Overview and Method

An overview of simulation procedure can be seen in Figure 4.5. The simulation involves binaural and B-format recordings of a MLS signal being taken at 24 seat positions in a simulated concert hall then replayed through simulated reproduction systems for further IACC measurements. The individual stages of the simulation procedure are outlined below. Stages 1 to 4 refer to the concert hall simulation, Stage 5 to the reproduction simulation and Stages 6 and 7 to both. All signal processing and calculations were undertaken using Matlab software. Program coding can be seen in Appendices B and C.

Stage 1

Dimensions of the hall, absorption coefficients and source/receiver positions selected.

Stage 2

Calculation of source to receiver distance and reflection to receiver distances and angles.

Stage 3

Appropriate time delays and attenuations applied to the MLS signals.

Stage 4

HRTFs, corresponding to direct and reflection angles convolved with MLS signals.

Stage 6

MLS signals summed to form the left and right ear signals

Stage 5A

MLS signals encoded to B format

Stage 5B

B-format signals decoded to various reproduction methods.

Stage 7

Impulse responses extracted from ear signals and IACC measurements calculated and compared

Stage 5C

HRTFs, corresponding to loudspeaker positions convolved with signals.

Figure 4.5 Overview of the simulated procedure for comparing IACC measurements in original and reproduced environments

Stages 1, 2 and 3 – Calcualtion of Time delays and Attenuation

The dimensions of the simulated concert hall could be specified in the program. These were set to approximate the dimensions of the concert hall later used for the ‘real’ measurements (see Section 4.5). This was approximately a ‘shoe-box’ shape; however other concert hall designs could have been investigated such as a reverse fan shape which would result in stronger lateral reflections and lower IACC measurements.

Absorption coefficients could be selected for each surface of the concert hall. Absorption coefficients that are typical of materials present in concert halls were selected that ranged from 0.5 to 0.95.

The source position was set to be in the centre of the stage and at a height of 1.6m. The source was assumed to be omnidirectional. As the simulated concert hall was symmetrical about its centre line, receiver positions were only required for one half of the hall. In order to generate a wide range of IACC measurements, evenly spaced receiver positions throughout all of one half of the concert hall were selected. For each of the 24 receiver positions, the source-to-receiver and reflection-to-receiver distances and angles were calculated using geometric methods. The receiver height was also set at 1.6m. For each receiver position, attenuations due to path differences and absorption and time delays due to path differences were calculated for the six, first-order reflections. In Furlong’s simulation a centrally placed receiver position was assumed and 25 measurements within a 1m2 area around this position were taken.

Whilst this method will result in some variations in the measured IACC, a wide variation in receiver positions (and hence IACC measurements due to differences in the proximity to side walls) was selected in the current simulation. This was chosen as a comparison between the ranges of IACC measurements taken in the two environments may indicate the limits of spatial impression delivered by varying reproduction systems. Certain systems may not be able to recreate spatial conditions below a certain IACC value. Furthermore, by selecting a potentially wide range of IACC measurements, the reproduction systems’ ability to recreate a wide and varying range of spatial conditions may be determined. This may also indicate the spatial capabilities of the reproduction system.

Six, single period identical copies of a 16383-point maximum length sequence (MLS) signal were generated then delayed and attenuated accordingly. With only six reflections, a limited simulation of a concert hall was created. However, Ando shows that the measured degree of spatial impression of a synthetic soundfield can converge to a final value after only four reflections [Ando 1985].

Stage 4 – Convolution with HRIRs (Concert Hall)

In order to simulate pinna filtering and interaural time and level differences, each of the seven MLS signals (the direct sound and the six reflections), particular to a seat position, were convolved with a head- related impulse response (HRIR) that corresponded to the source-to- receiver or reflection-to-receiver angle. This method varies from Furlongs’

in that in his simulation the binaural ear signals were derived from a widely spaced pair of omnidirectional microphones.

The HRIRs used in the simulation were taken from Gardner and Martin’s set of anechoic KEMAR head measurements [Gardner and Martin 1994]. Whilst covering a large number of possible source positions, this set of HRIRs has a limited angular resolution ranging from 5° to 30° in azimuth and 10° in elevation. For each source and reflection-to-receiver angle calculated, the HRIR that was closest to the intended angle was selected for convolution.

Stages 5A and 5B – B Format Encoding and Decoding

Using the methods outlined in Stages 1, 2 and 3, a simulated B format microphone recording of the MLS signal was made. Having calculated the source-to-receiver and reflection-to-receiver angles, the velocity components (X, Y and Z) of the B format signals could be synthesized along with the omnidirectional W component.

For simulated loudspeaker reproduction, the B format signals were decoded to a number of reproduction systems. These were; mono, stereo, four and eight loudspeaker pantophonic ambisonic, eight loudspeaker periphonic ambisonic and 3/2 loudspeaker arrangement (with non-Vienna ambisonic decoding).

Stage 5C - Convolution with HRIRs (Reproduction System)

Having decoded the loudspeaker signals for each reproduction method, each individual loudspeaker signal was convolved with the closest HRIR corresponding to the angular position of the loudspeaker, relative to the listening position.

Stage 6 – Summation of Ear Signals

To simulate binaural recordings, the left and right ear signals created in Stages 4 and 5C were summed for both the simulated concert hall and simulated reproductions, respectively. From these binaural ‘recordings’ of the MLS signals, the impulse response was extracted and from this the IACC was calculated.

Stage 7 – IACC Comparisons

Impulse responses were extracted from the summed MLS signals and IACC measurements calculated for both the simulated concert hall and reproductions. Contrasts between the degree of spatial impression delivered by each reproduction method could be carried out by comparing concert hall to reproduced IACCmeasurements.