CUTTING EDGE

Streaming Digital Audio

Published in SOS September 2002
Printer-friendly version

Reviews : Computers

While streaming digital audio via network cables isn't as easy as you'd think, there are some interesting developments in this area that could change the way we connect our studios in the future.

Whether we like it or not, networks are becoming more pervasive in our homes and studios — even people who'd rather entertain themselves with a root canal filling than log onto a network will soon find it unavoidable. While your studio isn't going to stop working if you ignore networking technology, you may, in a surprisingly short time, find you're missing some great new ideas for working with audio.

Until now, the use of networks in studios has been limited, although some digital mixing consoles use Ethernet to connect their control surfaces to the digital audio processing hardware. However, the user wouldn't necessarily be aware of this — and that, in my opinion, is exactly the way it should be.

You can find more information about CobraNet at www.peakaudio.com.

It would be great if you could use networks to replace all the complex wiring and routing in a studio, but there's a problem: networks are intrinsically unsuitable for the real-time, low-latency delivery of audio. Sending audio over a network is not the same as simply sending a signal over a wire. This isn't just because the audio is digital, though — there are already enough formats to prove there's no problem in moving multichannel digital audio around a studio. And there's no problem with latency or signal clocking either, since the basic rules about having a master clock and slaving all other elements in a configuration from it are well known and easily applied.

Where the simple 'signal on a wire' arrangement falls down is in its flexibility. For example, MADI (Multichannel Audio Digital Interface), a 15-year-old multitrack digital audio format doesn't allow you to isolate a track, or a group of tracks, without decoding and re-encoding them all, meaning that it's strictly 'point to point'. This is fine for dubbing between digital multitrack devices, or connecting a digital multitrack tape deck to a digital mixer, but it limits MADI's potential as a digital distribution system. And that's where networks begin to show their potential.

Every network architecture is built around the idea of sending packets of information to individual users or addresses. It's very like the way the postal service works. And of course, one of the most distinctive things about the postal service is that you definitely can't predict with any precision when your post is going to arrive, even if everyone were to post their letters and parcels at the same time or at regular intervals. As you might expect, this doesn't bode well for the timing-critical delivery of real-time audio in a studio.

Sending audio over a network isn't impossible, though, and Peak Audio's CobraNet is making some interesting progress in this area. Although CobraNet uses an industry-standard network infrastructure, the system inserts special 'timing' packets (called 'beat' packets) 750 times a second, which can be used at the destination points to reconstruct the timing data. Using this technique, Peak Audio appear to have achieved a level of performance not previously thought possible using Ethernet. When the beat packets are received, they're fed into what I imagine to be a Phase-Locked Loop (or PLL), which is a kind of electronic flywheel used to regenerate the original clock signal. Timing is accurate to plus or minus a quarter of a sample period, and the actual jitter is less than a nanosecond, which can be further improved by buffering and re-clocking.

CobraNet is different from conventional audio streaming in that it doesn't use audio compression — once the packets have been decoded, the audio data that comes out is the same as the audio data that went in. Resolutions of 16, 20 and 24-bit audio can be accommodated, and while 64 channels of 20-bit audio can be sent via a 100Mbit/s cable, the system is compatible with Gigabit Ethernet, which has 10 times the capacity. As you'd imagine, CobraNet's signal routing is particularly versatile: any source can be instantly routed to any destination, and connections can be one-to-one (unicast) or one-to-many (multicast) — beat packets are always multicast, of course.

CobraNet is good enough for work in recording studios, although most of the current installations are in stadiums and conference venues. Depending on your musical tastes (or sense of humour), you may have already heard CobraNet in action since it was used to distribute audio around this year's Eurovision Song Contest in Tallin!

At this stage in the evolution of 'audio over Ethernet', I'm not sure I'd want to rely on an all-network approach, although there's no reason why it should be any less reliable than a conventionally connected studio. However, there are other ways in which networks could change the way our studios work, based on the thing that networks do best: communication.

You only have to look at the history of MIDI to understand what kind of influence a standardised digital conection between equipment can have, particularly on the physical architecture of studio devices. Keyboard-less synth expanders, like the classic Korg M1R for example, make very little sense without the ability to connect them to keyboards or sequencers. However, via MIDI they become fully functional modules within a studio or even a live set up.

Since networks are capable of far more than MIDI is their use should, at least in theory, lead to even more dramatic changes in the form and function of studio gear. With fast enough networks, it should be possible to distribute not only audio, but audio processing across disparate pieces of equipment. And if there's suitable 'intelligence' in the network, studio devices might be able to negotiate exactly how they're going to work together, potentially leading to a result that's very much greater than the sum of the individual parts.