In most modern recording studios, audio is recorded and processed primarily in the digital domain. And even though many of these functions are now performed within a single computer, the use of multiple computers and other outboard equipment is often inevitable. In that case, audio is commonly transported from one device to another in any of several different forms, including S/PDIF, AES/EBU, Lightpipe, and analog.
Commercial studios and large, live venues have addressed the problem of audio transport by using networks to carry digital audio between multiple devices, but those systems are generally very expensive. However, one such system will soon become available for smaller studios that have more modest budgets. Developed by Peak Audio (www.peakaudio.com) for large commercial installations, CobraNet has now been implemented on a single, inexpensive microprocessor chip by Cirrus Logic (see Fig. 1), which acquired Peak Audio in 2001.
CobraNet uses Fast (100 Mbps) or Gigabit (1 Gbps) Ethernet to carry uncompressed digital audio between devices on the network. Audio can easily coexist with other data on the network, making it ideal for studios that already have an Ethernet network for high-speed Internet access and other types of data sharing.
Ethernet is designed to carry “bursty” traffic, not continuously streaming audio. On the other hand, Ethernet is the most widely used LAN in the world, which provides a preinstalled infrastructure on which to base an audio network. So Peak Audio decided to create an isochronous protocol that divides the streaming audio data into time-stamped packets and guarantees a certain level of network performance in terms of available bandwidth, latency, and jitter.
As a result, CobraNet supports three well-defined latency settings: 5.33, 2.66, and 1.33 ms. These are the available isochronous cycle periods, which define the time required to buffer the streaming data into packets; shorter latency settings mean fewer audio channels can be transmitted at a time.
CobraNet can accommodate audio sampled at 48 or 96 kHz with a resolution of 16, 20, or 24 bits. The packets that carry audio data are called bundles, each of which can accommodate up to eight channels. A single 100 Mbps link can transport up to eight bundles (64 channels) of 48 kHz, 20-bit data in each direction for a total of 128 channels. As you might imagine, being ten times faster, Gigabit Ethernet can support up to 1,280 channels of 48 kHz, 20-bit data. Bundles can be multicast (from one source to many receivers) or unicast (from one source to a single receiver).
One of the devices on the network, called the conductor, controls the system timing by sending a master clock signal. This role can be filled by any device on the network; if the current conductor is switched off or disconnected, another device on the network can take over within milliseconds. Clock accuracy is ±0.25 sample period (5 µs at 48 kHz), and cycle-to-cycle clock variation is maintained at less than 1 ns. The bottom line here is that the system exhibits very low jitter.
Current CobraNet licensees include many familiar names, such as JBL, Mackie, Crown, Peavey, dbx, QSC, Rane, DigiTech, Shure, Soundcraft, Symetrix, and Yamaha. Many of these companies are working on CobraNet devices for smaller studios, including mixers, DSP boxes, amplifiers, and powered speakers; we should see the initial fruits of that labor at Summer NAMM 2004. Not only will CobraNet carry digital audio to and from these devices on a single cable, it will also allow you to control and monitor them from your computer, which should simplify and integrate the operation of your studio. What more could you ask of a newly available technology?