MAKING SURE IT WORKS: REDUNDANCY

Sound systems are often used in critical applications - for example in large scale sound reinforcement systems where a total system failure would result in a large audience having to be sent home, or in voice alarm systems where system failure is simply not an option. For these reasons, a degree of redundancy can be designed into audio systems.

In the ‘analogue age’ (which, for this tutorial, is arbitrarily set to the days before 2000 when digital audio and audio networking were in their initial phases, later to become the standard for live and installed sound systems), redundancy was often limited to the most likely single point of failure in an analogue mixing system. In those days, analogue mixing consoles were modular, with many single channel modules in the console’s frame. The connections from/to stage sound sources and amplifiers to/from the console’s channel modules were made using multiple analogue cables - usually bundled together in the ubiquitous ‘multicore’.

This meant that the failure of one channel in the console frame or one connection in the multicore would only affect one sound source - all the other sound sources would still work. The solution was very simple: reserve some extra channels and multicore connections to serve as back-ups and switch over only the affected channel by hand. The only real single failure point was the power supply - for this reason many consoles offered the option of connecting dual power supplies.

After 2000, digital mixing consoles conquered the world, abandoning the safe modular concept of analogue. This meant the whole mixer became a single point of failure. As it’s usually too expensive to use two mixing consoles to achieve redundancy, most manufacturers use extremely strict quality management in their design and production processes for the power supply and the mixer’s hardware, as well as the software, to reduce the failure rate to an absolute minimum.

This approach worked nicely - almost all mixers in the world are now digital and prove their value every day and virtually never failing. For larger mixing systems, redundant DSP engines and redundant power supplies are available, of course.

One thing that has changed drastically with the introduction of digital and networked connectivity in the ‘digital’ age is that the input A/D and output D/A convertors are often no longer inside the mixing console or DSP unit, but placed on the stage, or distributed over a venue. Instead of many individual physical cables - one for each sound source - making up a multicore, these systems use either digital point-to-point cables or network cables, mostly carrying an Ethernet audio protocol such as Dante. In all of these cases a single cable carries all audio channels, becoming a single point of failure. No matter how good and reliable the cable is, it can always be disconnected by accident, e.g. by someone tripping over it.

Because the IT world has exactly the same reliability challenge, Ethernet was designed to include redundancy as a standard by implementing routing algorithms that allow for two or more physical connections to be used. The most used protocols are ‘Trunking’ - basically using two cables for one connection - and ‘Spanning Tree’, which is an algorithm to allow multiple physical routes between network devices, including the support for ‘redundant ring’ topologies.

For applications using audio networks, we want the ‘switch-over’ after a failure to be as inaudible as possible - which basically means that it has to happen within one audio sample; about 21 microseconds in an 48kHz system. Because Ethernet’s Spanning Tree protocol is not fast enough, most audio network companies designed an additional redundancy method which can support inaudible switch-over. Most common methods apply a method similar to Spanning Tree, either in the form of a redundant ring (e.g. ES100, Optocore, RockNet, TwinLane), or a ‘double star’ topology (e.g. CobraNet, Dante, Ravenna). Most pre-networking ‘point-to-point’ systems use simple double cable redundancy similar to Ethernet Trunking (e.g. AES10/MADI, AES50/Supermac).

The good thing is that where Ethernet redundancy in the past was too expensive to apply in medium and small scale sound systems, it is now a standard feature of Gigabit Ethernet audio protocols such as Dante, to be used at the cost of double cables and switches - which are relatively inexpensive compared to even a small sound system.