The Scoop on Codecs for IP Audio

CodecIllustrationUsing the Internet for audio distribution makes sense, but the problem is a little like the holiday rush at the Post Office.

There are simply too many packets of data for the pipeline.

You need a codec to bit-reduce the audio stream. So what’s it going to be? AptX, Opus, G.722 or AAC, and if so, which version of AAC? We asked Charlie Gawley from Tieline, “The Codec Company” and a Wheatstone technology partner, to fill us in on Opus, the EBU ACIP standard, and how the AES67 factors into the use of codecs for IP audio delivery.

WS: Before we get started, I have to ask you about the new Opus audio codec that everyone’s talking about. What is your experience with this codec?

CG: This algorithm is extremely robust thanks to development by a number of programmers from Xiph.org, Skype and other partners of Xiph. At the low end, it will do voice at the equivalent of G.722 but at one-fifth the audio bit rate. That’s a huge benefit for anyone wanting to send audio over a wireless network, which, as you know, has severe bandwidth limitations. You can run Opus at 14.4 kbps and have near the same audio quality as G.722 at 64 kbps.

It can also go to the extreme, where it will be used for full band audio at, say, 384 kbps. We added Opus as one of the many codec options for our Genie distribution and Merlin remote products because it’s so dynamic and robust.

The actual Opus algorithm is made up of two algorithms: one that’s called SILK, which is for low bit rates, and the other is CELT. Some folks say it’s like a hybrid between AAC and AptX, but it’s license free and royalty free.

WS: We’ll have to keep an eye on Opus. But, not everyone is going to use Opus, right?

CG: Right. We have a myriad of coding algorithms because there are a myriad of broadcast environments. For example, a network here in Australia might have a standard to implement MPEG-2 across the network. Then again, you might have another network, say, in France, that doesn’t like E-AptX or AAC; they just don’t like the sound of it. Also, you will find that AAC is widespread in the U.S. Whereas Enhanced AptX might be used more in other markets, AAC is still popular because it is known for its high audio quality at low bit rates, and at reasonably low coding delay. When broadcasters began migrating from ISDN to IP, it remained popular because it could produce a constant bit stream.

Then again, some people might think that music sounds better over Enhanced AptX than AAC, and of course, there’s the preference for G.722 for mono voice. It’s such a subjective thing. That’s why you need all these algorithms to match all those different requirements and preferences.

WS: Tell us about the European EBU ACIP standard. Most of us know that it is a standard adopted by codec manufacturers for audio contribution over IP, or the delivery of audio, but what is this standard really all about?

CG: We were part of the committee for this and the first Non-European to implement and support the standard, which is known as the 3326 ACIP standard. So broadcasters can integrate a number of IP audio codecs into one environment. ACIP stands for Audio Contribution over IP, and this standard addresses the use of IP connections for the purpose of streaming audio to production facilities and also the withdrawal of ISDN services that had been used for contributions in the past. All the audio codec manufacturers got together and standardized on three codec categories, starting with the codecs identified as needing to be implemented in order to achieve interoperability. These are the mandatory ones. Next are the recommended codecs that are advisable for interoperability, and finally, we identified an optional category of codecs that could be implemented for better interoperability.

(Mandatory audio codecs include Linear Audio (PCM), ISO MPEG-1/2 Layer II, ITU G.711 and ITU G.722. Recommended audio codecs are MPEG-4 AAC, MPEG-4 AAC-LD, and Standard/Enhanced AptX. And, finally optional audio codecs supported are ISO MPEG-1/2 Layer III, MPEG-4 HE-AACv2, AMR-WB/AMR-WB+ and the new Opus algorithm.)

WS: What about codecs and AES67, that other interoperability standard that has more to do with synchronization and timing?

CG: Bandwidth is the liberator of AES67, but it is also the inhibitor. What I mean by inhibitor is that you don’t have enough bandwidth over the public internet. There will come a day when that can be achieved, but today, there are bandwidth limitations to the internet and just as important latency.

So if it’s a microwave link from your studio to your transmitter, you've got lots of bandwidth and you can run AES67 straight out to your transmitter. Or, if you have dark fiber, you can run an AES67 stream straight out between a set of (WheatNet-IP) BLADEs. Or you can run it from two (Tieline) Genies with WheatNet-IP, and use those BLADEs for other things – like if microwave links are taken out by lightning or something happens to the dark fiber and you can only rely on the public internet, then that’s where the codec can take over and keep you live on air.

But not all those options are going to be available all the time, which is why people use the public internet wherever they can. And that’s going to require a codec to solve the fundamental requirement for audio over IP: a constant bit stream. Otherwise, it starts to fluctuate and that results in audio break up and audio loss.

WS: Thanks, Charlie.

Editor’s note: Tieline is a Wheatstone technology partner. Its Genie distribution unit is available with a WheatNet-IP audio card inside for easy and seamless integration with a WheatNet-IP audio network.

Site Navigations