Managing Measurement Uncertainty in Building Acoustics

Buildings 2015, 5, 1389-1413; doi: 10.3390/buildings5041389 Article OPEN ACCESS buildings ISSN 2075-5309 www.mdpi.com/journal/buildings/ Managing Measurement Uncertainty in Building Acoustics Ciara Scrosati * and Fabio Scamoni Construction Tecnologies Institute, National Researc Council of Italy, via Lombardia 49, 20098 San Giuliano Milanese (MI), Italy * Autor to wom correspondence sould be addressed; E-Mail: c.scrosati@itc.cnr.it; Tel.: +39-02-9806-432; Fax: +39-02-98280088. Academic Editor: Umberto Berardi Received: 30 October 2015 / Accepted: 14 December 2015 / Publised: 18 December 2015 Abstract: In general, uncertainties sould preferably be determined following te principles laid down in ISO/IEC Guide 98-3, te Guide to te expression of uncertainty in measurement (GUM:1995). According to current knowledge, it seems impossible to formulate tese models for te different quantities in building acoustics. Terefore, te concepts of repeatability and reproducibility are necessary to determine te uncertainty of building acoustics measurements. Tis study sows te uncertainty of field measurements of a ligtweigt wall, a eavyweigt floor, a façade wit a single glazing window and a façade wit double glazing window tat were analyzed by a Round Robin Test (RRT), conducted in a full-scale experimental building at ITC-CNR (Construction Tecnologies Institute of te National Researc Council of Italy). Te single number quantities and teir uncertainties were evaluated in bot narrow and enlarged range and it was sown tat including or excluding te low frequencies leads to very significant differences, except in te case of te sound insulation of façades wit single glazing window. Te results obtained in tese RRTs were compared wit oter results from literature, wic confirm te increase of te uncertainty of single number quantities due to te low frequencies extension. Having stated te measurement uncertainty for a single measurement, in building acoustics, it is also very important to deal wit sampling for te purposes of classification of buildings or building units. Terefore, tis study also sows an application of te sampling included in te Italian Standard on te acoustic classification of building units on a serial type building consisting of 47 building units. It was found tat te greatest variability is observed in te façade and it depends on bot te great variability of window s typologies and on workmansip. Finally, it is suggested ow to manage te uncertainty in building acoustics, bot for one single

Buildings 2015, 5 1390 measurement and a campaign of measurements to determine te acoustic classification of buildings or building units. Keywords: measurement uncertainty; building acoustics; Round Robin Test (RRT); sampling; acoustic classification 1. Introduction Tis paper is a revised and expanded version of te paper Uncertainty in Building Acoustics [1] presented at te 22nd International Congress on Sound and Vibration ICSV22. Wen reporting te result of te measurement of a pysical quantity, it is compulsory tat some quantitative indications of te quality of te result be given so tat tose wo use it can assess its reliability. Witout suc indications, measurement results cannot be compared, eiter wit one anoter or wit reference values given in a specification or standard. It is terefore necessary, in order to caracterize te quality of te result of a measurement, to evaluate and to express its uncertainty. Generally, it is widely recognized tat, wen all of te known or suspected components of error ave been evaluated and te appropriate corrections ave been applied, an uncertainty about te correctness of te stated result still remains; tat is, a doubt about ow well te result of te measurement represents te value of te quantity being measured. Te word uncertainty means doubt, and tus in its broadest sense uncertainty of measurement means doubt about te validity of te result of a measurement. Te formal definition of te term uncertainty of measurement developed in te Guide to te expression of uncertainty in measurement (GUM) [2] is as follows. Uncertainty (of measurement): parameter, associated wit te result of a measurement, tat caracterizes te dispersion of te values tat could reasonably be attributed to te measurand. Tis definition of uncertainty of measurement is an operational definition tat focuses on te measurement result and its evaluated uncertainty. However, it is not inconsistent wit oter concepts of uncertainty of measurement, suc as a measure of te possible error in te estimated value of te measurand as provided by te result of a measurement; or an estimate caracterizing te range of values witin wic te true value of a measurand lies. Altoug tese two traditional concepts are ideally valid, tey focus on unknowable quantities: te error of te result of a measurement and te true value of te measurand (in contrast to its estimated value), respectively. 2. Te Uncertainty in Terms of Repeatability, Reproducibility and in Situ Standard Deviation Tests performed on samples made of materials presumed to be te same, in identical conditions, generally do not give te same results. Tis condition is due to inevitable errors (systematic and random) in test procedures, caused by te difficulties in controlling te several factors tat influence te test. To determine te accuracy of a measurement metod, bot accuracy and precision sould be considered; in particular, te latter indicates te correlation between te test results.

Buildings 2015, 5 1391 Precision is a general term for te variability between repeated tests. Two measures of precision, termed repeatability and reproducibility, ave proved necessary and, for many practical cases, sufficient for describing te variability of a test metod. Repeatability refers to tests performed on te same test object wit te same metod under conditions tat are as constant as possible, wit te tests performed during a sort interval of time, in one laboratory by one operator using te same equipment. On te oter and, reproducibility refers to tests performed on identical test items wit te same metod, in widely varying conditions, in different laboratories wit different operators and different equipment. Tus, repeatability and reproducibility are two extremes, te first measuring te minimum and te latter te maximum variability in results. Te building acoustic quantities include airborne sound insulation of internal partitions, airborne sound insulation of façades, impact sound insulation of floors and sound pressure level from service equipment in buildings. Te quantities tat ave to be measured and teir measurement metods, for all aspect involved, are described in te international standard series EN ISO 10140 [3] for laboratory measurements and in te international standard series ISO 16283 [4] for field measurements. Te accuracy of tese measurement metod depends on several factors tat influence te test, suc as acoustic instrumentation, acoustic metod (micropones and sources position), context (regular rooms or semi-open space, of any size), constructive details of te building (tat could ave effect on acoustic measures) and workmansip, and, concerning sound levels, influence of instrumentation working conditions (repeat configuration). Detailed information for eac of tese factors is ardly available. Bot random and systematic errors affect te acoustic measurements results. Te random effects can be determined by repeated independent measurements in essentially identical conditions. Te systematic effects, owever, are not easy to determine, but, as a general rule, tey can be determined tanks to comparative measurements to be executed in different test facilities (for laboratory measurements) or carried out by different laboratories (for field measurements), and te knowledge of te random errors in tose conditions. Terefore, it is necessary to refer to te concepts of repeatability and reproducibility, wic provide a simple means for te expression of te precision of a test metod and of te measurements performed according to te test metod. Te best metodology to study te repeatability and reproducibility of building acoustic measurements is to carry out an Inter-Laboratory Test (ILT), or a Round Robin Test (RRT), tests consisting of independent measurements executed several times by different operators. Due to te particular nature of te sample in building acoustics, in addition to repeatability and reproducibility standard deviations, anoter standard deviation is defined, te in situ standard deviation (defined, for te first time, in ISO 12999-1 [5]), wic could be useful to estimate. Te in situ standard deviation is a particular kind of reproducibility standard deviation tat is measured in te same location on te same object. In fact, in te case of RRT field measurements, wen different operators, wit teir own equipment, perform measurements on a particular building element, bot te location and te object under test are te same. Terefore, location is te only difference between reproducibility and in situ standard deviation: for te in situ standard deviation, te location is exactly te same as is te test object, wile in te case of reproducibility standard deviation te locations are different and te test object can be eiter te same test object or identical test objects tested in te different locations. Te in situ standard deviation, terefore, corresponds to a reproducibility standard deviation of te same object in te same location.

Buildings 2015, 5 1392 2.1. Round Robin Test Generally, cooperative tests (ILT or RRT) assess te uncertainty of measurement metods using a reference value. One of te main aspects of tese tests is te determination of tis reference and its uncertainty. A reliable, low-uncertainty reference value is required in order to minimize te uncertainty of a cooperative test. Due to te typology of te sample test in acoustic measurements, a reference value does not exist; terefore an estimated value is used. Te best measuring reference is te mean value. A RRT of sound insulation field measurements of building elements was carried out as part of a researc sponsored by te Lombardy Region [6 8]; tis study was based on te cooperation of tree different bodies: a researc body, ITC-CNR (Construction Tecnologies Institute of te National Researc Council of Italy); a university laboratory, DISAT (Department of Eart and Environmental Sciences of te University of Milano-Bicocca); and a control organization, ARPA-Lombardy (Regional Agency for environmental protection) and it was coordinated by ITC-CNR. In te first approac to te problem [6], te analysis was centered on te single number values of te Italian regulation [9] and on te narrow frequency range (from 100 to 3150 Hz). In later studies [7,8], te analysis considered all te possible descriptors of te different European national legislations and was extended to te enlarged frequencies range (from 50 to 5000 Hz). Anoter study on te uncertainty of façade sound insulation [10] was carried out at te initiative of te Building Acoustics Group (GAE) of te Italian Acoustic Association (AIA). Tis study was focused on te low frequencies (from 50 to 80 Hz), in particular on te comparison between te procedure stated in ISO 140-5 [11] and te new low frequency procedure stated in ISO 16283 [4]. Te main results of tese studies are summarized in te following section. 2.1.1. Airborne Sound Insulation Notwitstanding te importance of te uncertainty of te measurement metod in building acoustics, te uncertainty of field measurements was not compreensively investigated. Tere are only few examples in te literature [12,13] compared to tose of laboratory tests [14 18]. Te studies regarding laboratory tests conclude tat te main influences are caused by te laboratory geometry and materials, te flanking transmissions, te type of border material, and te different test opening dimensions [15,16]. Nine teams coordinated by ITC-CNR were involved in te study about te uncertainty of airborne sound insulation [7]; eac of tem as replicated te tests five times, including te reverberation time. No deviations occurred from te test procedure laid down in ISO 140-4 [19] but, repeating te measurements several times, te parameters left open in te measurement procedure were represented as best as possible. In particular, te set of micropone positions and source positions were selected anew, more or less randomly, for eac repeated measurement. Te measurands were a floor witout floating floor (surface mass of 550 kg/m 2 and surface of about 19 m 2 ) and a ligtweigt wooden partition wall (surface mass of 30 kg/m 2 and surface of about 8.5 m 2 ). Considering te goal of European armonization of acoustic parameters [20], te differences between te various descriptors (R, Dn and DnT) were analyzed in terms of average, maximum and minimum values, and in terms of standard deviation of repeatability and reproducibility (in situ standard deviation, referring to ISO 12999-1 [5], were te reproducibility standard deviation of te same element is measured in te same location).

Buildings 2015, 5 1393 Figure 1 sows te standard deviations of repeatability sr and in situ reproducibility standard deviation ssitu of all analyzed quantities. Te descriptors extension at low frequencies (from 50 to 80 Hz) (LF) was also analyzed. From te graps of Figure 1, it is evident tat te uncertainty at LF is muc greater tan te uncertainty in te narrow frequencies range from 100 to 5000 Hz. From te comparison of te RRT ssitu values wit te values of te ISO 12999-1 [5] for situations A (sr) and B (ssitu) (see Figure 1), it was found tat te values of situation B underestimate te uncertainty of in situ measurements in particular at low-medium frequencies. Moreover, te values of ssitu [7] obtained are iger also tan te sr values, in particular for te floor at low-medium frequencies from 80 to 200 Hz, and for te wall from 160 to 250 Hz. s situ [db] 8 7 6 5 4 3 F-DnT F-Dn F-R F-D W-DnT W-Dn W-R W-D RT ISO12999-1sR ISO12999-1ssitu s r [db] 8 7 6 5 4 3 F-DnT F-Dn F-R F-D W-DnT W-Dn W-R W-D RT ISO12999-1sr 2 2 1 1 0 50 100 200 400 800 1600 3150 f [Hz] (a) 0 50 100 200 400 800 1600 3150 f [Hz] (b) Figure 1. ssitu (a) and sr (b) of floor (F) and wall (W) of R, Dn, DnT, D and RT [7], wit te comparison wit te reproducibility, in situ (a) and repeatability (b) standard deviation of ISO 12999-1 [5]. Te results of SNQ calculations are sown in Table 1. Two different ways to determine te SNQs ave been considered for te above-mentioned study [7]. Te former is to determine SNQ according to ISO 717-1 [21] by sifting te reference curve (value in te range from 100 to 3150) in steps of 1 db toward te measured curve, until te mean unfavorable deviation is as large as possible but not more tan 32 db; all te laboratories involved in te RRT ave followed tis procedure. Te latter is to determine SNQ plus te spectrum adaptation terms C and Ctr according to ISO 717-1 [21] bot in te narrow frequency range from 100 to 3150 Hz, and in te enlarged frequency range from 50 to 5000 Hz; in bot cases rounded to integer and wit 1 decimal place (subscript 01), using Equation (1) [21]. Te SNQs plus te spectrum adaptation terms were determined using a 0.1 db resolution, following from te work of Wittstock [22], to obtain more accurate data for te analysis of standard deviation tan te 1 db resolution. X Aj ij i 10 lg 10 X C db (1) i ( L X ) / 10 w j

Buildings 2015, 5 1394 were j is te index of te spectrum No. 1 to calculate C or No. 2 to calculate Ctr according to ISO 717-1 [21]; i is te index of frequencies; Lij is te level indicated in ISO 717-1 [21] at frequency i for spectrum j; Xi is one of te quantities considered, Ri, Dni or DnTi; at frequency i for te spectrum j; Xw is te single number; and Cj is te spectrum adaptation term C or Ctr if calculated wit spectrum No. 1 or No. 2, respectively. Table 1. sr and ssitu of SNQs of floor (F) and wall (W) in narrow (100 3150 Hz) and enlarged (50 5000 Hz) range [7]. Narrow Range 100 3150 Hz Enlarged Range 50 5000 Hz s situ s r X X + C X + C tr X 01 + C X 01 + C tr X 01 + C X 01 + C tr F-D nt 1.3 1.3 1.5 1.3 1.5 1.4 2.8 F-D n 1.2 1.2 1.5 1.3 1.4 1.4 2.8 F-R 1.2 1.2 1.5 1.3 1.5 1.4 2.7 W-D nt 0.7 0.9 1.2 0.9 1.2 0.8 1.4 W-D n 0.9 0.9 1.3 0.8 1.2 0.8 1.4 W-R 0.8 0.9 1.3 0.9 1.2 0.8 1.4 F-D nt 0.7 0.6 0.6 0.5 0.7 0.6 1.3 F-D n 0.5 0.5 0.7 0.5 0.7 0.6 1.3 F-R 0.5 0.6 0.9 0.5 0.7 0.6 1.3 W-D nt 0.2 0.2 0.3 0.2 0.2 0.2 0.3 W-D n 0.3 0.3 0.3 0.2 0.2 0.2 0.4 W-R 0.2 0.2 0.4 0.2 0.2 0.2 0.4 Te internal partitions considered in tis RRT were a ligtweigt wall and a eavy floor. It was demonstrated tat te uncertainties of ligtweigt samples are lower tan te uncertainties of eavy types of construction; terefore it will be important for datasets of different constructions to be considered separately. A similar difference between te uncertainty of eavy and ligtweigt test samples was sown by Dijckmans and Vermeir [23] wo made a numerical investigation of te repeatability and reproducibility of laboratory sound insulation measurements by investigating bot te pressure metod and te intensity metod. Dijckmans and Vermeir [23] found tat for large, eavy test elements, like concrete walls, te reproducibility in te lowest frequency bands is not improved by using te intensity metod, wile, for double plasterboard walls, te teoretical uncertainty is decreased by 1 db by using te intensity metod. Te results of Table 1 sow tat te one-tird-octave band uncertainty at LF sligtly affects te SNQs in te enlarged range plus C spectrum adaptation term but greatly affects (almost double tan te narrow range standard deviation) te SNQs in te enlarged range plus Ctr spectrum adaptation term. Tis is mainly due to te fact tat te spectrum adaptation term Ctr considers predominantly te low-medium frequencies noise components. In teir recent study on te correlations and implications of SNQ for rating airborne sound insulation in te frequency range 50 Hz to 5 khz, Garg and Maij [24] sowed tat Rtraffic (as defined in ISO CD 16717-1 [25] and corresponding to Rw + Ctr50 5000) is igly sensitive to low frequency sound insulation as compared to te current SNQ and Rliving (as defined in ISO CD 16717-1 [25] and corresponding to Rw + C50 5000). Finally, te measurement uncertainty in te low frequency range

Buildings 2015, 5 1395 (due to te presence of te normal modes of vibration, tat imply tat at te first tree one-tird-octave bands te measured levels can be strongly influenced by te measurement position) is too ig to justify te decision to perform field measurements down to low frequencies, and terefore te scientific evidence for including te low frequency range sould be significantly improved. Moreover, te fact tat te iger uncertainty at LF is not well represented in te SNQs uncertainty confirms tat furter studies are needed to better understand all te implications of te inclusions of LF in te SNQs, from bot a pysical point of view and from a legislation point of view. Garg and Maij [24] found interconversion equations applicable for sandwic gypsum constructions and roof constructions. Tey stressed te fact tat testing of sound transmission loss caracteristics in te extended frequency range of 50 Hz to 5 khz also implies te need to reformulate te sound regulation requirements in buildings including te low frequency spectrum adaptation terms. Some recent studies [26 29] on te uncertainty of SNQs extended to te low frequencies range sow an increase in te SNQs uncertainty due to te LF extension, confirming te results found in tis RRT. Man and Pearse [26] studied te effect on uncertainty of expanding te frequency range included in te calculation of te single number ratings, using laboratory measurements of 200 ligtweigt walls as data. Tey found tat te uncertainty of te single number ratings is igly dependent on te sape of te sound reduction index curve. Te uncertainty obtained for Rliving (Rw + C in te enlarged frequency range) was greater tan tat of te traditional weigted sound reduction index for 98% of te 200 ligtweigt building elements included in te evaluation. Hongisto et al. [27] focused teir study on te two most important SNQs proposed by ISO CD 16717-1 [25]; tat is, Rtraffic (Rw + Ctr in te enlarged frequency range) and Rliving (Rw + C in te enlarged frequency range), and ow teir reproducibility values differ from te reproducibility values of teir counterparts Rw + Ctr and Rw. Tey found tat te reproducibility values of te proposed single-number quantities (50 5000 Hz; Rliving, Rtraffic) are larger tan te reproducibility values of te present SNQs (100 3150 Hz; Rw, Rw + Ctr) wit sound insulation measurements made wit te pressure metod; wit te sound intensity metod, te reproducibility values increased very little. Macimbarrena et al. [28] presented an alternative procedure, aiming at evaluating te need of performing individual uncertainty calculations and te effect of extending te frequency range used to calculate sound insulation single number quantities. For tis purpose tey performed calculation in a set of 2081 field airborne sound insulation measurements on 22 different types of separating walls partitions of in situ airborne sound insulation measurements. Te results of Macimbarrena et al. [26] sow tat te frequency range used for te evaluation affects te uncertainty of te single number quantity. In almost all te cases sown in teir paper, te uncertainty is increased wen te frequency range is extended. António and Mateus [29] studied te influence of low frequency bands on airborne and impact sound insulation single numbers for typical Portuguese buildings. Tey found tat te uncertainty is iger for te DnT,w + Ctr descriptor tan for DnT,w + C, confirming wat was found in tis RRT. Tey also found tat wen te low frequency bands are included in te calculation, te uncertainty of te descriptor increases on average and tis increase is more evident wen te adaptation term is for a spectrum of traffic noise.

Buildings 2015, 5 1396 2.1.2. Façade Sound Insulation Te uncertainty of field measurements, in particular façade sound insulation, as not been compreensively investigated. Tere is only one example in te literature of a Round Robin Test conducted on a window of a façade [12]. In te study about te uncertainty of façade sound insulation [8], te measurand was a prefabricated concrete façade wit a 4 mm single glazing wood-aluminum frame window wit a MDF (Medium Density Fiberboard) sutter box. Te façade is situated at first floor level. Nine teams coordinated by ITC-CNR were involved in tis study; eac of tem as replicated te tests five times, including te reverberation time. One laboratory sowed a significant presence of stragglers and outliers. After a statistical examination of tis result, te laboratory was excluded. In fact, it turned out tat te random effect estimated for laboratory was, in absolute value, te igest value [8]: te Grubbs test [30,31] for one outlier identified te laboratory as te first outlier. Terefore ere are te eigt reported laboratories results. In tis study, te igest values of sr and ssitu were found at te frequencies of 50, 63 and 80 Hz. Tat paper [8] also underlined tat te uncertainties in Dls,2m,nT are eavily contaminated by te inappropriateness of te reverberation time correction at low-frequencies and a comparison between te uncertainties of te standardized level difference Dls,2m,nT and te level difference Dls,2m sows te magnitude of te reverberation time at low frequencies (see Figure 2). Tis influence is noticeable in particular at 63 Hz and at 80 Hz, wile at 50 Hz te uncertainties of Dls,2m,nT and Dls,2m are coincident. Figure 2. Comparison between te in situ and repeatability standard deviation of Dls,2m,nT and Dls,2m [8] and te reproducibility, in situ and repeatability standard deviation of ISO 12999-1 [5].

Buildings 2015, 5 1397 Te variations between laboratories at low frequencies are still very ig even if te reverberation time correction is not included in te calculation (i.e., just considering Dls,2m), wic implies tat for te sound pressure level measurements te low frequencies also ave a ig uncertainty. Te ssitu and sr beavior of Dls,2m is similar to te beavior of te uncertainties of ISO12999-1 [5], wic increase steadily and rapidly below 100 Hz. Tus te trend of te standard deviation curve at low frequencies of in situ reproducibility and repeatability standard deviation calculated from te RRT study is attributable to te reverberation time measurements. In Table 2 are sown te SNQs uncertainties, in terms of repeatability and in situ standard deviations. Te SNQs were determined according to ISO 717-1 [21] sifting te reference curve bot in steps of 1 db and 0.1 db (subscript 01), toward te measured curve, until te mean unfavorable deviation is as large as possible, but not more tan 32 db; all te laboratories involved in te RRT ave followed tis procedure. Te sift in increments of 0.1 db was evaluated because te 2013 update of te ISO 717-1 [21] provides for increments of 0.1 db for te expression of uncertainty. Te SNQs plus spectrum adaptation terms C and Ctr according to ISO 717-1 [21] in te extended range (from 50 to 5000 Hz), bot at integer and wit one decimal place (subscript 01) were calculated using Equation (1). Table 2. ssitu and sr of SNQs, calculated as one of te levels j of RRT [8]. Frequency Range SNQs s situ s r Dls,2m,nT,w 0.8 0.3 Dls,2m,nT,w + C 1.0 0.4 narrow range Dls,2m,nT,w + Ctr 1.1 0.3 100 3150 Hz Dls,2m,nT,w01 0.9 0.3 Dls,2m,nT,w01 + C 1.0 0.2 Dls,2m,nT,w01 + Ctr 1.1 0.3 enlarged range 50 5000 Hz Dls,2m,nT,w01 + C 0.9 0.2 Dls,2m,nT,w01 + Ctr 1.1 0.3 In te study about te airborne sound insulation [7], it was found tat te extension at low frequencies range increases te uncertainty of te SNQs. In te case of te façade, calculating te SNQs uncertainty andling te SNQs values as a level of te RRT itself (see Table 2), no significant differences are observed weter including or excluding te low frequencies. In tis case, te low frequency uncertainty is not well reflected in te SNQs uncertainty. Considering te extension to low frequencies, te suitability of te reference spectra for rating airborne sound insulation sould be validated. On tis topic, Masovic et al. [32] made a study on te suitability of ISO CD 16717-1 [25] reference spectra for rating airborne sound insulation. Te ISO CD 16717-1 [25] spectra living and traffic correspond to te reference spectra C (50 5000 Hz) and Ctr (50 5000 Hz) of ISO 717-1 [21], respectively. Masovic et al. [32] demonstrated, wit an extensive noise monitoring in a number of dwellings recordings of 38 potentially disturbing activities, tat te reference spectrum for living noise (Lliving), sould be redefined to better matc te typical spectrum of noise in dwellings because it seems to be rater ig at lower frequencies, especially below 100 Hz. Moreover, in te case of noise generated by sources of music wit strong bass content te reference spectrum for traffic noise (Ltraffic) seems to be more appropriate above 100 Hz tan Lliving. Tis could suggest one of te reasons wy te low

Buildings 2015, 5 1398 frequencies uncertainty is not adequately reflected by te SNQs uncertainty extended to low frequencies and sould be considered deeper before deciding to perform measurements down to LF range. Terefore, considering tis kind of façade (prefabricated concrete façade wit a single glazing window and wit a sutter box) including te low frequencies range in te façade sound insulation measurements, brings no obvious advantage, but rater te disadvantage of complicating and lengtening te measurement. In literature, tere are some studies (e.g., Rindel [33] and Park and Bradley [34]) on te annoyance of noise from neigborood at low frequencies tat stress te importance of investigating te LF noise; neverteless, at present time, effective protection systems against low frequency noise are still an open callenge bot for researcers and components manufacturers, as underlined by Prato and Sciavi [35]. Hongisto et al. [27] suggested tat scientifically valid socio-acoustic evidence for te need to include te frequency range 50 80 Hz sould be significantly improved before deciding tat te low frequency measurements are included in te calculation of te SNQs. Last but not least, if LF measurements are aimed at te protection against LF noise, te fact tat te ig uncertainty of te one-tird octave LF band affects te reliability of te performance of te test element implies tat te potential effectiveness of te protection system against low frequency noise is not quantifiable. A prefabricated concrete façade wit a PVC frame wit double glazing 4/12/4 window was tested in te furter RRT study concerning façade sound insulation uncertainty [10], focused on te new low frequencies measurement procedure stated in ISO/DIS 16283-3 [36], tat will soon replace te standard ISO 140-5 [11]. Ten teams, coordinated by ITC-CNR were involved in tis RRT, eac of tem operating wit its own equipment and replicates te tests 5 times, including te new low frequencies procedure (explained below) and te reverberation time measurements. All teams performed measurements following te global loudspeaker metod, wic yields te level difference of a façade in a given place wit respect to a position 2 m in front of te façade. All teams positioned te outside micropone 2 m in front of te façade, and te loudspeaker on te ground, wit te angle of sound incidence equal to 45 ± 5 ; as positioned directly in front of te façade by some teams, and in a lateral position by oter teams. Te statistical analysis of te data provides a tree-step procedure for te identification of stragglers and outliers. Following tis procedure, two teams were identified as outliers and excluded because tey sowed a significant presence of stragglers and outliers starting from 500 Hz to 3150 Hz [10]. Te comparison of standard deviation values, repeatability and in situ standard deviation, from RRT (calculated for bot Dls,2m,nT and Dls,2m) and from ISO 12999-1 [5] are plotted in Figure 3. Regarding te low frequency range (from 50 to 80 Hz), te reasons for te ig values of sr and ssitu can be sougt in te presence of te normal modes of vibration, in fact at te first tree one-tird- octave bands (50, 63 and 80 Hz), te measured levels can be strongly influenced by te measurement position. At low frequencies, te ssitu and sr beavior of bot Dls,2m,nT and Dls,2m is not similar to te beavior of te uncertainties of ISO 12999-1 [5], in terms of reproducibility sr and in situ standard deviation, wic increase steadily and rapidly below 100 Hz, as it can be seen in graps of Figure 3. Contrary to wat was found in te previous RRT [8], tis difference is not attributable to te reverberation time measurements. Tis different beavior could be attributable to te differences of te façade test samples: te façade of te previous RRT [8] is a prefabricated concrete façade wit a 4 mm single glazing wood-aluminum frame window wit a MDF sutter box; te façade of te second study is a prefabricated

Buildings 2015, 5 1399 concrete façade wit a PVC frame wit double glazing 4/12/4 window. Also te loudspeaker position could be relevant and its influence is under investigation. Figure 3. Comparison of standard deviation values from RRT (calculated for bot Dls,2m,nT and Dls,2m) and from ISO 12999-1 [10]. Wit respect to te ig frequency range, in particular at 4000 and 5000 Hz, te RRT and ISO 12999-1 [5] standard deviations values sow te same beavior, i.e., an increase wit frequency, but te RRT ssitu values are iger tan te ISO 12999-1 [5] values. Moreover te RRT ssitu values are iger tan te low frequency ssitu values of bot RRT and ISO 12999-1 [5]. Tis is probably due to te different positions of te loudspeaker wit respect to te façade [10] and it is still under investigation. In te previous RRT [8], were all te teams involved placed te loudspeaker in te same position (directly in front of te façade), te ig frequency uncertainty was lower, in particular lower tan ISO 12999-1 [5] values and muc lower tan te low frequencies uncertainty. In te first RRT on façade sound insulation [8] a beavior similar to te beavior found by Lang [12] in te Austrian RRT was observed, were te RRT values exceed te values of te ISO 140-2 [37] (te standard on acoustics measurement uncertainty available at te time of Lang s RRT) in te range of mass-spring-mass resonance frequency and in te range of te coincidence frequency of te double glazing. Lang suggests tat suc beavior may be caused by te difficulty of arranging te loudspeaker at an angle of incidence of 45. Te first RRT [8] faced no difficulty wit te arrangement of te loudspeaker at an angle of incidence of 45. Suc beavior is tus exclusively attributable to te nature (i.e., critical frequencies) of te measurand itself. However, te uncertainty dependence from te loudspeaker position could be found at ig frequencies as sown in te second RRT [10] and, as already said, it must be more deeply investigated. Berardi et al. [38] and Berardi [39] considered te position of te loudspeaker as a variable, but its influence on te ig frequencies was not compreensively evaluated.

Buildings 2015, 5 1400 In tis RRT [10] all te participating laboratories repeated te measurements wit te low-frequency procedure included in te upcoming standard ISO 16283-3 (ISO/DIS 16283-3 [36]). In is recent paper Hopkins [40] gives te background to te revision of te ISO 140 standards relating to field measurement of airborne, impact and façade sound insulation tat form te new ISO 16283 series. Te low-frequency procedure was first studied and proposed by Hopkins and Turner [41] in a work about te airborne sound insulation between rooms. For eac of te 50, 63 and 80 Hz bands, tey proposed tat te average low frequency sound pressure level in te room, LLF, be calculated from LISO140-4 (te average sound pressure level in a room measured according to te normative guidance in ISO 140-4) and Lcorner (te corner sound pressure level measured according to te normative guidance in ISO 16283-1) according to: L LF 2 10 10lg 0.1L ISO 1404 0.1L 10 corner [ db] Te low-frequency (LF) procedure is mandatory in case of room volume lower tan 25 m 3. As te volume of te receiving room in tis RRT is 40 m 3, it was possible to compare te results of te two procedures: te LF procedure and te default procedure. Te results of tis comparison, for te LF range are sown in Table 3. Te results refer bot to 8 and to 10 teams, as te two outlier teams tat are excluded from te calculation of standard deviation for te all frequencies considered (from 50 to 5000 Hz), can be included in te evaluation of te LF standard deviation because tese teams sowed a significant presence of stragglers and outliers starting from 500 Hz to 3150 Hz. Table 3. Low frequency sr and ssitu values for te two measurement metods (default and LF) for bot 8 and 10 teams [10]. Standard 50 Hz 63 Hz 80 Hz Deviations Default LF Default LF Default LF s r _10 2.7 db 2.5 db 3.1 db 4.5 db 1.4 db 2.3 db s situ _10 3.1 db 3.1 db 4.8 db 5.5 db 4.0 db 4.1 db s r _8 2.3 db 2.3 db 3.3 db 5.0 db 1.4 db 2.5 db s situ _8 2.9 db 3.2 db 4.3 db 5.2 db 4.1 db 4.2 db Wit te low-frequency procedure tere is an increase of te uncertainty, particularly noticeable at 63 Hz: te repeatability standard deviation increases by about 1.5 db wile te in situ standard deviation increases by about 1 db. Te results sown in Table 3 indicate tat te low-frequency measurement procedure does increase te uncertainty. Tis cannot be attributed to te operators wose experience is well proven; tis aspect is still under investigation. To deal wit te measurement issue in te low frequency domain, Prato and Sciavi [35] and Prato et al. [42] suggest te modal approac. At frequencies below 100 Hz, te acoustic field is non-diffuse, as it is caracterized by large fluctuations of sound pressure levels in space and frequency domains. Because of te inomogeneity of te acoustic field, Prato et al. [42] suggest to move from a statistical approac typical of diffuse sound field (average sound energy) to a discrete one, focused at igest noise and annoyance points, i.e., te points of igest sound pressure level in space (corners) and frequency (resonance modes): te so-called modal approac. 3 (2)

Buildings 2015, 5 1401 In tis RRT [10], it was found tat te differences between including and excluding low frequencies are a little iger for SNQ plus Ctr wen using standard measurement procedure and are very ig for SNQ plus Ctr wen using te LF measurement procedure, as sown by comparing Tables 4 (SNQs witout LF) and 5 (SNQs wit LF), contrary to wat was found in te previous RRT [8] tat sowed tat te differences between including or not te low frequencies were practically negligible. Table 4. Standard uncertainties of SNQs witout low frequencies for te 8 teams [10]. Descriptor (SNQs) s r (db) s situ (db) D ls,2m,nt,w 0.4 0.7 D ls,2m,nt,w + C (100 3150) 0.6 0.8 D ls,2m,nt,w + C tr(100 3150) 0.8 1.0 D ls,2m,nt,w01 0.3 0.7 D ls,2m,nt,w01 + C (100 3150) 0.5 0.8 D ls,2m,nt,w01 + C tr(100 3150) 0.7 1.0 D ls,2m,nt,w01 + C (100 5000) 0.6 1.2 D ls,2m,nt,w01 + C tr(100 5000) 0.7 1.0 Table 5. Standard uncertainties of SNQs wit low frequencies for te 8 teams [10]. Descriptor (SNQs) s r (db) s situ (db) Default LF Default LF D ls,2m,nt,w01 + C (50 3150) 0.5 0.6 0.8 1.0 D ls,2m,nt,w01 + C tr(50 3150) 0.8 1.9 1.0 2.1 D ls,2m,nt,w01 + C (50 5000) 0.6 0.6 1.2 1.3 D ls,2m,nt,w01 + C tr(50 5000) 0.8 1.9 1.0 2.1 Tis different beavior could be attributable to te differences of te façade test samples: te façade of te previous RRT [8] is a prefabricated concrete façade wit a 4 mm single glazing wood-aluminum frame window wit a MDF sutter box; te façade of te second study is a prefabricated concrete façade wit a PVC frame wit double glazing 4/12/4 window. In fact, from te experience derived from many measurements of façade sound insulation [43,44], te lower te insulation of a window, te lower te spectrum adaptation term Ctr and vice versa, te iger te window insulation, te iger Ctr. For tis reason, in te case of te previous RRT [8] (a façade wit low insulation window) te difference between Dls,2m,nT,w and Dls,2m,nT,w + Ctr averages, was not a large one, only 1.5 db, wile in te case of te present study (a façade wit iger insulation window), te difference between te average values of Dls,2m,nT,w and of Dls,2m,nT,w + Ctr,50 5000 is 5.3 db for default measurements and 6.8 db for te low-frequency metod. 3. How to Manage te Cooperative Tests Uncertainty As stated in te introduction, current knowledge in building acoustics suggests tat te best metodology to study te measurements uncertainty is to carry out an Inter-Laboratory Test or a Round Robin Test. Terefore te results of ILTs and RRTs are very important to know te uncertainty magnitude tat is reasonably expected for a measurement result. However, even if an ILT or RRT gives te uncertainty of a measurement metod, te uncertainty magnitude depends also on te measurand.

Buildings 2015, 5 1402 An example of te dependence on te metod can be drawn from te results of uncertainty of façade sound insulation measurements discussed in Section 2.1.2, were te ig frequency uncertainty depends on te loudspeaker position (wic is still under study). On te oter and, te uncertainty magnitude also depends on te test sample, as sowed in Section 2.1.1 concerning te sound insulation of internal partitions were it was found tat te uncertainties of ligtweigt samples are lower tan te uncertainties of eavy types of construction. Te dependence on te measurand, in particular for including or not te LF in SNQs, was also found in te case of façade sound insulation uncertainty (see Section 2.1.2) were te comparison of te two RRTs results igligted tat tat te differences are attributable to te windows, on wic te Ctr coefficient depends: a single glazing window and a double 4/12/4 glazing window. ISO 12999-1 [5] gives te medium uncertainty on all te ILTs and RRTs considered (and available at te time wen te standard draft was being written) in tat standard, for airborne sound insulation, witout distinction of te type of measurand. At te current level of knowledge and due to te number of cooperative tests available, tis seems to be te only way to give an idea of te uncertainty magnitude. Te fact tat te values of ISO 12999-1 [5] are te best estimates for te uncertainty of sound insulation measurements tat can be obtained today, was also underlined by Wittstock [45] in is paper tat describes ow te average uncertainty values standardized in ISO 12999-1 [5] were derived. Terefore, it is important to keep tat standard constantly updated in order to increase te number of available data on wic te average uncertainty values could be calculated. Tis specific standard is inaccurate as far as te façade sound insulation is concerned, because its uncertainty is considered equal to te airborne sound insulation uncertainty; indeed, te façade sound insulation measurement metod is extremely different from te airborne sound insulation measurement metod for party walls and floors. A priori, te reproducibility standard deviation is iger tan te in situ standard deviation because of, as far as reproducibility is concerned, te geometry of te rooms and wall can cange, wile tis is not te case for te in situ standard deviation as defined in Section 2. Because te geometry (i.e., modal beavior) as a large influence at low frequencies, sr is larger tan ssitu (cf. Table 6). Te use of ssitu is tus only appropriate wen te geometry is te same. In te case of façade sound insulation, owever, tere are no literature data tat referred to RRT of te same object in different situations and it will be appropriate in te future tat ISO 12999-1 [5] include tis difference (i.e., reproducibility and in situ standard deviation for façade sound insulation), considering te following: te measurement metod of façade sound insulation is extremely different from te laboratory measurement of airborne sound insulation; te uncertainty at ig frequencies (wic exceed, in te case of te second façade RRT [10], te sr values of ISO 12999-1 [5] as sown in Figure 3) is mainly dependent on te loudspeaker position (as supposed in te case of te second RRT of Façade [10], as said before), and te RRT [12] values exceed te values of te ISO 140-2 [37] in te range of mass-spring-mass resonance frequency and in te range of te coincidence frequency of te double glazing. At te present state of knowledge, te reproducibility standard deviation values included in ISO 12999-1 [5] seem to be te only available uncertainty tat could be used also in te case of façade sound insulation, keeping in mind tat te façade sound insulation measurement metod is very different.

Buildings 2015, 5 1403 Table 6. Standard uncertainties for single-number values in accordance wit ISO 717-1, as per ISO 12999-1 [5]. Descriptor s R db s situ db s r db R w, R w, D nw, D nt,w 1.2 0.9 0.4 (R w, R w, D nw, D nt,w ) + C 100 3150 1.3 0.9 0.5 (R w, R w, Dnw, D nt,w ) + C 100 5000 1.3 0.9 0.5 (R w, R w, D nw, D nt,w ) + C 50 3150 1.3 1.0 0.7 (R w, R w, Dnw, D nt,w ) + C 50 5000 1.3 1.1 0.7 (R w, R w, D nw, D nt,w ) + C tr,100 3150 1.5 1.1 0.7 (R w, R w, D nw, D nt,w ) + C tr,100 5000 1.5 1.1 0.7 (R w, R w, D nw, D nt,w ) + C tr,50 3150 1.5 1.3 1.0 (R w, R w, D nw, D nt,w ) + C tr,50 5000 1.5 1.0 1.0 Terefore, in te case of a single measurement, te uncertainty tat sould be associated to tis measurement is te reproducibility standard deviation given in ISO 12999-1 [5] multiplied by te appropriate coverage factor to obtain te expanded uncertainty. Now, considering wat was stated in te introduction, wen reporting te result of te measurement of a pysical quantity, it is compulsory tat some quantitative indications of te quality of te result be given. Suc an indication sould be independent on te final use of te results (verification of a requirement or determination of predicted values), and sall be stated as follows, as provided by GUM [2] and ISO 12999-1 [5]: Y y U (3) were Y is te measurand; y is te best estimate (obtained troug te measurement) of te value attributable to te measurand; and U is te expanded uncertainty, calculated for a given confidence level for te two-sided test, defined as te product of te measurement uncertainty u (wic is te reproducibility standard deviation sr) wit a coverage factor k. Terefore, for example, for a single measurement of te airborne sound insulation of a partition floor R w (C;Ctr) = 53 ( 1; 4), considering te values given in ISO 12999-1 (see Table 6), te airborne sound insulation of tis partition wall sall be given to one decimal place (R w = 52.6; C = 1.0; Ctr = 4.1) to state also its uncertainty and sould be designated as [5]: ' R w R ' w ' R w 52.6 2.4dBk 1.96 two sided, 51.6 2.6dBk 1.96 two sided C, 48.5 2.9dBk 1.96 two sided C, tr were k = 1.96 corresponds to a confidence level of 95% for a two-sided test. On te oter and, wen a measurement is made in order to verify a requirement, te expanded uncertainty tat sould be given wit te result, sould be calculated using a coverage factor for one-sided test, as laid down in ISO 12999-1 [5]. Ten te expanded uncertainty sould be added to or subtracted from te measurement result to ceck weter tat measurement result is smaller or larger tan te requirement, respectively. (4) (5) (6)

Buildings 2015, 5 1404 Te Italian standard on te acoustic classification of building units UNI 11367 [46] first considers te measurement uncertainty from RRTs as a basis for te expanded uncertainty U. Wen a national regulation as to be met, te coice of te confidence level is very important. Te Italian standard on te acoustic classification [46] as faced for te first time te problem related to te confidence level. In te case of measurement uncertainty, te standard recommends to use a coverage factor k for one-sided test equal to 1, wic corresponds to an 84% probability; for buildings performances, in fact, in order to meet te limit, it is not realistic to use a 95% or 90% confidence level, wic is normally used in oter contexts. As te update of te ISO 717-1 [21] allows applying te weigting procedure by 0.1 db steps for te expression of measurement uncertainty, it could now possible also be to use, in building acoustics, a coverage factor k for one-sided test equal to 1.65 corresponding to a 95% confidence level. Generally, wen measurements are made to verify te acoustic requirements of buildings, one single measurement migt not be enoug to tis end, and terefore more measurements and more results for te same requirement are necessary. In tis case, te measurement uncertainty is combined in a certain way wit te uncertainty due to te number of tests performed. 4. Sampling Tere are two different types of surveys tat can be used to analyze te acoustic requirements of building units, or buildings: a census (te entire population is taken into account) or a sample survey (only a part of te elements tat make up te population are considered). For building acoustics, a sample survey is te best solution in terms of cost and time. To make meaningful comparisons wit bot national regulations and acoustic classification, it is terefore necessary to determine te type and amount of te measurements. In order to make any sample survey on certain features (acoustical) of a finite population, it is essential to formulate a strategy of selection, wic is closely connected wit te purposes, te cost and te execution time of te survey. In addition, te sample obtained from it, is te only valuable information tat could be used for te interpretation of te results. Among te different sampling strategies currently available, te two main ones used in building acoustics, for te time being, are te following: te stratified sampling as adopted by UNI 11367 [46] (see next section) and a sampling procedure taking into account a certain percentage associated wit a selection criterion as adopted by UNI 11444 [47] and proposed by ISO/WD 19488 [48]. Only te former strategy (stratified sampling) includes te sampling uncertainty. Te strategy of UNI 11444 [47] consists in te selection of a minimum number of Building Units (BUs): not less tan 10% of te total amount of BUs composing te building system and not less tan 2 BUs, if te total amount is 4, and not less tan 3 BUs for building systems up to 30 BUs. Tese BUs must be te most critical BUs from an acoustic point of view. Te selection of te most critical BUs must take into account all te critical acoustic features of te building elements of te BU. Te selection criteria for eac type of acoustic performance (façade sound insulation, sound insulation of orizontal and vertical partitions, impact sound insulation and equipment noise level) are stated in standard UNI 11444 [47]. Tis standard does not include te sampling uncertainty but, for eac measurement, it includes te measurement uncertainty as stated in UNI 11367 [46].

Buildings 2015, 5 1405 Te standard proposal ISO/WD 19488 [48] considers, as a general principle, tat, wen verifying te acoustic class of a unit, a sufficient number of measurements of eac relevant acoustic caracteristic must be performed in order for te result to represent te unit. It also suggests tat care sould be taken to include te critical site/rooms, e.g., partitions wit critical flanking constructions. At te current stage, te proposal includes neiter te sampling uncertainty nor te measurement uncertainty, but it considers tat compliance is granted if te average results comply wit te class limits and no individual result deviates unfavorably by more tan 2 db. Moreover, if classification for different dwellings, rooms or acoustic caracteristics varies, te classification assigned is te minimum class obtained. Considering te pros and cons of tese two sampling strategies, te first ting to keep in mind is te scope of te measurements; i.e., to determine a class witin te acoustic classification or to verify te legal requirements. In te former case, it is obvious tat a value as close as possible to te value of all te elements is suitable. In te latter case, te scope is to identify te worst acoustic performances and to verify if also te critical site/rooms is/are in compliance wit te legal requirements. Te stratified sampling strategy allows increasing te efficiency of a sampling plan, witout increasing te sample size. Wit tis strategy it is possible to obtain te best representative value of a class to be attributed to te entire building system, as if te entire population were taken into account. Anoter pro is te stratified sampling uncertainty related to te final result tat gives a confidence level, wic is important bot for te owners and te builders. Te con of tis strategy is tat it requires a large number of measurements (a minimum of tree measurements for eac omogeneous group). A strategy tat takes into account a certain percentage of te population, including all te critical site/rooms, could not be representative of all situations but would give te worst results and terefore, if tis result complies wit te legal requirements, te wole building complies wit tem. On te oter and, not all te critical site/rooms may ave been taken into account and terefore te confidence level and te sampling uncertainty to be associated wit te results is not known. Moreover, te sampling strategy proposed in ISO/WD 19488 as te obvious drawback tat it cannot guarantee to ave spotted all critical situations: for example a workmansip failure tat cannot be detected by visual inspection can be identified only after te measurements. Tus, a sampling criterion based on generic rules cannot find it. However, te con of tis strategy is tat in general te number of measurements is limited. 4.1. Stratified Sampling Te stratified sampling is te most direct procedure tat allows increasing te efficiency of a sampling plan, since it allows reducing te order of magnitude of te sampling error witout increasing te sample size. Stratification is made possible by means of additional information about one or more caracters of te population, wic is about te structure of te population itself. Tis allows, based on informed coice, dividing te population into a number of layers as omogeneous as possible, as meaning tat witin eac layer, te considered caracter as a lower variability. A simple random sample is extracted from eac layer; terefore tere are as many simple samples as tere are layers. Tese samples are independent of eac oter and can ave different sample sizes. Te stratification, due to te way it is implemented, allows obtaining an improvement in te estimates for te same sample size, or to contain te sample size at te same level of efficiency [49].

Buildings 2015, 5 1406 Considering te above mentioned advantages offered by te stratified sampling, tis latter is te solution adopted by UNI 11367 [46] in te case of classification of serial type buildings. Te part of te Italian standard on te classification of buildings and building units tat refers to te stratified sampling procedure can be applied in te case of a serial type building. Te stratified sampling procedure is based on te concept of omogeneous group. Te population of all te building elements tat ave to be measured for te acoustic classification as to be divided in te omogeneous groups tat are defined in te Italian standard on classification. Referring to UNI 11367 [46], generally, a set of test items can be considered omogeneous and terefore subject to a possible sampling (in reference to a specific requirement), if te following conditions are satisfied: item dimensions (wit 20% tolerance); dimensions (wit 20% tolerance wit respect to te volumes) of bot transmitting and receiving rooms were te test item is located; te same test metodology; stratigrapy, materials and surface mass; structural constraints (flanking transmissions); presence of equipment passing troug te test item; installation tecniques. In tis section an example is given wit reference to te paper presented by te autors at te 38t National Congress of te Acoustical Society of Italy in 2011 [50], concerning te acoustic classification of a building system of a total volume of about 15,000 m 3, consisting of two similar buildings, identified as body A1 and body A2, on tree floors, wit apartments on te ground floor, first and second floor and, in te body A1, a tird floor attic. In total, te building system consists of 47 Building Units (BUs), distributed according to teir type: six four-room apartments, eigt tree-room apartments, 25 two-room apartments and eigt studios. Te building system was considered a serial type building system, based on te following considerations: it is possible to identify a typical floor (see Figure 4) in wic te distribution of BUs is symmetrical wit respect to te stairwells; te two-room apartment type is repeated 25 times; te rooms wit te same intended use (bedrooms, living rooms, kitcens, etc.) ave te same sape and size. Figure 4. Typical floor of te building system considered: te BU typologies are igligted, in green te four-room apartments, in red te two-room apartments and in blue te studio [50]. For te application of te stratified sampling procedure defined in UNI 11367 [46], it would ave been sufficient to use a minimum number of items to be tested equal to at least 10% of te total number of elements of te omogeneous group and not less tan tree for eac omogeneous group. However, in order to obtain te most useful data for a critical examination of te results, te number of items to be tested was iger tan te minimum required. In particular, 84% of te vertical partitions were measured. For some requirements, te number of items to be tested of some omogeneous group was equal to two, wic is less tan te minimum required of tree for reasons related to te impossibility to perform