There is a curious interface between science and the senses. Perception is often wrapped in psychological or emotional elements. This comes in many forms. It can be a group trying to share their impression of the taste of a particular wine. It can be people listening to music. Every persons experience of such things, being filtered by all that is us, both physically and mentally, is unique.
In a recent post I referred to a silly trend in very high sample rate music. This has been promoted by a variety of people, including HD Tracks, a music reseller that promises to provide “audiophile grade high-resolution recordings.” This term they use to generally describe digital recording a structure beyond the 44.1 KHz sampling and 16 bit linear word length defined by the CD format. They offer music at up to 24 bits and 192 KHz sample rate.
I’ve come to regard this trend as having very little merit. I came to this conclusion after many years dealing with digital audio production, after some experimentation and a lot of research. Like everything else on this site, I don’t profess to be any kind of expert, but I am happy to share my experience.
The samples and explanations provided are first rate. They clearly illustrate the merit in ultra-low-latency for some applications, as well as exemplify how CELT currently fares against other common codecs at a variety of bit-rates and with various types of source material.
As someone who’s passionately involved spreading the gospel of HDVoice I’ve been following the mailing list of the IETF CODEC Working Group. They’ve been working towards a new IETF RFC on a brand new wideband codec standard.
Starting with four submissions from Broadvoice, SpiritDSP, Skype and Xiph.org I think that they have made startlingly good progress over the past year. The group has actually arrived at a solution that provides for a codec that is a hybrid of SILK and CELT. It was recently announced on the mailing list that this new hybrid codec is to be known as “Opus.”
At IETF78 in Nagasaki, Japan the working group met to further their efforts. There’s a good recording of the session that, amongst other things, gives considerable detail about the hybrid nature of Opus.
There has been close co-operation between developers at Skype and Jean-Marc Valin of Xiph.org, such that they already have sample code running and have conducted some structured listening tests. The results of the listening tests are reported to be excellent.
It’s very interesting how the hybrid codec works. It can leverage both CELT and SILK principles acting on different frequency bands to generate the most optimal audio for a given bit-rate. It supports a wide range of sample rates from 8 KHz (PSTN narrowband) to 48 KHz (production quality audio) and bit-rates from 8 kbps to 128+ kbps per audio channel.
Opus may have a huge role to play in our transition away from the legacy PSTN. It’s very encouraging to hear that the CODEC Working Group is progressing so quickly.
The xiph.org foundation have just released a 30 minute video that they call A Digital Media Primer For Geeks. This video is a great overview of the basics of digital media. It’s nicely written and very clearly stated.
Given my interest in IP telephony and background in production/broadcasting I find that there’s a lot to like in this introduction. It nicely describes such things as; digital vs analog signals, audio sampling rates and alaw vs ulaw encoding….all good stuff to know. It even includes a few rudimentary examples of audio at different sample rates.
Available in a variety of formats up to 720p HD, I think that this primer is a great offering. It makes clear what might be for some fuzzy concepts in digital media. Highly recommended viewing!
BTW – I get a kick when someone actually says “Google is your friend” in such a presentation. It’s so true, but rarely said without being crude.