First, you replaced your 8-track system with a cassette player. Then, you threw out your turntable and bought a CD player. What next? Digital speakers? Tony Hooley thinks so. He believes that the digital loudspeaker will completely change audio electronics and that analogue systems will simply disappear. Power amplifiers too, will be outmoded, as low power loudspeakers will be driven directly by digital signals. His company, 1…, is already working on a purely digital design.
‘Loudspeaker design has not fundamentally changed in 75 years,’ says Hooley. ‘We simply have better magnet, suspension and cone materials.’
Conventional loudspeakers suffer from some fundamental problems too. In their design, a suspension system holds a coil within a gap in a permanent magnet. Due to the non-linear effects of the magnetic field along and at the edge of this gap, the resulting force on the coil in the field is non-linear as its position changes. In addition, the suspension system produces increasing force as the cone is moved off centre, due to the fact that it too acts like a non-linear spring.
‘A speaker is an inherently doubly non-linear device the coil produces non-linear effects in the same sense that the reactive force does. If you put in a sine wave signal, the result that you hear is a sine wave with a flat top a principle cause of harmonic distortion,’ says Hooley.
Another problem with conventional speakers is that to produce a lot of sound energy at low frequencies, a large volume of air needs to be displaced. That can either be achieved by a small cone and a big displacement of the coil, or a large cone with a smaller displacement. However, in a small speaker, the coil movement will be extremely non-linear even if the cone just moves for a fraction of a millimetre. That is why speakers built to handle low frequencies are usually large.
For most hi-fi speaker systems, three speakers are used for low, midrange and high frequency signals with a crossover circuit used to deliver the right signals to the right speakers. ‘The problem,’ Hooley says, ‘is that cross over units also put a phase on the signal so you start to get phase shifts in the output of the signals where crossover occurs. Worse still, the apparent size of the sound source changes as you go through the crossover region, giving strange directional effects.’
And they are inefficient too. A coil in a loudspeaker may have a DC resistance of 6. ‘When it is driven by a power amplifier, just about all the energy is dissipated as heat in the coil. Only about 1% of the energy comes out as sound energy,’ says Hooley. Because they are inefficient, expensive power amplifiers are required to drive the speakers. Overall, you might be talking about 250W in for 1W of sound out,’ he adds.
BACK TO BASICS
So the team at 1… went back to basics and asked themselves some fundamental questions. How could they make a loudspeaker that was completely digital? The answer was to design an array of devices of different areas driven with pulses. The result would be a pressure wave proportional to the area of the speaker.
At first, they considered using a binary signal to drive the signal, an approach that has been under investigation at Sony in Japan. ‘It turns out that doesn’t work because using a binary signal, the transducers cannot be made accurately enough,’ says Hooley. ‘Even if you could make them, such a system still wouldn’t work because the code pattern produced would have transients in it that could be perceived by the human ear,’ he adds. ‘Incrementally moving from an output of 7 to 8 in binary, for example, produces a large transient effect as the binary bits switch from 0111 to 1000.’
So the 1… team decided upon a unary system in place of a binary one. ‘With a unary system, only the number of ones determine the output value. All the transducers can be identical,’ explains Hooley.
On that basis, two years ago, the company built a demo machine to prove the feasibility of the concept. It used 80 smoke alarm transducers driven by a unary signal converted from a digital CD input. Hooley says that although the transducers had a terrible frequency response and resonated very badly, the result was that when connected to a CD, music could clearly be heard. The demo had proved that the old concept of driving one transducer with varying power could be replaced with a new paradigm where a variable number of transducers could be driven with the same power output.
Currently Hooley and his team are working on a second demonstration system that will use digital signal processors (DSPs). Specifically, an array of 256 transducers or speakers will be driven by a bank of Analog Devices Sharc DSPs.
In the system, a 16-bit signal from a CD player is taken and oversampled by a master Sharc based DSP board. This produces a 176kHz data rate that is then quantised to 8 bits. A feedback loop and filter determine the spectrum of the quantisation noise, and this noise is moved from the 0-20kHz band into the 21-78kHz region, which is outside the audible range. ‘The bit of the spectrum that we do care about has virtually the same signal to noise ratio and dynamic range as the original signal,’ says Hooley.
The master Sharc board drives a further 8 Sharc slave boards that take the 8-bit oversampled signal from the first board and drive an array of 32 transducers each. Not only do these slave boards perform the binary to unary conversion, they also allow a digital delay to be introduced unique to each transducer so that the phase of the waveform can be shaped to allow the resulting sound to be ‘steered’.
In parallel with the digital electronics work, 1… is also developing the digital transducers. These, however, will not be ready for about a year. In the meantime, the new demonstration vehicle will use conventional loudspeakers as transducers driven by a set of custom electronics that correct for the worst aspects of the speaker.
Once they are built, the brand new 1… digital transducers will comprise just three parts: an outer helical bender, an inner cylindrical piston manufactured using silica aerogel, and a pressurised gas filled bearing between the two which allows the piston to move freely along the axis of the helix with essentially zero friction. In operation, the cone-like deformations of the helical bender squeeze the gas bearing radially, which in turn imparts axial forces to the piston which then rolls on the bearing up and down the centre of the helix up to 10mm in either direction.
Each of the elements presented a considerable design challenge. Although flat planar PZT benders have been manufactured for many years, all existing manufacturing methods were found to be incapable of manufacturing helical or other curved multilayer structures, and were generally limited to planar devices.
The solution was found at the University of Birmingham’s Interdisciplinary Research Centre in Materials for High Performance Applications. ‘We overcame the traditional problems by suspending small PZT particles in a plasticine like substance. The PZT material can then be formed into a multi-layer rod and mechanically shaped as required,’ says Dr David Pearce, a Research Fellow in High Performance Metals.
For the piston itself, the company chose a silica aerogel material with about 503 air density, a figure which Hooley says can be lowered, to perhaps, 103. The benefits of the material is that it is relatively stiff and has a very low mass around 20mg. The challenge was to find a manufacturer who could make the piston in such a small size, a challenge that led Hooley all the way to the West Coast of the US.
The last challenge was to make a very low friction bearing that moves in between the helix and the piston. Hooley designed and patented just such a bearing in 1997 but could find no one capable of manufacturing it. Eventually, he found a manufacturer of thin walled polymer 20 m to 30 m thick, and, using the polymer, designed what is in effect a thin polyurethane toroidal belt with a layer of gas inside that separates the coil from the piston. The design provides a very low cost, low friction bearing because it is simply made from gas and polymer with only rolling friction.