The VoIP Users Conference October 10, 2008 Was Great!

As is often the case some interesting stuff on the weekly VUC call was recorded in the session after the formal call ends. This weeks post call discussion about conferencing, Talkshoe, the cancellation of Skypecasts and HDVoice was really interesting. Thankfully Randy, our esteemed host, records the post-call chatter and uploads it when it’s useful. You can download it here.

Karl Fife, a VUC regular and fellow wideband voice advocate, corrected his assertion made a few months back about Polycom working some magic to enhance basic G.711 calling on the IP550 and IP650 phones. As he tells it, he had at one time made a call dialing out through an ITSP to a PSTN number, where the call was received via another ITSP and passed to a Cisco IP phone. The Polycom phone display had indicated “HD” and the call quality was excellent. This gave rise to the suspicion that Polycom had implemented some form of enhancement to the basic G.711 encoding used on the PSTN.

Polycom IP550 indicating HD, which means a G.722 encoded call
Polycom IP550 indicating HD, which means a G.722 encoded call

While at Astricon 08 he spoke to people from Polycom and found that what he thought he had experienced, while an excellent quality call, was likely not “enhanced” G.711 as he suspected. It was just a best case call between a pair of excellent end-points over well run networks. The fact that the Polycom phone indicated “HD” (as shown above) might be explained as simply a bug in the phones firmware.

I am especially happy to get this clarification as I had been pondering what might be going on under the covers for quite some time. Karl is a knowledgeable guy and one of the reasons why I’m drawn to the weekly VUC. It’s always an interesting group of VoIP geeks doing what they do best…Geek Speak!

For the record, I own two IP650s and they are simply awesome phones. Highly recommended! You can read a full review of them that I wrote earlier this summer right over here.

The pity is that I had to discover this weeks call after the fact as I was flying home from Seattle at the time it occurred. Earning a living is such a bother sometimes.

  • Many of us tend to use “PSTN” and “G.711” and “narrowband” as synonyms, and talk about “wideband” and “G.722” (together) as an alternative technology. That is not strictly correct. Here are a few things to bear in mind when contemplating these two types of connections:

    1) The frequency range carried over the PSTN is often listed as 300 – 3400 Hz (“narrowband”). Historically, this was pretty accurate. The microphones and speakers in the telephone sets themselves, as well as various filters, load coils, and other elements dictated that range.

    2) The G.711 codec, which was introduced when the network started “going digital” some decades ago, is actually capable of representing frequencies from 0 to 4000 Hz. All of the digital sampling in the PSTN is done at 8000 Hz; the Nyquist theorem (see Wikipedia) tells us that we have to sample at a rate equal to twice that of the highest frequency we want to capture.

    3) Even with this great codec and digital sampling, however, there are still network elements that may impose stricter limits. Many systems explicitly filter out signals below 300 Hz, to get rid of noise caused by 60-cycle AC (50 cycles in other parts of the world) that routinely “leak” into the analog portions of telecom systems. High frequencies are filtered out to avoid aliasing (the mis-representation of signals beyond the range of the sampling rate as lower frequency sounds). And in most modern end-points designed to work with the legacy systems, nobody has invested in better microphones and speakers — why bother, if there are going to be other elements filtering out the very low and high frequencies? (Sometimes the phones themselves contain these filters.) So “typically” a PSTN voice call, with traditional analog phones on each end (or even on one end) will have a “narrowband” frequency range of about 300-3400 Hz.

    4) If you connect two VERY GOOD digital phones (like Polycom 650’s) to each other, in G.711, you will get results better than “typical” narrowband. These phones have built-in immunity to 50 and 60 cycle noise, so those frequencies (and their harmonics) don’t have to be filtered out of the speech path. Presumably at the high end, the cut-off frequency is more precisely imposed. This, in combination with the better microphones and speakers that are in these phones (which have to be there to support “wideband”), means that we can come much closer to the 0-4000 Hz that G.711 can actually provide.

    5) So if you make a phone call from, say, an IP-650, via SIP, to a PSTN gateway, that then takes you over the PSTN to another gateway and another IP-650, and nobody along the way “monkeys” with the G.711 samples that are originated by each phone, you’ll have a call that sounds better than in (3) above. You still won’t have the higher (than 4KHz) frequencies that G.722 would offer, so the consonants will still be muffled. Remember that the simplest PSTN gateway is just moving the G.711 samples (8000 per second) between their packet encapsulation on the VoIP side, and the TDM circuit on the PSTN side.

    6) My understanding of (and experience with) the “HD” indicator on the Polycom phones is that it shows up when the phone is sending (and receiving) in G.722. So when you see this indicator, it means that the RTP stream AT THE PHONE is in G.722. That doesn’t guarantee that the stream is end-to-end G.722. There can easily be some other element in the connection (gateway, conference bridge) that is transcoding to something else. So the rule is: If you DO NOT see the HD indicator, then you know your call is not HD. If you DO see it, then your call MIGHT be HD all the way to the far end. The phone “negotiates” the codec with whatever is terminating the stream. In the case of a call that is end-to-end SIP, that could easily be the phone at the far end. But if the call is transiting the PSTN, there will be gateway(s) along the way, and the gateway terminating your packet stream will be the one with which your phone negotiates. The far-end phone, in this case, can’t “tell your phone” to turn its HD indicator on (or off). The indicator can, by the way, be toggled mid-call (by changing the codec mid-call). We (ZipDX) do this routinely in our wideband demo. (I have never myself, by the way, observed the HD indicator active when the phone was not, in fact, operating in G.722, but certainly there could be a bug along those lines.)

    There’s my 2 cents on “Why do my narrowband IP-650 calls sound better?”

    Also of note on the VoIP Users Conference call you reference was the discussion of numbering / addressing. This is indeed a giant issue, and an appropriate solution would go a long way in promoting the adoption of wideband.

  • David,

    Thanks for taking the timer to post such a detailed comment. It’s about a I suspected. I think that we need to get you as a VUC guest and go over the reality of wideband implementation on day soon. It would be very cool to actually host the call on ZipDX for this who have suitable end-points, bridging into Talkshoe for those who don’t and to sustain our usual workflow.

    Michael