Recording - January 2011 - 26

Daisy-chaining multiple audio devices and disk drives seems like a wonderful and convenient way to hook up a music studio computer... until something fails. Learn the pitfalls of this approach, and simple rules to avoid them.

By Mike Metlay

EVERYBODY OFF THE BUS!
are being transferred won’t get garbled... the worst that will happen is that they take a few seconds longer to copy. When we talk about video over cables, often that’s not live video data but video files from a camera’s onboard data storage, which again can be interrupted without harm. Streaming audio and video from the Internet are at very high compression, and most of us accept the occasional stutter as second nature. Moving data to and from “the cloud” can happen as fast or slow as it wants to; you may grumble at a file taking a few seconds more to open, but that delay isn’t going to hurt the file itself, right? With professional-grade audio, though, these interrupts can be disastrous, because audio data has to be isochronous (each data word stamped with a very specific time) and delivered reliably. If other devices grab the bus and the computer has difficulty making sure that the audio isn’t prioritized properly and getting through in one piece, you’ll get dropouts, stuttering, or no audio at all. In other words, no matter how well it works for folks who’ll never notice a short delay in file transfers, “just daisychain and you’re good” is a pretty good recipe for significant audio problems. Now, it’s true that some companies allow you to daisychain multiple interfaces on a single bus for more tracks—but that’s multiple units of theirs, which were designed and tested to share a bus and work with one another in a seamless and reliable fashion. Add a couple of hard disks to that bus, or an interface from another company, and watch what happens. Fortunately, this problem has a simple solution— don’t daisychain. Just put every data-heavy device on its own bus. Unfortunately, there often aren’t enough buses to go around. Missing the bus If you want to hook up an audio interface to your computer along with external hard disks, input devices, and control surfaces, you’ll want at least two separate buses to work with, and some would argue that three is even better. But nowadays many computers, especially laptops, may have only two that are solidly reliable (if they’re PCs) or may only have two or even one, period (if they’re Macs). As we do more with wireless connectivity, physical ports go away, and the audio user is the first to suffer. If we’re to get reliable results from our audio systems, we need to make proper use of the buses we have, knowing what they’re best at. What are our choices? For current desktop machines, there’s SATA/eSATA, PCI (which, for our purposes, includes its modern variant PCIe), USB, and

Recently I had reason to revisit a topic that’s been a sore spot with me for quite some time now. The Talkback question and answer I wrote up concerning the fellow who was having issues with his audio system’s data throughput (Talkback, November 2010) caused me to revisit the entire question of connectivity to a music computer, and how the community right now has a set of choices that aren’t entirely appetizing. The fact is, with the boom in wireless connectivity and the rise of handheld data devices and netbooks, the “computer with lots of boxes wired up to it” is starting to look pretty old-fashioned. Which is fine, I guess, if you’re a high-powered business executive or young media surfer, but tends to push us musicians into a corner. Let’s consider the plight of the musician trying to get his computer to do what he wants with his audio gozintas and gozoutas, and how the modern fashionable computer often doesn’t do it very well. Buses and bandwidth The computer’s CPU has various data highways, called buses, that it uses to connect itself to the outside world. Each of these buses has a certain amount of data it can push through to the CPU for processing or take back out again before it starts to choke. We can call this the bus’s bandwidth or throughput. In drastically oversimplified terms (please, no hate mail from IT people!), each bus can coexist without stepping on the others in an attempt to get the CPU’s attention, or at the very least, we can expect the CPU not to trip over itself when talking to all these buses even if each one is pushing a lot of data. They’re separate highways into the hub of the system, and if each one is happy, they’re all happy, and we’re happy. Okay so far? Now theoretically, in a perfect world we should be able to stack devices on any given single bus until we get close to the maximum rated bandwidth and never see any problems. In the real world, however, devices on a bus interrupt one another; some items have a higher priority than others, and the computer’s operating system tries to keep such devices from coming into conflict by making them take turns. What might seem to a human like a steady flow of data from multiple devices is actually, at the speed of the computer, a series of bundles of data interrupting one another, stopping and starting again, with the CPU keeping track of what’s going where and stitching all the bundles together again when they get where they’re going. For the vast majority of computer users, these stops and starts aren’t an issue. Hard disks can interrupt one another hundreds of times per second, but the files that
26
RECORDING JANUARY 2011

Recording - January 2011

Table of Contents for the Digital Edition of Recording - January 2011