Skype to give away wideband audio codec Skype this week said it will soon be offering royalty-free licenses to its new SILK wideband speech codec to interested third-party developers and hardware makers. The wideband codec recently debuted as part of Skype 4.0 for Windows (with a Mac version coming in April.) With a claimed 400 million Skype users registered …
Greater availability of wideband codecs is a very welcome development - the quality improvement is like comparing speech on a regular telephone vs. FM radio, and helps greatly with intelligibility and naturalness of communication. Even better if Skype can do this at bit rates of 40kbit or lower and good packet loss resilience.
What is remarkable is how long it has taken for high quality telephony to take off. Basic high quality codecs with a bandwidth up to 7kHz at 64 kbit/s data rate were available more than 20 years ago, and would have been usable over early ISDN lines even before widespread adoption of Broadband. Anyone who had used them would wonder why people put up with poor quality telephony speech, particularly on loudspeaker phones. I suppose the awful quality of many mobile calls has made people think that telephony speech on advanced systems will always be lousy.
Some of us living away from "home" find that <insert any IM service here> does video calling perfectly well anyway without having to fanny around with Skype.
Most of SkHype's success revolves around the generally mistaken view that "to do this you need Skype".
I'm not sure about the quality argument here. My SIP service sounds as good, if not better, than the landline service. But then the landline service is probably being VOIP routed using standard SIP protocols and the usual codecs "behind the scenes" anyway. The words "fit for purpose" spring to mind. Why the f*** would you need high-quality audio for chatting on the dog 'n bone? Is there a market for auditioning Operatic Divas by telephone?
@David Hadworth: I agree
However, I've been wondering how a conversation that seems recorded at 8kHz-12kHz even has the range to encompass the plain old 300Hz-3.4kHz. I would have thought that if you don't have the signal there in the first place, the if you have one non-compliant system along the way which works at a different (old) frequency all you'd get is a lack of sound.
I'm guessing that Skype have got the units wrong on the original posting.
Surely only the two endpoints of the conversation need to support the codec? I can't see any need for steps in between to know how to decode the audio data, they just need to be able act as a transport layer. By analogy - you don't need to upgrade all your routers every time you invent a new application level communications protocol.
I would guess that the press release comment about 8Khz says "improving audio bandwidth going from 8 kHz to 12 kHz" Is referring to the bandwidth changing dynamically, not the frequency range.
Presumably there is some kind of low pass filter that will remove rumble and DC but a bandwidth of 8Khz would suggest that the frequency range goes from 0-8Khz
If it's for speech, there's plenty already, and most of them do perfectly intelligible speech without needing a whole 12kHz of bandwidth (though I note the article doesn't mention the actual data rate required).
If it's for reasonable quality audio, ditto basically.
Skype needs a bit more free advertsing maybe?
Andy could be right about the figures being the sampling rate. Telephone systems typically have a top end of around 3-4KHz as the high frequency response. The sampling frequency will need to be double this, coming in at around 8KHz.
So the figures Skype could be quoting are an increase in the sampling frequency from 8KHz to 12KHz, but that would only increase the audio bandwidth up to 6KHz. Not a lot! And I would suggest probably not worth bothering with.
Perhaps it is just a typo and the audio response goes down to 8Hz, but this is actually determined by the audio circuitry in the PC and almost certainly doesn't go down to 8Hz.
The software probably doesn't care what the lowest frequency is, and if it could go down to near DC it would. It tends to be the upper frequency which matters, because that is set by the sampling frequency.
Are they still using their resource sapping P2P technology? That is the question.
There's a reason why most hardware manufacturers go for SIP instead, despite consumer awareness of SIP being low. It's standard, it's cheap, and easy to implement in embedded devices. Not to mention easy to support in ISP QoS routers and servers to ensure call priority and quality.
Skype is a kiddies Windows app that has been shoehorned into a few embedded devices with some difficulty (requiring external boxes, a lot of power in the device itself, or a 3rd party proxy server to do the main work).
The other question is does 'open' mean they'll allow peering with SIP services to allow SIP in and out calls (for free)?
"Speex can do that, for free and without having to install spyware on your computer."
It may very well be capable of doing that, problem is, none of the 37.5 people who have that
codec on their end are ones I wish to speak to.
Oh goody! It's free. Let me see, how much money does that save me over using Skype?
I didn't have to install spyware. If you can prove Skype to be spyware, there's money to be
made by suing them over their claim to be free of adware, spyware and malware.
>There's a reason why most hardware manufacturers go for SIP instead
Yes they want to sell you expensive SIP routers and charge you consultancy to configure them.
The nice thing about skype's resource sapping P2P technology is that you don't have to send a CISCO certified engineer to 10million homes to configure their router.
So you ring up your granny and talk them through installing and configuring Asterisk and SIP as well as getting their BT wireless router to port forward the SIP packets to their machine ?
The only advantage of Skype is their P2P stuff that means two people behind a NAT/Firewall can talk to each other - yes, everthing else about the app is crap - but they solved the bit that was stopping customers using voip.
Main thing with SIP is it needs packaging by operators really to sell it in an easy to use form. i.e. by using an app on a phone or PC that the consumer just installs and it configures everything for them.
e.g. Gizmo on Nokia mobiles. Includes a Gizmo account and configures pre-existing SIP services, with a funky UI wrapped around it. Makes it easier to use for consumers.
Problem is there are so many operators and few are popular. Some of the bigger names that are associated with traditional telecoms companies charge too much also. Skype is unfortunately what people associate with VoIP even though it's a single service locked to calling Skype users (for free) and proprietary inefficient protocol.
The original bullet from Skype reads
"- improving audio bandwidth going from 8 kHz to 12 kHz" which is lazy-English for "Raising the upper frequency limit from 8kHz to 12kHz", .......... no mention of the lower end.
For any pedants watching, 8Hz lower limit would be a waste of engineering resource: there is an empirical "rule" which works quite nicely and is based loosely on what "sounds right". This states that the product of lower & upper frequency limits (actually more like -3dB points) should be 1E6, i.e.have a geometric mean of 1000 (so, 300 to 3400 is fine, as is 100 to 10kHz etc. etc.) If you flout this guidance big-time, things can sound quite odd, as you can sample by listening to a European AM station on Hi Fi speakers.
So, for the Most Satisfying Listening Experience, we'll settle for 80Hz-12kHz ta very much
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
