VoIP Calls Double In Quality

anthm writes "From Newsforge and LinuxPR
FreeSWITCH, an open source soft-switch and IVR platform, have announced that they can support 16khz audio calls thus doubling the potential voice quality. They have had successful tests with a conference bridge, a pass-through SIP call and an IVR that reads RSS news feeds with the Cepstral Text-To-Speech Engine."

Voip-Info.org has a good list of business VoIP providers.

Please get the rest of the telcomms to follow.

[rob_squared]

Everything else is stuck at 8khz, so unless your call uses this service end-to-end, there's going to be a downconversion if you're calling someone on a land line. And you'll be stuck with 8khz if you get any calls from someone not on this service.

Still, its a good piece of news, onward and upwards.

*crosses fingers* Please nobody mention video phones. *crosses fingers*

Define: IVR

[theGreater]

Google gives the definition of IVR [google.com] as Interactive Voice Response.

So I knew what one was, I just didn't know there was a TLA for them. This inane personal revelation brought to you by the captcha "accuse".

-theGreater.

Re:Please get the rest of the telcomms to follow.

[anthm]

Yes, you are correct. The benefit comes when both ends of the call are using a 16khz device. Situations where you are connecting to the PSTN would obviously be better suited at 8khz. The media description on the pstn gateway would advertise only 8k so the client would know better than to operate at a higher frequency.

Re:Doubling? hardly

[jdmicklos]

The only real advantage to adding in "unused" octaves is in order to transmit overtones. Overtones shape the sound you can hear even though they may not be hear directly. Think about it as if you were to have a G note at 120 dB playing in an octave that you couldn't hear. It would still cause all things around with a fundamental frequency that is a "G" to vibrate as well as color certain audible noises.

Only a slight improvement

[riflemann]

Actually, I've used Asterisk to pass through 24KHz Speex encoded audio - very impressive sound quality, but only works when the SIP channel is client to client.

In theory a SIP server doesn't need to know all of the codecs a client supports - the clients themselves negotiate any compatible protocol.

Of course, if the sip server puts itself in the path (such as when it needs to pass through to PSTN or firewalled clients), then 8KHz is the (till now) maximum supported rate.

Nothing special here...

[Anonymous Coward]

They're just using a higher quality codec than G.711 (which is the standard for the back-end digital phone system).

The phone people (probabably AT&T) chose that standard since it gave pretty good voice quality given the limitations of current technology.

People are generally happy with the voice quality of the phone system - which is different from the voice quality of the last mile - the analog copper loop to your house, or CDMA/GSM/TDMA to your cell phone.

It's highly unlikely this new codec will catch on - the installed base of G.711 phone systems out there is enormous.

Re:Marketing BS

[anthm]

FYI: 20ms of 16khz audio (the typical size of 1 RTP packet) encoded with the Speex Codec http://www.speex.org/ [speex.org] is 43 bytes. 20ms of 8khz audio encoded with the Speex Codec http://www.speex.org/ [speex.org] is 29 bytes which is only 1.4 times as big as it's 8khz counterpart. 20ms of 8khz g711 is 160 bytes so with speex at 16khz, you can still fit 3 calls in the same amount of bandwidth that it takes for one 8khz call. The biggest overhead in VoIP is the various headers on each RTP packet per level of encapsulation, not the size of the payload.

Because it covers almost all of the human voice

[Sycraft-fu]

Our voices don't have that wide a frequency range, there's little up in the high frequencies. A voice sample recorded at 22kHz (11kHz frequency range) is very hard to distinguish from one recorded at 44kHz (22kHz frequency range). In fact you'd need to be using a fairly good mic to really get much of the higher frequencies anyhow. 8kHz works since F1 and F2 (the frequencies of the first two peaks in the harmonic curve) fall under 4kHz for essentially all speakers. F1 and F2 are what we primarly use to determine vowel sounds and thus are what's realy relivant. Well with an increase to 16kHz you get F3 and even F4 which leads to pretty natural sound as far as most listeners are concerned. Past that, there's just not a whole lot that affects your perception of speech.

The reason for chosing 16kHz is probably simply that it's twice what you have before. Thus if you are interfacing with an old system that doesn't support it, just discard every other sample, no sample rate conversion needed (which is CPU intensive).