At last count, there were maybe 8 or 9 wideband or superwide band codecs floating around that I'm aware of. Let's assume there's at least two or three more that I haven't heard about to make it an even dozen or so HD voice codecs in existence.
Out of this wide selection, I've come to the conclusion that only four -- maybe 5 -- really count when it comes to developing products and deploying HD voice networks. It's a harsh assessment that's likely to offend some who love the latest open source codecs floating around.
You have to start off with placing G.722 -- the original gangster wideband codec -- at the top of the list. It is the touchstone for defining what HD voice means and is incorporated into literally millions if not tens of millions of IP handsets around the world. It is nearly impossible to buy an IP business handset today that doesn't have G.722 support built on or available via firmware upgrade.
G.722 gets a lot of love from manufacturers because it is a royalty-free codec, easy to implement, and doesn't burn a lot of processing time relative to newer options. Codec junkies will dog G.722 in three areas: It's old, it requires 64 kb/s of bandwidth, and it's 'only' wideband, sampling at 16 kHz, not at a full 24 kHz for superwideband.
In an ideal world - OK, my ideal world - G.722 would be adopted by the mobile world so we could have seamless end-to-end capability, but the cellular world is locked into a mentality that voice and data are separate animals and that bandwidth should be conserved at all costs. This argument is weakened by the world-wide proliferation of two-way mobile video services that consume more than 64 kb/s.
AMR-WB, an evolution of the AMR codec in the GSM cellular world, is the current wideband leader for HD voice in the mobile world. It has the backing of France Telecom, Nokia, and Ericsson (News - Alert)and the codec is in active use in France Telecom's rollout of HD voice across Europe. Handset manufacturers, including Nokia and Sony Ericsson, have lined up to support it.
The cellular world likes AMR-WB because it can deliver a wideband experience in 24 kb/s -- a fraction of the 64 kb/s for G.722. It's also no secret that the vendors most actively promoting AMR-WB -- France Telecom (News - Alert)and Ericsson -- also hold patents for the codec, so there's a definite financial interest by those companies in the proliferation of the codec.
Needless to say, use of AMR-WB requires a royalty payment. Device developers have said implementing AMR takes much more work than G.722 because of the many different modes it can operate in. And while it may be bandwidth frugal compared to G.722, it comes at the price of using more compute cycles for data compression which leads to shorter battery life.
Skype (News - Alert)'s ultrawideband SILK codec straddles two worlds. Developers like it because the source code has been published as an IETF standard under an open source license and it's royalty-free. It is a variable-bit-rate 'VBR' codec, adopting its performance from superwideband to wideband to narrowband depending on the bandwidth, network, and compute power available.
However, in order to deliver higher voice performance, SILK uses more CPU cycles, resulting in lower battery life on mobile devices; one tweet I recently saw said, 'An iPhone (News - Alert)on a Skype call makes a wonderful hand warmer.'
Rounding out my foursome of HD voice codecs is Global IP Solution's iSAC. The wideband codec is proprietary and requires licensing from GIPS, but it has been incorporated into a huge number of softclients around the world, ranging from AIM, GoogleTalk, and Yahoo to WebEx and IBM (News - Alert)Lotus Sametime. While GIPS won't venture to estimate the total number of HD voice-using soft clients that are in operation, I think it'd be fair to say the number is likely in the hundreds of millions.
While there are other HD voice codecs floating around, developers cannot realistically support an infinite number of them - even if they are 'free' in the sense of being open source and royalty-free . At this point in time, this is the foursome I'd bet on.
Doug Mohney is a contributing editor for TMCnet and a 20-year veteran of the ICT space. To read more of his articles, please visit columnist page.
Edited by Stefania Viscusi