Precision of ISO 3382-2 and ISO 3382-3 – A Round-Robin test in an open-plan office

Abstract

International standards ISO 3382-2 and ISO 3382-3 are increasingly applied to determine the room acoustic conditions in open-plan offices because attention to noise control in such workplaces has increased. The precision of those standards in open-plan offices has not been published. The purpose of this study was to determine the precision of ISO 3382-2:2008 and ISO 3382-3:2012 in an open-plan office following the instructions of ISO 5725 standard. Furthermore, the results were analyzed in the light of ISO CD 3382-3:2020 involving improvements related to uncertainty. An accuracy experiment (a.k.a Round-Robin test, inter-laboratory test, intercomparison test) was arranged where nine independent participants conducted the measurements in the same open-plan office. For ISO 3382-3:2012, the reproducibility standard deviations were 530.5 dB, 1.3 dB, 1.3 dB, and 18% for spatial decay rate of speech (D_2,S), A-weighted SPL of speech at 4 m distance (L_p,A,S,4m), A-weighted SPL of background noise (L_p,A,B), and distraction distance (r_D). The corresponding values for ISO CD 3382-3:2020 were 0.3 dB, 1.1 dB, 0.6 dB, and 16%, respectively. The reproducibility standard deviations for reverberation time were 4.5–16% within 1/1-octave bands 125–8000 Hz. The measurement uncertainty of ISO CD 3382-3:2020 is smaller than that of ISO 3382-3:2012. The values depended mainly on between-laboratory differences while within-laboratory differences had only a marginal impact. The results can be used in the development of the standards.

Introduction

International standard ISO 3382–3 [16] describes a method for determining the room acoustic properties of open-plan offices using measurements on site. The method is largely based on the scientific work of Virjonen et al. [23]. The standard provides four single-number quantities (SNQs), which are spatial decay rate of A-weighted sound pressure level (SPL) of speech, D_2,S [dB], A-weighted SPL of speech at 4 m distance, L_p,A,S,4m [dB], A-weighted SPL of background noise, L_p,A,B [dB], and distraction distance, r_D [m], i.e. the distance where Speech Transmission Index, STI, falls below 0.50.

ISO 3382-3 [16] is under revision in working group ISO/TC43/SC2/WG34. The accepted committee draft ISO CD 3382–3 [17] introduced a fifth SNQ, r_C [m], comfort distance. It is the distance where the A-weighted SPL of speech falls below 45 dB. Comfort distance integrates the information of D_2,S and L_p,A,S,4m into a quantity having unit in meters. It is expected to facilitate the communication about the level of spatial attenuation.

Measurements according to ISO 3382-3 [16] are increasingly used by consultants because an increasing proportion of new office buildings and refurbished offices may involve mandatory or voluntary target values based on the above-mentioned SNQs. Furthermore, scientific research on factors affecting environmental satisfaction in open-plan offices benefits from this standardized method because it enables an internationally harmonized way to describe the room acoustic properties of an office under study. Objective description of the acoustics of an open-plan office is important since noise and lack of privacy are among the greatest sources of environmental dissatisfaction in open-plan offices [4], and the percentage of occupants disturbed highly by office noise has been found to be higher if the distraction distance is longer [5].

Because ISO 3382-3 [16] introduced a completely new measurement procedure and new SNQs in 2012, the precision of the method could not be reported at that time. The precision is usually determined by standards ISO 5725-1 [10] and ISO 5725–2 [11]. They call for accuracy experiments where participants from different laboratories or companies (preferably 8–15) conduct the measurements according to the same method for the same object (here: an open-plan office) to obtain the between-laboratory standard deviation (s_L, variation of single measurements between different participants and different apparatus in the same space) and repeatability standard deviation (s_r, variation of repeated measurement of the same participant with the same apparatus in the same space). These can be used to calculate the reproducibility standard deviation, s_R, which is a superposition of s_L and s_r.

Yadav et al. [24] studied the repeatability of ISO 3382–3 but they did not conduct repeated measurements in the same office as ISO 5725–1 suggests. Instead, they studied several paths in several offices. ISO 3382–3 [16] states that the measurement shall be conducted at least in two measurement paths per office and the SNQs of each path are reported separately. If only one path can be found, then two measurements are conducted to opposite directions on that path. Therefore, Yadav et al. [24] focused on the uncertainty of single-direction measurements since it represents the smallest reported unit according to ISO 3382–3 [16]. They derived the apparent repeatability standard deviation of single-direction measurements using boot strapping method based on the differences of SNQs obtained for 20 pairs of single-direction measurements along 20 paths in different offices. The repeatability standard deviations of these single-direction measurements were 0.90 m, 0.6 dB, 1.0 dB and 1.0 dB for r_D, D_2,S, L_p,A,S,4m, and L_p,A,B [dB], respectively. The values seemed surprisingly large. Therefore, Yadav et al. [24] suggested that at least two measurements per office zone are recommended and if only one path is possible, measurements are conducted along the same path in both directions. However, receiving two results per open-plan office is a burden for the user especially, if one result fulfils the target value and the other does not. Therefore, ISO CD 3382–3 [17] has already taken the progress that only one SNQ value is reported per open-plan office zone, and it is the mean of the SNQs of two separate paths in the open-plan office zone or two opposite directions along the same path. Yadav et al. [24] suggested that taking the mean of two directions reduces the measurement uncertainty. Yadav et al. [24] concluded that future studies should determine the reproducibility standard deviation. It requires that several participants from different laboratories or companies conduct the same measurement in the same office. This is challenging since several participants need to travel to the same address in turns. To our knowledge, neither reproducibility standard deviation nor repeatability standard deviation has been published for ISO 3382–3 [16] based on the recommendations of ISO 5725-1 [10] and ISO 5725–2 [11].

Due to historical reasons, reverberation time, T [s], is still often measured in open-plan offices. It is usually conducted according to standard ISO 3382–2 [14] within octave bands 125–8000 Hz. Many standards and guidelines still contain target values for T, although it has been shown that reverberation time is weakly associated with spatial decay rate in open-plan offices [23]. SNQs of ISO 3382-3 [16] might better reflect the perceived attenuation of office noise in open-plan offices than reverberation time because the latter is a local quantity ignoring spatial attenuation of sound.

We could not identify a prior accuracy experiment concerning ISO 3382-2 [14] which meets the procedure of ISO 5725-1 [10] for conducting accuracy experiments. It should be noted that there was only one reverberation time measurement standard until 2008–2009, i.e. ISO 3382 [12]. It was replaced by ISO 3382-1 [15] and ISO 3382–2 [14] for performance spaces and ordinary rooms, respectively. Zehner et al. [25] reviewed prior interlaboratory studies on reverberation time measurements. Lundeby et al. [21] conducted a study with seven independent participants. However, they focused on MLS measurements and reported only octave bands 125, 1000, and 4000 Hz. Bork [1] conducted a study of three participants. James [18] conducted a study of eight participants but all participants had to use fixed positions ignoring an important source of variation in measurement results, i.e. the selection of source and microphone positions. Eggenschwiler and Machner [3] conducted a study with 20 participants. However, the participants were not required to follow ISO 3382, but the method could be freely chosen. Finally, Zehner et al. [25] conducted a study with 17 participants. However, the participants were not instructed to follow ISO 3382–1 or ISO 3382–2, but the method could be freely chosen. Many of the prior studies have been conducted before 2008 when ISO 3382–2 was published and the only study after that did not require to follow ISO 3382–2.

It is worth mentioning that sound insulation determinations involve reverberation time measurements according to ISO 3382–2. Therefore, some accuracy experiments dealing with sound insulation may involve reverberation time analyses (see e.g. [22]). Since the sound insulation measurements are usually conducted in ordinary living rooms which are small compared to open-plan offices, and the values are reported in 1/3-octave bands instead of 1/1-octave bands, we did not survey accuracy experiments on sound insulation.

Due to the lack of experimental data on uncertainty, Annex A of ISO 3382–2 [14] reports only theoretical estimations of standard uncertainty for precision, engineering, and survey grade measurements of T. These expressions do not meet the requirements of ISO 5725–1 [10] and ISO 5725–2 [11] about how the precision of accuracy experiment should be described using s_L and s_r. We are not aware of a previous experimental study, which reports the precision of ISO 3382-2 [14] in this way in open-plan offices. Because there is an obvious lack of accuracy experiments of ISO 3382–2 [14], it is very important to conduct such a study to improve the standard in the future.

The purpose of our study was to determine the precision of ISO 3382–2 standard (engineering grade of accuracy) [14], ISO CD 3382-3 draft standard [18], and ISO 3382–3 standard [16] by means of an accuracy experiment where several participants from different laboratories or companies conduct the measurements for the same open-plan office. The exact purpose was to determine the reproducibility standard deviation and repeatability standard deviation according to ISO 5725–1 [10] and ISO 5725–2 [11]. This work takes into account the progress of ISO CD 3382-3 [17] according to which SNQs are no longer reported for single paths but the reported SNQ is the mean of the SNQs obtained from two paths in the same open-plan office zone. However, we also reported the precision which conforms with the current version of ISO 3382–3 [16].