STE VE CASSIDY
The Royal Albert Hall’s peculiar acoustics present a unique challenge to those attempting to create perfect audio. And the key is VoIP
The Royal Albert Hall’s peculiar acoustics present a unique challenge to those attempting to create perfect audio – and the key is VoIP.
What connects the Royal Albert Hall with Paralympic cycling? This month, the link was your humble correspondent, who was foolish enough to agree to meetings on both subjects within the same day. Not a show-stopping issue at first glance, but it definitely gained some tension when I figured out the Paralympic event was in Eindhoven.
Eindhoven represents a travel lacuna if you’re starting from the western end of London. Yes, you can hop on a plane, but it won’t be landing especially nearby. So it was that the day saw a somewhat deranged Cassidy in a Volvo, swiftly moving from one to t’other. Actually, that wasn’t as hard as it sounds, because the Royal Albert Hall was taking invited visitors at the ungodly hour of 7.30am to show off its new sound system. Any later in the day and it would be interrupting booked performances because there was essentially no downtime to the project: the team had to roll in the replacement speakers in a way that let performances carry on using hybrid arrangements of old and new.
Come on, Steve. Speakers? Really? Where’s the tech in this? Well, okay, not only speakers. While I was busy misspending my youth during the 1980s and 90s, we thought we were posh if the rented PA gear came with XLR connectors and trip cut-outs to ward off the fumbling fingers of budding ravers. In those years, the major risks were from clipping amplifiers and busted up speaker cones, which could do bad things to the unwary, youthful ear. But speakers have no computational element to them, right? Much less a network.
Not strictly true. The Royal Albert Hall seats 5,000 people and is impressively round. This
is an acoustician’s nightmare and, to make it that bit worse, the architects added a glass roof (now obscured) and surfaces conducive to the worst kinds of echo. To simply wire every seat with a headphone socket would have been impressive, but massively wrong-headed: the overall system must even out the experience for every seat position. Several other journalists asked whether in fact this was just a network, with 400 devices on it, all with a single-purpose specification of “speaker” despite being plugged into a complete network. The Albert Hall techies immediately looked rather pained.
It’s not really like that, they said. The system is a network at the centre, with the various amplifiers and groups of speakers being addressable from the central DSP racks. Their interest wasn’t in making Ethernet carry sound but in using computers to rectify the foolish mistakes of Victorian architects. A few more hints and I suddenly realised what they were doing: they were treating the hall like a giant phone system.
Phones these days almost invariably use VoIP, and VoIP is frankly a perversion of Ethernet as we know it. Ethernet was designed to both cause and then recover from
collisions. One of the core protocols is CSMA/CD (carrier-sense multiple access with collision detection). Zealots will be ready with that upheld finger, as in a twisted-pair Ethernet there is no multiple access. But all the code that deals with collisions, retries and dropped packets is still in there, and occasionally damage or malfunction will bring it to the fore.
VoIP (and the extended architecture used in the Royal Albert Hall) doesn’t like any of those events, because sound is one of those annoying senses that fits poorly with the techniques that work so well for vision. Sound is the sense that doesn’t turn off when we sleep. That’s why we have alarm clocks, not alarm lights. It’s the warning sense, the “what’s that noise?” sense. It doesn’t tolerate gaps in transmission, down to impressively short intervals of resolution. A 1% glitch in a JPEG can go completely unnoticed; a 1% glitch in Beethoven’s fifth, or The Cure live at Glastonbury, is headline news.
So the conundrum for the Royal Albert Hall is this: you have to control and present a perfect stream of bits, managing the flow of traffic over the network so that everything that has been sent is received. That includes the non-network part of the wiring that links the amplifiers to the speakers. It might be possible to do this with Ethernet right to the back of the speaker, but then they would have been obliged to deliver the amplifier power and a decoder to every speaker location. Even if their architecture has one satellite amplifier for every ten boxes (which would be 30 speakers), the logic is good, and is subordinate to a problem that had the hairs on the back of my neck standing up.
They’re not just delivering noise into a row of little rooms. Those rooms are in a circle and sound from one will propagate clear across the open space to the opposite side more easily than it will travel left/right along the row. This isn’t ideal for music. So the major part of the effort is spent in modelling the acoustics of the space, and shaping the delivery of the sound to a relatively small groups of speakers (I think they said three boxes is their smallest configurable unit) to minimise echo, delay or distortion. This is DSP (digital signal processing). The couple of racks of computers hidden in an upper room calculate which cell gets which audio packet with what level of delay, so that you get the same sound wherever you sit.
Why does this creep me out? The delay intervals are well into the milliseconds – the unit of network performance much prized by gamers. Not only is the hall’s audio network clean, it’s also unbelievably fast – and, to a certain degree, self-aware. Most networks just chuck out the packet and wait to see if anyone grabbed it. In an audio deployment such as this one, that’s nowhere near good enough. The DSP has to know how long the overall data transit time is, from the mixing desk to the furthest speaker, in milliseconds – and adapt what’s sent where and when, based on that information.
To guard against audio naiveté, I took Ron with me to look the system over. Ron is one of my best friends from the pre-millennium days, but more important than that, he’s a DJ. That he trained as a structural engineer was a considerable bonus. Once the Albert Hall people were running some demos (string quartet; bit of Peter Gabriel; some mid-quality technopop), Ron set off at a smart pace round the circle of seats. Only in the last couple of metres by the stage did the sound alter, in very subtle and not at all unpleasant ways.
That’s the wonder of DSP: the various subcontractors and engineers said they could in fact give the sound the style of almost any venue in the world. Mr Gabriel’s string ensemble version of In Your Eyes was delivered with the apparent acoustics of the Vienna Opera House, for example. It might be telling that they refused the offer of my iPad and Pentatonix’s Daft Punk medley as a blind trial: I suspect the format of sound storage or input for a heavily DSP-reliant system might not work with anyone’s old music player.
The modelling piece is hugely significant. Screens showed us a rotating, translucent Royal Albert Hall, shot through with ominous red lines. The DSP logic has to know what sort of space it’s working with, and that includes what it’s made of as well as how big it is. Some of the more mainstream news journalists were mumbling about digital twins, a concept I thought was more about people than vast Victorian piles of brick. In any case, at that point I was chasing after Ron, who was off looking at some of that brick and asking hard questions about things such as echoes, and what they had done to the roof-mounted flying saucers, which were a mid-20th century attempt at improving the sound quality. He was spot on with the questions, though, because this overall audio project included the option of altering some of those early efforts to make the whole DSP thing run at lower levels of compensation. The engineer in charge of the design said it was a bit of a shock to be allowed to drill more or less at will into a listed building, just to put a speaker up.
PS not PSV
And so to Eindhoven! The invite came from the most unassuming professor I think I’ve ever met: Bart Blocken, who runs the sports
“Not only is the hall’s audio network clean, it’s also unbelievably fast”