Pilot tests were conducted with a limited number of participants to establish the suitability of the audibility threshold and equal annoyance procedures. Additionally the equal annoyance procedure was adapted to ask specifically for Equal Loudness (EL) to give an indication whether the equal annoyance results would be different from EL results.
The number of pilot test participants for each test is summarised in Table 14.1
Audibility
thresholdannoyance Equal loudnessEqual
No masking noise 5 4 3
Indoor 7 6 3
Outdoor 4 6 3
Table 14.1 Number of pilot test participants for Part A Tone audibility threshold, equal annoyance and equal loudness in the three different scenarios as specified in Section 5.5.
14.1Audibility threshold
The pilot tests used a tracking Békésy method to assess 94 detection thresholds: First the participant heard the garden noise played and a user interface on screen (Figure 15.1, Appendix VI) explained the procedure. When the participant pressed start an audible tone would be played in addition to the masking and a user control GUI (Figure 15.2) appeared. This required the participant to press the “Audible” button until the Tone level is reduced to be inaudible. The participant then pressed the “Inaudible” button which caused the Tone level to increase until audible. The process was terminated when steps converged as can be seen in the trial operator control window in Fig. V.3.
Audibility threshold for Tones without masking
Figure 14.1Audibility threshold of tones without masking noise
Fig. 6.1 shows the results of the nominal audibility thresholds for five pilot test participants and the average values in the absence of masking noise. It can be seen that the audibility threshold
decreases towards higher frequencies in agreement with standard literature (Fastl & Zwicker, 2007). The exception occurs at 70 Hz where the audibility threshold seemed to be consistently higher than the 53 Hz threshold. This is a test artefact and was traced back to the effect of a room mode. It is also evident that the variation between participants expressed through the standard error bars are larger from frequencies at and below 100 Hz compared to the 200 Hz and 400 Hz test tones. This could be due to individual differences of perception or to spatial variability of the low frequency tones with head movement. While spatial variability was found to be a problem at 70 Hz the tone levels did not change perceptibly at other frequencies.
Audibility threshold in indoor and outdoor scenarios
Figure 14.2 Audibility thresholds of tones in wind turbine noise masking, no garden noise. a) indoor, b) outdoor
When comparing audibility thresholds in the indoor and outdoor scenarios in relation to the no masking scenario, Fig. 6.2 shows the same general shape of the audibility curves with consistently higher audibility threshold levels in the outdoor scenario. This is entirely expected as the dB(A) level values in the figure refer to the outdoor levels and the façade attenuation therefore causes the indoor levels to be considerably quieter.
The figure also shows that the three different chosen turbine levels show very similar tone audibility with many of the data points lying within the error bars of the other data points and some audibility threshold levels of the 39 dB(A) wind turbine noise higher than the levels of the 44 dB(A) wind turbine noise. At the same time participants reported to be able to distinguish clearly between the different scenarios. This suggests that at least part of the masking spectrum was audible if not necessarily the low frequency masking bands.
The high standard deviation at 43 Hz (Figure 6.2 a) was due to problem with tone identification when a participant erroneously identified part of the masking sound to be the tone which caused an unrealistically low audibility threshold. Two measures were taken to avoid this type of problem for the final test. Firstly, the room calibration procedure was enhanced (see Appendix IV for details of the final algorithm). Secondly the audibility threshold methodology was changed to Two- Alternative Forced Choice (2AFC) to enable detection of erroneous tone identification.
When comparing these results with those that contain additional masking garden noise audibility threshold does not change significantly.
14.2Equal annoyance of tones
Figure 14.3 Equal annoyance, tones only for two different reference tone levels 5 dB apart and two participants.
When two participants were asked to compare the annoyance of a reference tone to the annoyance of the test tones defined in Section 5.5 P9 adjusted the tone levels of equal annoyance to be higher towards the lower end of the spectrum and lower towards the higher end of the spectrum which is the expected behaviour given the audibility threshold results presented above. P3 chose a higher equal annoyance level for the 400 Hz tone which could suggest a certain habituation effect with respect to the reference tone. The 70 Hz irregularity is the same as in the audibility threshold results. Fig. 6.3 shows the difference between the SPL of the test tone and the SPL of the reference tone. If annoyance scaled linearly with loudness the green and blue lines should therefore fall together. Instead the figure shows parallel lines with the lower reference tone levels at a higher than expected tone level. The curves show a more arbitrary gradient at the lowest 3 frequencies
Equal annoyance with and without garden noise
Tone levels of equal annoyance at garden noise and no garden noise annoyance (Fig. 6.4) for the indoor and outdoor scenarios are mostly parallel and different especially for the indoors scenario. Statistical significance is to be explored in main listening tests.
Conclusions of pilot tests
The pilot tests were conducted to explore the appropriateness of the methodology with respect to participant response and expected outcomes. The effect of using a variety of LAeq_WT, reference tone
levels and types of masking noise was explored.
In general the participants found the tests to be intuitive and workable in spite of prior concerns about the length of the procedure. Results on audibility threshold and equal annoyance look promising and reasonable. The audibility threshold threshold method was changed from Békésy tracking to Two-Alternative Forced Choice (2AFC) to generally avoid false identification of tones in noise. Stimuli loop length was also increased to avoid periodicity being perceptible; this solved the identification problems with the 35 Hz tone.
In response to technical problems during the pilot tests and first round of listening tests, the loudspeaker calibration procedure was enhanced to use multiple microphones to avoid measurement nulls, as compensating for these can create tonal artefacts, and a technical issue with the subwoofers (which particularly affected the 70Hz tone) was corrected.