12th October 1999 - Measurement and simulation results comparing the binaural acoustics of various direct radiators

The UK Section's October lecture was given by Neil Harris of NXT and the University of Essex. The lecture was based on PhD research completed by Neil over the period of approximately a year, the centre of the research being the analysis of the flat panel loudspeakers designed by NXT.

Neil started by introducing how conventional loudspeakers are usually measured on-axis in an anechoic environment. He argued that this measurement is not particularly relevant to how the loudspeakers are used, and tells little about the true acoustical properties of a loudspeaker.

He went on to introduce the flat panel loudspeaker and explain its differences from a conventional loudspeaker driver. The flat panel is an unusual loudspeaker, the sound being created by a uniformly distributed free vibration in a stiff light panel. Neil equated how the loudspeaker works by explaining that striking a table makes a resonance, the trick is in making the panel sound less like a table! There were a number of flat panel loudspeakers present, including a large demonstration model which was passed round the meeting, and a commercial system including a subwoofer unit.

Neil used computer simulations to show the radiation patterns of a conventional loudspeaker driver, explaining the reasons for the uneven side lobes and beaming in the polar response. He contrasted this with the flat panel loudspeaker, which due to its random vibrating across the surface, manages to keep a circular wavefront both at low and high frequencies.

Other qualities of the flat panel loudspeaker were illustrated, including the frequency response and the impulse response. Unusually, the flat panel's impulse response rings on much longer than a conventional loudspeaker, up to 10ms or more. This is due to the low mass of the panel. On the other hand, this low mass means that the leading edge of the transient response is much steeper than that of a conventional loudspeaker. If the long tail of the impulse response was present on a conventional loudspeaker, this would seriously affect the perceived sound quality and would sound very coloured. However, because the decay of the flat panel is diffuse in time and space, with careful engineering it is not as much of a problem. Neil explained that the leading edge of the impulse response seems to be much more important.

The work which Neil has undertaken included modelling the flat panel, a conventional piston-like driver, and an ideal point source. These models were used to simulate the 3 drivers in an anechoic environment. Simulating measurements of the on-axis pressure revealed very few differences between the very different drivers.

In order to separate the drivers from each other, Neil looked at measurements off-axis. The measurements taken in the simulations included the directivity polar pattern, and the correlation of a sound emitted at a certain angle against the sound emitted on-axis. These measurements showed a large difference between the three drivers. The piston and the point source had very similar correlation curves - approximately completely correlated in the front hemisphere. The directivity pattern of the piston was obviously different at high frequencies due to the cancellation which was caused by the physical size of the driver.

In contrast, the flat panel loudspeaker had a fairly even hemispherical directivity pattern at all frequencies. More distinct, however, was that the sound emitted towards the sides of the panel was very uncorrelated with the sound emitted on-axis.

This was compared with the subjective results of a previous localisation experiment comparing conventional loudspeakers and flat panels. Neil asserted that the reason why the flat panel performed better in the localisation experiment was due to the fact that the off-axis radiation of a flat panel loudspeaker was more diffuse. This is apparently less distracting for the listener.

To sum up, Neil concluded that the on-axis anechoic response is of limited use as an acoustic measurement as it was unable to separate the three drivers tested here. In contrast, the off-axis response differentiates well between the drivers.

Neil finished the lecture with a fascinating computer animation that demonstrated the pressure waves created by two flat panel loudspeakers and their interaction with each other and a human head. After a number of stimulating questions, the meeting closed with thanks to Neil for an interesting lecture.

Russell Mason