Already dominant elsewhere, audio over IP is poised to take over the studio world. We explain how it works — and why there are so many protocols!
Today, anyone wishing to add a high-quality, low-latency, multi-channel audio interface to a computer is spoiled for choice, with devices available in a whole host of formats: PCI, PCI Express, Cardbus, Expresscard, FireWire 400 and 800, USB Full-Speed, High-Speed and Super-Speed and Thunderbolt 1 and 2. Until recently, however, the oldest and most ubiquitous connection format of all was absent from this list. It wasn’t until 2011 — 26 years after the publication of the IEEE 802.3 standard — that the first DAW-focused Ethernet recording interfaces appeared. Yet, in principle, Ethernet has unique advantages over other formats, and it has already come to dominate the worlds of installation and high-end live sound.
So why has it taken so long for Ethernet to reach the recording studio, and what key benefits does it have to offer? To find the answers, we need to take a look at the development of Ethernet itself, and that means going back, back through the swirling mists of time...
It’s May, 1973. In Britain, VAT and the Austin Allegro have recently been introduced. In America, Skylab is launched, and Richard Nixon confesses to his role in the Watergate cover-up. Meanwhile, at Xerox’s Palo Alto Research Centre (PARC) in California, a 27-year-old recent PhD graduate called Robert Metcalfe has been tasked with adding networking functionality to the company’s new Alto computer — the machine whose revolutionary graphical user interface would famously ‘inspire’ Steve Jobs. Metcalfe delivers a memo to his employers in which he coins the term ‘ethernet’, and outlines the essentials of what will become the world’s pre-eminent local area networking technology.
The protocol which Metcalfe described in his memo is now known as Carrier Sense Multiple Access with Collision Detection, or CSMA/CD, after the title of the first IEEE 802.3 standard. The context in which CSMA/CD operates is that of a fully distributed, packet-switched network consisting of autonomous nodes connected to a shared transmission medium (see Figure 1). The autonomy of the nodes means that any given device is free to transmit at any time, while the sharing of a common medium — originally a single coaxial cable — means that if two (or more) nodes decide to begin transmitting at the same time, those transmissions will collide. Since collisions effectively randomise the electrical state of the wire, all such transmissions are void. The only recourse is to try again (which again entails the risk of failure). Clearly this state of affairs has serious implications for the efficiency of a network.
CSMA/CD seeks to mitigate the problem in two ways. Firstly, all nodes continuously monitor the transmission medium for the presence of a signal (carrier), and will only attempt to send data when none is detected. That’s the ‘carrier sense’ part of the protocol. Carrier sensing alone, however, provides only a statistical defence against collisions, the reason being that signals take time to propagate throughout the network. The fact that a node doesn’t detect a carrier doesn’t mean that another node has not begun transmitting, merely that said transmission has not reached the first node yet. So while carrier sensing reduces the likelihood of collisions, it doesn’t eliminate them, and this is where the second part of the protocol comes in. When any node in the process of transmitting data detects a collision, it stops transmitting that data and instead sends a predefined jamming signal. The duration of the jamming signal is such that it is guaranteed to propagate throughout the entire network, and thus be received by every node. Upon detecting the jamming signal, each node will ‘back off’ for a random period before making any further attempt to transmit. In effect, the network is ‘reset’ as a means of quickly recovering from a collision. The randomisation — Metcalfe’s key insight — mitigates against a ‘repeat’ of the same collision.
Naturally, the description above omits or skates over many technical and historical details for the sake of brevity, but hopefully it provides a clear enough picture of the original Ethernet protocol to make one thing apparent: It was fundamentally unsuited to the job of transporting high-performance audio. However, while CSMA/CD may have been inimical to real-time applications, some found the underlying infrastructure of Ethernet too attractive to ignore — particularly as prices fell and capabilities increased — and a number of companies adopted Ethernet as a vehicle for their own proprietary audio networking protocols.
The first of these Audio over Ethernet (AoE) systems, as they came to be known, was CobraNet, developed in the mid ’90s by a small Colorado company called Peak Audio, and later acquired by chip makers Cirrus Logic. CobraNet works by designating one device on the network as the ‘conductor’, while the remaining devices are known as ‘performers’. Every 750 s, the...
You've only read 10% of this article, so to continue reading...
Option 1: Login to read this article if you have a Digital Subscription or Industry Controlled Circulation account
- To read the full article online (in HTML browser format), please LOG IN at the top of this page.
- Note: Your Digital subscription does not include downloadable PDF articles free of charge.
Option 2: Buy a Digital sub from our shop
- A Digital sub can be bought from our Shop and used immediately, or contact our Subs staff to discuss an upgrade price to add Digital access to your existing Print subscription.
Option 3: Buy and download this SOS article in Adobe PDF format
- Buy this article now and immediately download the PDF file to your computer.
- PDF articles look identical to the printed magazine layouts (but exclude advertisements).
- Note: Some shorter articles don't always have a PDF version.