mo, presuming a 30 day month

It seems unlikely that typical user (ie not a call center) would crest their ISPs transfer cap based largely on the VoIP calling habit.