More on the HD Communication Summit
As pointed out in John’s blog, the HD Communication Summit last week was a great gathering of industry experts, all with the same goal of advancing HD voice deployments. At times the discussions were fairly but always very constructive.
The hottest issue related to the number of wideband codecs that need to be supported. Some suggested that in order for HD Voice to really take off a very limited (two to three) set of codecs has to be agreed upon. Clearly there are scenarios where interoperability is no issue and hence any codec can be used in such scenarios. Others said that it is unrealistic to assume that such a small set of codecs will be agreed upon and that transcoding will be a necessity. AudioCodes, for example, suggested that this is the most likely scenario. Dave Frankel, CEO of ZipDX also suggested that we need to accept transcoding, at least initially, to get HD Voice going. If not, we run the risk of not getting anywhere by waiting for the codec “war” to come to an end. Currently, it seems like G.722 is the most common choice for interoperability. As Jason Fischl with Skype pointed out, not only is the bitrate high but G.722 is also extremely sensitive to packet loss and therefore not a good choice for VoIP anywhere outside the managed networks.
Another discussion related to codec choices that created some debate was the topic of licensing. The codec landscape contains everything from open source, through license free proprietary codecs to standards with very complex licensing situations. For many, I believe, this was the first time they realized that even a codec that is labeled free is rarely truly free. Many free licensing agreements include marketing and IPR conditions that will be, by some, considered as having a high cost. In addition, indemnification from IPR claims does not come for free.
Most people I talked with agreed that the fastest way to create an end user pull for HD Voice is by widespread deployments in the wireless networks. It was very interesting to hear Benoit de Boursetty, Director FTNA, at Orange describe Orange’s deployments of HD Voice. Clearly this is an operator that takes HD Voice seriously and sees it as a key differentiator. Benoit said that they don’t see a distinct pull for HD Voice but on the other hand he claimed that it does increase customer retention.
The message in my own presentation was that the codec is just one piece of the HD Voice puzzle. I.e., in order to experience true HD Voice, all other parts, including acoustic hardware, echo cancellation, and other signal processing have to support HD Voice and provide the best possible quality. It doesn’t matter how good of a codec you have if the other parts of your solution are not up to par. I was glad to see that several other of the speakers at the event, including Martyn Humphries of Broadcom and Christian Stredicke of Snom, made the same observation. Christian also suggested a HD Voice label to be put on HD Voice capable devices.
For more details about the summit, check out the twitter feed here.
Jeff Pulver announced that the next event will be held on September 15 – 16 this fall.
Some additional blogs on this event:
http://pulverblog.pulver.com/archives/008925.html
http://blog.radvision.com/voipsurvivor/2009/05/20/can-you-hear-me-now-2/
http://www.mgraves.org/voip/2009/05/hdvoice-summit/
Tags: HD Communication Summit, HD-voice, VoIP, wideband
May 26th, 2009 at 10:35 am
Is it not true that G.722–and all such codecs trying to squeeze ever-growing amounts of data through pipes that are not growing at the same pace–is it not true that all of them are terribly susceptible to packet loss? Our experience at IPeak Networks has shown that the more powerful the codec, the more pronounced the damage when even a small measure of packet loss hits. We have also observed that when it comes to best efforts IP networks (and it is clear enough that everybody wants to use best efforts networks) packet loss is a constant if unpredictable threat.
By now you’ll have guessed that IPeak Networks is in the business of reducing packet loss (over best efforts networks in particular) but we are very interested to know if others would agree with our assessment of the situation. It seems that the legacy methods for dealing with packet loss do worse harm to wideband, HD, hi-rez and you name it media streams that the market is demanding. FEC and block encoding add latency, a no-no in the voice world and a travesty in real-time applications like video conferencing. We just can’t see any affordable way to get performance and quality out of a best efforts network without actually attacking packet loss right in the telecomms stack, the way we do it.
Speaking of videoconferencing, we note that there’s a good reason telepresence being restricted to custom rooms at this point. Yes, the lighting and the arrangements of cameras and microphones relative to people at the table is an importqant part of telepresence, but the other importsnt part is the fact that the high definition audio and video can’t be trusted to the public Internet. It takes serious big and seriously expensive pipes to move that kind of data around and keep it together.
May 27th, 2009 at 5:58 am
While some, as yourselves, may feel that G.722 is inappropriate on IP networks my experience with it has been good. The VoIP Users Conference (http://www.voipusersconference.org/) is a weekly gathering of telecom sys admins, VARs and end-users that has been holding a weekly one hour conference call. For the past several month we have been using the ZipDX (http://www.zipdx.com) wideband (G.722) bridge as well as the Talkshoe (http://www.talkshoe.com)narrowband (G.711) bridge.
We’ve had as many as 30 participants on the G.722 bridge, which is primarily accessible via SIP URI. We’ve has no serious issues of the nature you describe. Our regular members join using a variety of connection schemes, from DSL, Cable even wireless ISPs. Those problems that we have experienced are things that would be a problem for any sort of IP network, like excessive latency and poor/incompatible protocol/codec implementations.
Please feel free to you us any Friday at 12 noon EDT. Connect details are at the URL given above.
May 27th, 2009 at 6:00 am
One last thought…G.722 is at least as appropriate as G.711, which as we all know is widely deployed over IP.