12th January 1999 - Audio Signal Processing

There was standing room only for January 12th's talk by David Karlin of BSS. David has been with BSS, traditionally a company making products for the touring and live sound markets, for some years, having come from a background in telecommunications, computing and software engineering. The lecture was split into three major sections which dealt with the history the technology, and some of the commercial aspects of designing and manufacturing audio signal processors.

The talk began with a brief history of how signal processing has been applied to audio signals. It was noted that the first applications were corrective filters applied at the two ends of the replay chain, for example passive loudspeaker crossovers and RIAA equalisation. Gradually filters and dynamics processors made their way towards the middle of the chain in the form of parametric EQ, graphic equalisers and noise gates etc., initially as 'problem- solving' devices. Over some time people began to realise that these processors also have creative potential and started using them in new ways, which further encouraged development of processors designed specifically for creative use.

The advent of digital processing technology has made possible sound processors that would be practically impossible in analogue technology. He suggested that the ability to delay a signal was the single most important new processing element. For the future he believes that applications such as compensation for loudspeaker distortions will be important, but that such applications will require a level of co-operation amongst the audio industry that has not traditionally been present.

The history of BSS, founded in 1978, follows almost precisely this course. Its first products were a modular crossover (serial numbers 201-204 sold to one F. Zappa!) and a DI box - at the two extremes of the signal chain. It followed these with various EQ and dynamics processors, including the 'dynamic EQ' which attempts to make up for the lack of microphone technique amongst stage vocalists. More recently BSS has had a range of digital and hybrid analogue/digital products.

David pointed out that one of the most critical areas for design in order to get a good-sounding product is the user interface. There is no point having the most clever, innovative signal processor in the world if it cannot be easily controlled.

The second section of the evening was a discussion of some of the commercial decisions and compromises that are made when designing a piece of equipment and bringing it to market. For a high-volume product good design is effectively free (although not always used wisely). At low volumes the design cost is highly significant, and it is this tradeoff between a specialist box (with a small market) and a 'Swiss army knife' (that does a lot reasonably well, but excels at nothing) which is critical at determining the success of a product.

The target is for a new device to be general-purpose enough to have a large market, while being very nearly as good as a specialist device at each of its jobs. In the course of this design process customer feedback is essential, and as a result the new product tends to do those jobs required by the existing customer base very well. This can hamper diversification, and tends to result in particular companies being connected with particular applications; for example a live-sound EQ and a recording studio EQ are essentially the same device, but the chances are they will be from different manufacturers as a result of this pigeon-holing.

David also touched on the issue of perceived-versus-real quality, and pointed out that it is rarely possible for a (potential) customer to assess the quality of a product until it is too late. Specification sheets are effectively useless, and long-term reliability and after-sales service are impossible to judge until the equipment is bought, and probably several years old.

The aim from this point of view is to design, engineer and implement a product well, and then pitch it at a price and margin such that the volume is adequate, and the product can be supported.

The third section of the talk covered the various technologies available for implementing audio signal processors.

Analogue electronics are still the most effective way of performing simple functions such as DI and very small mixers. Analogue electronics are not getting significantly cheaper as time goes on. Digitally-controlled analogue was dismissed as a dead technology because of the engineering difficulties and compromises associated with getting high- quality analogue circuitry in the same box, or on the same silicon die, as high speed digital circuitry. In addition, as digital technology becomes cheaper the hybrid option becomes economically as well as technically difficult to justify.

Digital technology by contrast is becoming very good very quickly. Traditionally performance has been limited by the ADC and DAC performance, but these too are becoming very good, and very cost-effective. The one remaining problem is that of signal delay, especially when a number of devices are daisy-chained, each with its own ADC and DAC as these typically introduce a few milliseconds of delay each.

In addition digital technology makes possible algorithms and processes that would be impossible in analogue. The two key features here were cited as the ability to arbitrarily delay a signal with no quality loss, and the ability to add resettable controls to any device.

The good design of a digital processor necessarily centres around the control interface. It is easy to fit a lot of signal processing into a very small space using DSP, but the issue of how to control it is difficult. For example, analogue faders are very good at their Job, but the alternatives for a resettable digital control (motorised faders, shaft encoders, trackballs etc) are either very expensive or much less easy to use.

At the same time, however, digital technology allows great advances. A hotel, conference centre or airport can be equipped with a very complex sound system incorporating sound reinforcement, public address, emergency alarms, multi-zoning etc, whose controls can be reduced to a small number of presets (labelled, perhaps 'music' or 'wedding'), selected by a simple row of buttons.

Digital allows the control surface to be divorced from the audio processors, which can therefore be positioned anywhere that is convenient. Furthermore the processing can be distributed around a network giving a degree of inherent fault-tolerance, and potential for a drastic reduction in cabling. Many conventionally separate systems can then be integrated into one such distributed audio system. For example the same components can be used for a public address system, background music as an entertainment function, and a voice alarm emergency system. Previously these three would have been separate in, for example, a leisure complex, but a configurable distributed system can perform all three functions.

David's vision of the future included a system in which CAD data for a shed metal punch was produced as an integral part of the sound system design. This data could then be used to generate a custom control surface for each and every sound installation. This feature, which was at first presented as almost frivolous, was seen to be nearer than first thought. We were then given an interesting demonstration of 'Soundweb', BSS'S solution to this problem of distributed, networked audio processing. This system comprises any number of processing units, each of which holds a configurable DSP, eight analogue inputs, eight analogue outputs. and two network ports. The network allows daisy-chaining of units with 8 channels of audio between them, and more complex systems can incorporate 'hubs' which allow zoning and branching of the network, While retaining the ability to route audio and control information at will. After a lively question and answer session the evening was concluded with a vote of thanks by the Section Chairman.

Christopher Hicks