CAPÍTULO II. METODOLOGÍA GENERAL
2.5. ANÁLISIS ESTADÍSTICO
A practical use of the acquisition of 3-D impulse responses (3-D IR) is to measure the distribution of reflections during the decay time of any room and map it to describe it in different planes. Since this thesis is focusing in small rooms, the first target was to measure a control room and analyse its temporal decay. Specifically, how the room evolves from a very directive sound field when the direct sound arrives to the receiver until the sound in the room vanishes. One of the desirable properties of a control room is to be able to hear the direct sound and avoid flutter echoes and strong early reflections.
By applying the analysis of the impulse response ( IR) using the STFT method is possible
to analyse the diffuseness and assess the temporal and spectral properties of the sound decay. This experiment was done before having access to the p-p probe, therefore only
the Soundfield was used. A description of the measurement conditions follows:
A sample of 10 measurements of a 3-D IR in small rooms was done using the Soundfield
model SPS 422B microphone and applying the short Fourier transform method ( STFT).
Figure 5.13 shows the measurement scenario.
Only the horizontal plane (xy-axis) is investigated at this time. The measurements were
made in Blue Room 2 control Room from the facilities of recording studios at the
University of Huddersfield, whose dimensions are 4.16 × 2.75 × 2.87 m and its volume
is 32.1 m3. Measurements were performed at two different source positions. The first
was at the front of the microphone and the second at 120º from the origin, referenced at
4.12, Figure 7.1 and Figure 7.2). The sampling frequency resolution was set to 96 kHz, which gives a temporal resolution of 0.01 ms.
This room has a fairly short reverberation time RT = 0.17 s @ 1 kHz, and a Mean Free
Path lMFP= 2.2 m giving a minimum time window of Δt= 6 ms. However, smaller time
windows of 0.1 and 1 ms were tested in this analysis and seem more appropriate to
determine any isolated early reflection. Application of the Soundfield microphone model
SP 422 to small rooms. The results of this experiment are covered in Chapter 7 and the discussion of these results is found in Chapter 8,( section 8.2.1 ).
Figure 5.13: Photo and sketch of Blue room 2 measured with the Soundfield microphone model SP 422.
5.6 Summary
This section describes details of the Acoustic probe configurations, covering the measurement conditions and the repetitions needed to validate the accuracy of the method. The next section covered is the description of the measurement environment, which is comprised of: semi-anechoic chamber, monitor speaker influence, damping vibration on source and receiver, and the laser cross system to define spatial coordinates. It describes the laboratory test for single reflection measurements in the semi-anechoic
chamber using the p-p probe and the Soundfield microphone. Afterwards, the single
performance in angular accuracy of sound detection. An explanation of the effect of varying the spacing between the microphones is depicted in that section. The usable frequency range is discussed according to the optimum selection of spacing at 10.3 mm.
The next topic covered is the applications of 3-D IR measurements in a control room
where the application of a cheaper and more versatile Soundfield microphone for the
Chapter 6 : Results
of laboratory tests
This section presents the results of the experiments undertaken in the single reflection scenario ( for an explanation of it please refer to section 5.2.1 ). The case of measuring a
single reflection in the semi-anechoic chamber is tested with the time domain ( TD)
method and the frequency domain method ( STFT).
The detection of the angle of arrival of direct sound (θs
i) is a trivial issue, but the
detection of the angle of arrival of first reflection (θm
i) is not, and usually exhibits larger
errors in its estimation. Therefore, it is expected to have a larger errors and variations of the positions, which have the longest distance paths because the time of flight of the direct sound and the first reflection is closer than when the path difference is longer (Defrance et al., 2009). Owing to the fact of the high sensitivity of the system to the angle of arrival of first reflection and its tendency to deliver errors on the estimation, two
measurement methods were used. This is because the time domain method ( TD) and the
frequency domain method ( STFT) complement each other’s deficiencies.
Due to the great amount of data created in this experiment, only the mean absolute percentage error graphs and their uncertainty are presented. Initially the most intuitive graphs were the error graphs. They showed if the measure was over estimating the real value or if it was under estimating. Later the error graphs showed an analysis of the errors with positive or negative sign. This complicated the analysis of errors. It was needed to use an absolute norm of error. Mean squared error was discharged because it introduced a non-linear explanation of errors while mean absolute percentage error seem to preserve the error without introducing any non-linear behaviour.
Figure 6.1: Example of impulse response B-Format signals measured. It contains the direct sound and first reflection (signals not normalised).