Evaluating Laboratory Measurements for Sound Insulation of Cross-Laminated Timber (CLT) Floors: Configurations in Lightweight Buildings

Abstract

Cross-laminated timber (CLT) floors with supplementary layers or floating floors comprise a common solution in new multistory timber structures. However, bare CLT components provide poor sound insulation, especially in low frequencies during structure-borne sound propagation. Thus, floor configurations in wooden buildings deploy more layers for improved acoustic behavior. Twelve contemporary CLT floors were analyzed after laboratory measurements of airborne sound reduction and impact sound transmission utilizing the following indicators: $R_w$, $R_{w, 100}$, $R_{w, 50}$, $L_{n,w}$, $L_{n,w,100}$, and $L_{n,w,50}$ (per ISO 10140, ISO 717). An increase in sound insulation was achieved thanks to added total mass and thickness, testing layers of the following: elastic mat for vibration isolation, wool insulation, gypsum boards, plywood, concrete screed, and wooden parquet floor. The results indicate that multilayered CLT floors can provide improvements of up to 22 dB for airborne sound and 32 dB for impact sound indicators compared with the bare CLT slab. Floating floor configurations with dry floor solutions (concrete screed) and wooden parquet floors stand out as the optimal cases. The parquet floor provides a 1–2 dB improvement only for impact sound indicators in floating floor setups (or higher in three cases).

1. Introduction

This publication concerns contemporary configurations of wooden floors for lightweight CLT-frame buildings. Their production is part of a growing trend in the construction industry [1,2,3,4,5]. The eco-friendly properties of wood as a building material and its structural properties have made CLT a promising material for sustainable buildings [6,7]. This happens because the wood for CLT is renewable and has a considerably lower carbon footprint than concrete [6,8]. The CLT components can be produced with extended prefabrication, implying design flexibility and high protection from moisture, and provide great thermal performance [3,6,9]. All of these characteristics make CLT appropriate for the sustainable design of modern structures [2,5,9,10].

For instance, small- or large-scale high wooden buildings have been erected in Scandinavia [11], Canada [1], and Australia [5]. The construction of wooden multistory buildings was approved in Sweden in 1994 and has been developing constantly [11,12]. Thanks to this development, acoustic issues regarding CLT as structural material have arisen [5,9,11,12]. Sound insulation in CLT components can be insufficient owing to certain properties such as low mass, density, or stiffness [5,13], as well as because of extended flanking transmission [2]. Measurements have shown that CLT components offer better sound insulation than traditional concrete structures at middle and high frequencies, but offer considerably worse sound insulation at low frequencies below 100 Hz [10,12,14].

Further, multistory buildings are common structures with multiple tenants who often complain about noise and the quality of the sound environment [15]. Early research for sound insulation and noise annoyance in Swedish CLT or other wooden buildings (Akulite, Acuwood projects) revealed shortcomings owing to low frequency sound propagation and disturbances for occupants [12,16,17]. Most complaints in CLT apartments concern impact sound, namely noise from neighbors stepping barefoot or walking on heels [18].

Acoustic comfort studies also found significant differences in sound environment perception between the users of heavyweight concrete and lightweight CLT dwellings [14]. To correlate acoustic descriptors to residents’ noise annoyance, the use of low frequency measurement data from 50 Hz was suggested [12,16,17,18]. Thus, the obligatory limit from 100 Hz in ISO 717 −1 and −2 [19,20] is under debate. The same applies to regulations, threshold values, and the lack of harmonization [15,21], while in Nordic countries, the extended frequency range from 50 Hz has been adopted [22].

Floor coverings and resilient layers with insulation materials are commonly used for acoustic problems in buildings; floating floor solutions are preferred for enhanced insulation and they offer reduced subjective annoyance due to impact sound [5,10,23]. Modifications with additional mass and damping on a floor system represents a successful strategy for insulation improvements [5,10]. Some studies show measured data for wooden floors of the three usual types: wooden joist floors [24], CLT massive wooden slabs [2], or CLT timber volume elements (TVEs) [10].

However, limited measurement data for sound insulation of massive CLT floors have been published and no reliable prediction models for sound transmission in CLT structures have been found [2,3,5,10,25]. Although several approaches for modeling sound insulation of CLT floors have been developed, including analytical models [8,11,26], finite element methods [4,21,27,28], wave propagation modeling [26], statistical energy analysis [25], and modeling with neural networks [29], to estimate the acoustic performance of CLT floors, acousticians still rely on field or laboratory measurements, which still provide only indications [3,8,10,21].

Moreover, reference data are provided for concrete or wooden joist floors in ISO 10140-5 [30], but not for massive CLT floors, while this necessity has been reported [9,31]. This problem amplifies considering variations between the similar sample structures from different wood types (base material), manufacturers, and productions [21,32]. There is also limited characterization with acoustic descriptors for CLT components, while studies focus mostly on structural behavior [25,33]. Lately, airborne and impact sound insulation have attracted interest, especially for floor partitions [33].

In this manuscript, the standardized airborne and impact sound insulation of a bare CLT slab and 12 configurations with extra layers are investigated. The study aims to provide information on improved solutions for acoustically efficient CLT floor configurations, with laboratory data and insights in measurements. Each case, as a combination of materials and group properties, is characterized acoustically, attempting to narrow the widely reported gap in acoustic data [8,9,31,32,33,34].

2. Materials and Methods

The presented study concerns sound insulation of a cross-laminated timber (CLT) floor with a thickness of 180 mm, which is the reference structure and has additional layers for enhanced acoustic behavior in 12 versions (configurations). The reference floor is 4000 mm long and 3100 mm wide and comprises a five-ply massive CLT slab, as illustrated in Figure A4 (Appendix A).

The materials involved in this study are tabulated with their relevant properties in

Table 1. Overview of material properties of the study.

Material	Thickness (mm)	Density (kg/m3)	Mass per Unit Area (kg/m2)
Cross-laminated timber (CLT)	180	523.9	94.3
Gypsum board	12.5	1134.6	14
Plywood	12.5	641.0	8
Concrete screed 1	30	1850.0	55.5
Concrete screed 2	60	1850.0	111
Glass wool	20	330.1	6.6
Vibration isolation mat A	12	359.8	4.3
Vibration isolation mat B	12	340.2	4.1
Parquet floor (wooden)	13	573.7	7.5

, and the examined floor configurations are presented in

Table 2. Overview of test floor cases with acoustic descriptors in dB.

FLOOR	SECTION PLAN	CONFIGURATION	$R_w$	$R_{w,100}$	$R_{w,50}$	$L_{n,w}$	$L_{n,w,100}$	$L_{n,w,50}$
0 (REF.)		CLT 180 mm	32	32	32	86	80	80
1		1 * Gypsum 12.5 mm Vibration mat B 12 mm CLT 180 mm	41	40	40	67	67	67
		Case 1 + Parquet 13 mm	43	41	41	66	66	66
2		1 * Plywood 12.5 mm 1 * Gypsum 12.5 mm Vibration mat B 12 mm CLT 180 mm	44	42	42	65	65	65
		Case 2 + Parquet 13 mm	45	43	43	65	65	65
3		1 * Gypsum 12.5 mm Vibration mat A 12 mm Vibration mat B 12 mm CLT 180 mm	45	43	43	65	64	65
		Case 3 + Parquet 13 mm	46	44	44	64	64	64
4		2 * Gypsum 12.5 mm Vibration mat A 12 mm Vibration mat B 12 mm CLT 180 mm	47	45	44	62	62	62
		Case 4 + Parquet 13 mm	47	45	45	62	62	61
5		2 * Plywood 12.5 mm 2 * Gypsum 12.5 mm Vibration mat A 12 mm Vibration mat B 12 mm CLT 180 mm	48	46	46	61	60	61
		Case 5 + Parquet 13 mm	50	48	48	60	60	61
6		2 * Gypsum 12.5 mm Vibration mat B 12 mm Vibration mat B 12 mm CLT 180 mm	50	48	48	62	62	62
		Case 6 + Parquet 13 mm	51	49	49	62	62	62
7		2 * Gypsum 12.5 mm Vibration mat B 12 mm Glass wool 20 mm CLT 180 mm	51	49	49	60	60	61
		Case 7 + Parquet 13 mm	50	48	47	59	60	62
8		Concrete screed 30 mm Vibration mat B 12 mm Glass wool 20 mm CLT 180 mm	54	52	51	60	59	61
		Case 8 + Parquet 13 mm	53	51	51	57	58	61
9		Concrete screed 60 mm Vibration mat B 12 mm Glass wool 20 mm CLT 180 mm	54	52	52	61	59	60
		Case 9 + Parquet 13 mm	54	52	52	54	56	58
10 *		Concrete screed 60 mm Vibration mat B 12 mm Glass wool 20 mm CLT 180 mm Airgap 25 mm 2 * Gypsum 12,5 mm	-	-	-	-	-	-
		Case 10 + Parquet 13 mm	59	57	54	52	52	57
11		Concrete screed 30 mm CLT 180 mm	43	42	42	93	79	79
		Case 11 + Parquet 13 mm	43	42	42	70	70	70
12		Concrete screed 60 mm CLT 180 mm	46	45	45	87	75	75
		Case 12 + Parquet 13 mm	45	44	44	66	66	66

* Some values are missing because floor 10 was measured only once with a parquet floor.

. The thickness of the layers, their density, and the area density (or mass per unit area) are commonly used in product specifications and research [2,8,34].

Notably, some details may be unique in the measurement setup:

(a) No glue was used between the layers, nor any other connecting elements such as screws or spikes; the additional layers just lay one on top of another, covering the full area of the test floor.

(b) The gypsum board type employed has a higher density (1135 kg/m³) than average plasterboard (usual range between 800 and 900 kg/m³ [13]), because it is a special floor insulation product.

The measurements were performed in the acoustic laboratory of LTH (Lund University, Lund, Sweden) between the years 2020 and 2021, following the ISO 10140 procedures for airborne and impact sound insulation [35,36], in the extended frequency range of 50–5000 Hz in 1/3 octave bands.

Specifically, the airborne sound reduction index R (in dB) was estimated according to ISO 10140-2 [35] for each frequency band as follows:

$$R\left(f\right)=L_S\left(f\right)-L_R\left(f\right)+10log\left(\frac{A_R\left(f\right)}{S_R}\right)$$

(1)

where

• $L_s$ is the measured sound pressure level in the sending (source) room in dB;
• $L_R$ is the measured sound pressure level in the receiving room in dB;
• $S$ is the sample area of the testing partition in m², which is 12.3 m²;
• $A_R$ is the equivalent sound absorption area of the receiving room in m².

Similarly, the normalized impact sound pressure levels $L_n$ were calculated according to ISO 10140-3 [36] as follows:

$$L_n\left(f\right)=L_i\left(f\right)+10log\left(\frac{A_R\left(f\right)}{A_0}\right)$$

(2)

where

• $L_i$ is the measured sound pressure level in the receiving room in dB;
• $A_0$ is the reference equivalent sound absorption area, equal to 10 m².

Then, the single-number quantities (SNQs) for insulation ratings were calculated following ISO 717, with part 1 defining the following: $R_w$, $R_{w, 100}$ (=$R_w+C_{100-3150}$), and $R_{w, 50}$ (=$R_w+C_{50-3150}$) for airborne sound [19], and part 2 defining the following: $L_{n,w}$, $L_{n,w,100}$ (=$L_{n,w}+C_{I,100-2500}$), and $L_{n,w,50}$ ($=L_{n,w}+C_{I,50-2500}$) for impact sound [20].

3. Results

Table 2 summarizes all of the measured configurations in this study with their various materials or layers and the accompanying acoustic descriptors collected from the measurements. Each configuration was measured with and without the wooden parquet floor, which is a typical floor covering in Swedish buildings—mainly in residential buildings, but also in office facilities. The last configurations, namely floors 8–12, represent a subgroup of the so-called wet floor solutions—this term refers to layers of concrete on top of timber floors [2,34]. The cases without concrete screed on top are the dry solutions.

The rating indicators R_(w,50) and L_(n,w,50) with spectrum adaptation terms including the extended frequency range from 50 Hz are of high interest in this study, as suggested in [12,16,22,37,38]. The descriptor quantities mentioned in this text such as R_w (C) and L_(n,w) (C_I), with correction terms C and C_I, refer to the typical spectrum adaptation terms calculated from 100 Hz [19].

3.1. Airborne Sound Insulation Comparison

The airborne sound insulation performance increases from simple cases to multi-layered configurations, as depicted in Figure 1. This situation simply follows a generic rule: insulation ability increases as mass is added to a system (see Table 2) [2,5,13]. The reference CLT floor demonstrates the lowest $R_w$ in this study and the simplest configuration (floor 1 with vibration mat and a gypsum layer) shows an initial increase in reduction with R(f) in 1/3 octave band values ranging from 1.4 dB (at 80 Hz) to 16.5 dB (at 1250 Hz). The improvements in insulation performance ΔR(f) for all cases against the reference floor are depicted in Figure 2.

However, improvements for the bare CLT floor apply only from 160 Hz and higher as a constant trend (Figure 1 and Figure 2). A few cases show no reduction at 125 Hz (floors 1–4) or lower at 50–100 Hz (floors 7,10), but instead amplify the transmitted noise in certain bands. The first eigenfrequency of the combined mechanical systems of all layers (vibration mat, gypsum, and plywood) causes that distinct resonance around 125 Hz. Floor 7 exhibits those first eigenfrequencies even lower at 80–100 Hz, similarly to floors 11 and 12. Floor 10 shows some interactions owing to the suspended ceiling, which, as a layer, helps the overall insulation to increase, but resonates with airborne noise at the lowest frequency bands of 50–63 Hz.

Generally, most R(f) curves show low values at 2000 Hz, which relate to critical frequencies, where a plate-type structure radiates sound with the highest intensity [39]. The cases with fewer layers and mass (floors 1–4) show greater sensitivity to the critical frequency effect, which leads to reduced insulation around 2 kHz. Only floor 10 compensates with high reduction against this phenomenon.

Noticeably, floors 1 and 2 have a single vibration mat, while most configurations rather have two vibration mat layers (floors 3–6) or a vibration mat on top of glass wool insulation (floors 7–10). For cases with a double vibration mat, negative effects at 125 Hz can be observed (Figure 2). Further, floor 5, which has also two gypsum layers and two plywood layers on top of the isolation interlayers, would be expected to be the best in this cluster instead of floor 6 (two gypsum layers on top, the same type of vibration mat in both interlayers), which comes first.

The cases with a vibration mat and a glass wool layer (floors 7–10) provide the highest levels of airborne sound reduction alongside floors 5 and 6 (top 6 curves in Figure 1). They all have $R_w$ values of at least 50 dB. Firstly, floors 6 and 7 show that replacing a vibration mat layer with glass wool may provide a similar outcome, even if the material properties are different. They have $R_w\left(C\right)$ values of 51 (−2) and 50 (−2) dB, respectively. Floors 8 and 9 provide a higher reduction than floors 6 and 7 as wet solutions. Floor 9 with a 60 mm concrete floating layer performs better than floor 8 (30 mm concrete layer) in most frequencies (e.g., up to 3.9 dB improved R(f) values at 500 Hz).

Floor 10 is the only test structure with enhanced insulation and extra layers below the slab, namely, 25 mmm of air gap and two gypsum layers for suspended ceiling. It provides $R_{w, 100}$ = 57 dB and $R_{w, 50}$ = 54 dB. Floors 11 and 12 are also exceptions, as they have only concrete screed layers on CLT without any resilient layers for insulation, providing moderate results.

Figure 2 shows the differences in airborne sound reduction index ΔR(f) levels relative to the reference CLT curve. It depicts a clear trend of insulation improvements above 160 Hz and dips at 125 Hz or below. Summing up, the highest ΔR is 36.8 dB at 1600 Hz for floor 10 (special case with suspended ceiling), but it shows a rather adverse reduction at 50–63 Hz, displaying controversial behavior. Besides floor 10, the best improvement ΔR is of 30.6 dB at 2500 Hz (floor 9) and the worst is of −2.9 dB at 80 Hz (floor 7). Overall, most floors show low airborne sound reduction below 160 Hz, but greater performance above this level.

3.2. Impact Sound Insulation Comparison

The impact sound insulation also increases with the floor case order (Table 2), so the normalized impact sound pressure levels decrease from simple to multi-layered floors, as illustrated in Figure 3. The bare CLT floor shows the highest normalized impact sound pressure levels $L_n$ (worst performance) without any exceptions at frequency bands, as observed for floors 1–4 in the airborne sound reduction at the range of 50–125 Hz. The SNQ for the reference floor is $L_{n,w,100}$ = 80 dB, while it is $L_{n,w}$ = 86 dB without correction adaptation terms.

The first eigenfrequency of the configurations seems to affect most measurements (except floor 8), and thus enhance the impact sound at the 100 Hz band, instead of the 125 Hz band, which was observed previously for airborne sound. Similar, but weaker peaks are found at 160 Hz, probably due to overtone resonances for floors 1–9. Additionally, the critical frequency effects are visible with small peaks of $L_n\left(f\right)$ at around 2 kHz, except floor 10, which exhibits such a peak at 2.5 kHz (Figure 3).

The simplest case (floor 1 with single vibration mat and a gypsum layer) shows improved impact sound pressure levels ranging from 2.8 dB (100 Hz) to 41 dB (5 kHz) relative to the reference. The improvements in $\mathsf{\Delta }{L}_n$ levels against the bare CLT slab appear in Figure 4.

Floors 3–6 (with double vibration mat) show increasing impact sound insulation. In contrast to the airborne sound analysis, floor 5 (two gypsum layers and two plywood layers on top of double vibration isolation) performs slightly better than floor 6 in middle frequencies (only two gypsum layers on top of two vibration mat layers of exactly the same kind). This still defines the final result in terms of SNQ. Floor 5 has a value of $L_{n,w}\left(C_I\right)$ = 60(0) dB and floor 6 has $L_{n,w}\left(C_I\right)$ = 62(0) dB, hence floor 5 provides better impact sound insulation by 2 dB in the weighted index.

Further, floors 7–10 (vibration mat on top of a glass wool layer) provide the best impact sound performance, as observed for airborne sound. In Table 2, floor 7 stands out as the best case with $L_{n,w}\left(C_I\right)$ = 59(1) dB regarding the dry solutions (floors 1–7, without concrete floating floor on top). Most wet floor solutions, namely floors 8, 9, and 10, show the best performance, but not floors 11 and 12.

The non-floating floor wet solutions, floors 11 and 12, demonstrate peculiar behavior and have the worst impact sound insulation with $L_{n,w}\left(C_I\right)$ values of 70(0) dB and 66(0) dB, respectively. The results are even worse for floors 11 and 12 without a parquet floor (Figure A2, Appendix A), underperforming the bare CLT floor values. Those bad ratings change substantially with added correction terms, while they remain the worst acoustically.

In Figure 4, the difference $\mathsf{\Delta }{L}_n$(f) in normalized impact sound pressure levels relative to the bare CLT floor is depicted. The limited improvements at 80–100 Hz (floors 1–9) and the curve drop at 2000 Hz due to critical frequency effects are apparent. The highest contribution to insulation levels is recorded for floor 9 with 54.8 dB at 1600 Hz and the lowest one for floor 8 at 0 dB (80 Hz). A significant comparison is visible for floors 9 and 10; the former has higher contributions above 1 kHz, besides the overall higher performance of floor 10, which has additional suspended ceiling layers. Surprisingly, the improvements of floor 10 are mediocre above 2 kHz, suggesting that the impact sound radiation from the ceiling’s gypsum boards around the critical frequency at 2500 Hz dominates in that range of 2.5–5 kHz. Similar peculiarities have been observed with analytical models in [8] or high variation in statistical models in [29].

3.3. Linear Regression Model Comparison

Another comparison is presented by utilizing the linear model fit of each curve, as applied in previous studies, who used the linear slope as a test parameter [29,40]. Figure 5 and Figure 6 present the linear fits of $R\left(f\right)$ and $L_n\left(f\right)$ (corresponding to the measured data seen in Figure 1 and Figure 2, respectively). The fitted models deploy simple linear regression and the determination coefficient R² corresponds to the explained variance of every model fit in the range 0–1 [41]. All fitted models rely on 21 data points of frequency band levels, hence they cannot be further improved with extra data. The R² values lie between 0.70 and 0.96, which indicate great fitting, except for the impact sound of the reference floor (R² = 0.01, Figure 6). This model depends only on the intercept, estimated as a mean value at 78.7 dB, while the slope is −0.1 without statistical significance.

Airborne sound reduction fits and their slopes in Figure 5 present the linear trends of measurement curves. Floors 1 and 2 have the smallest slopes, after the reference curve, showing their initial trend shifts. Then, floors 3, 4, 5, 6, 8, and 9 show rather similar upward trends, with significant improvements for airborne sound insulation in the presented order. Floors 7 and 10 have the steepest slope, implying radically higher R values with higher frequency. Floor 10 (with added ceiling) demonstrates similar values to floor 9 at 50–100 Hz, while above 200 Hz, it appears to be the most efficient, with floor 9 clearly being second best.

Figure 6 illustrates the linear fits for impact sound curves where the floor cases show decreasing $L_n$(f) with higher frequencies. The reference CLT floor fit looks almost flat with an average value at 78.7 dB (intercept) having SNQs of $L_{n,w,100}$ = $L_{n,w,50}$ = 80 dB and $L_{n,w}$ = 86 dB without correction terms. The latter value is far from the model’s 78.7 dB, which proves the necessity of correction terms with a wide frequency spectrum (including low frequency bands) to yield a representative descriptor value. A first group of floors 1–4 can be observed (highest $L_n\left(f\right)$ estimated). Similar behavior appears between floors 9 and 10, which have the best impact sound performance while floors 5–8 have similar slopes (close parallel trend lines). Evidently, the best cases correspond to higher slopes in absolute value with steeper slopes upwards for $R\left(f\right)$ and downwards for $L_n\left(f\right)$ estimations.

4. Discussion

The sound insulation of the examined floor structures increases overall with the increased total thickness and total mass per unit area. The results verify, for cross-laminated timber (CLT) floors, this common conception for the usual heavyweight structures [2,8,13,16,34,39].

The SNQ descriptors for the reference CLT floor are $R_w$ = 32 dB and $L_{n,w}$ = 86 dB for airborne and impact sound, respectively. The latter impact sound SNQ is representative of five-ply CLT floors with similar thickness, as reported in [9,42], while an average value of $R_w$ = 40 dB was reported in a recent review for airborne sound [9]. Such discrepancies in acoustic performance can be easily explained though. Deviations may occur between CLT floors of similar type and thickness considering variations in the wood type as a raw material [21,31,42]. Changes in wood material have a high probability of shifting eigenfrequencies [43], which can consequently affect the sound insulation results in 1/3 octave bands. Furthermore, significant variations between the same floor types from different producers have been reported regarding fundamental properties such as the bending stiffness and elasticity modulus [32].

All presented floor configurations with resilient layers and floating masses provide acoustic improvements. However, those improvements are limited for airborne sound reduction index R below 125 Hz and for impact sound pressure levels $L_n$ below 100 Hz (frequency bands). This is not unexpected and can be justified for the following reasons:

-

The location of the first eigenfrequencies of the combined system of additional layers;

-

The overall difficulty to block low frequency noise in structures below 200 Hz [13,39];

-

The inability of CLT elements to resist structural vibrations at low frequencies [42,43].

The studied configurations with higher thickness and mass provide better sound insulation than others above the first eigenfrequency peaks, i.e., 100–160 Hz. In the airborne sound analysis, a few configurations resonate at 125 Hz (Floors 1–4) or below. However, for impact sound, no such effects occur. Impact sound displays differences in insulation trends (Figure 3 and Figure 6), but overall, most floors follow similar trends, as observed for airborne sound (Figure 1 and Figure 5). The resonance effects of multi-layered configurations at the critical frequency areas (mostly seen at the 1.6 kHz frequency band) with relatively decreased insulation are observable for both airborne and impact sound measurement data.

Regarding specific materials, the effects of resilient layers and floating floor components are clearly beneficial for enhanced sound insulation. Floors with a single resilient layer of 12 mm (vibration mat) provided moderate improvements of 9–13 dB for airborne sound and 13–21 dB for impact sound; this is in terms of single number quantities (SNQ) compared with the reference CLT. The added floating mass (plasterboard, plywood, wooden parquet) plays a role when comparing floors 1 and 2. Significant effects for blocking impact noise transmission are demonstrated even in those simple floor cases.

Double resilient layers work more efficiently and combinations of two different vibration mats provided further improvements of 14–18 dB for airborne sound and 16–24 dB for impact sound descriptors (floors 3–5). However, using two of the same vibration mats yielded 1–2 dB better airborne and impact-related SNQs in the case of floor 6. Floor 7 with a vibration mat on top of glass wool still performs slightly worse than floor 6.

However, the combination of a vibration mat above glass wool insulation yielded the best results for the so-called wet floor solutions, those that have a concrete screed layer on top of resilient ones. Such cases with parquet floor as the final covering (floors 8–10) demonstrate the highest insulation values, with SNQ improvements of 19–27 dB for airborne sound and 29–34 dB for impact sound.

The contribution of the parquet floor is rather arbitrary for dry floor solutions, adding 0–2 dB to the descriptors $R_w$ or $L_{n,w}$ in comparison with the same floors 1–7 without the wooden parquet layer (see Table 1). However, for the dry floors 8–10, the parquet covering has radical effects, increasing most SNQs, even up to 7 dB for $L_{n,w}$ of floor 9. Figure A1 (Appendix A) demonstrates that all cases without a parquet floor show a limited airborne sound reduction index R(f) below 160 Hz. Furthermore, Figure A2 presents larger deviations for impact sound measurement curves for the test cases without a parquet floor than those presented in Figure 3 with a parquet covering. Moreover, the performance of floor 9 is odd, deviating from the overall trends from 315 Hz and above, showing limited impact sound insulation above 800 Hz. Presumably, the thick concrete floating floor (60 mm) somehow resonates at middle and higher frequencies, which weakens the impact sound performance, especially above 500 Hz.

When the parquet floor layer is omitted, the impact sound SNQs of wet floor solutions (floor 8: $L_{n,w}\left(C_I\right)$ = 60(−1) dB, floor 9: $L_{n,w}\left(C_I\right)$ = 61(−2) dB) are similar or worse than the best dry solution cases (floor 6: $L_{n,w}\left(C_I\right)$ = 62(0) dB, floor 7: $L_{n,w}\left(C_I\right)$ = 59(1) dB). Floor 10 is excluded from this comparison as it is an enhanced insulation case. Overall, a parquet floor layer is applied in most Swedish dwellings, thus the results from Figure A1 and Figure A2 are not considered realistic.

Floor 9 has a considerably decreased impact sound performance above 500 Hz compared with floor 8 for the instances without parquet floors on top (Figure A2 in Appendix A). For those instances, the SNQ values coincide with $L_{n,w,100}$ = 59 dB, but when measured from 50 Hz, floor 9 remains slightly better with $L_{n,w,50}$ = 60 dB (while floor 8 has 61 dB). The thick concrete screed of 60 mm (floor 9) seems to induce higher transmission in high frequencies above 1 kHz for impact sound performance, while airborne sound reduction remains large, even without the parquet floor on top.

The case of floor 10 is a special configuration, similar to floor 9 with the addition of a suspended ceiling with an airgap and double plasterboard. This significantly improved both airborne and impact sound insulation and all of the SNQs (Table 2). However, there were adverse results in the frequency bands 50–80 Hz and 1.6–5 kHz for impact sound only. This case demonstrates the potential of suspended ceilings with gypsum boards below CLT slabs, which are already utilized in timber buildings [2]. Similar results with positive acoustic effects due to suspended ceiling and floating floors on CLT slabs have been reported in previous studies [2,8,34].

Floors 11 and 12 comprise exceptions in the study cases, which are the only wet solutions without a floating floor setup; they have concrete screed layers directly on CLT without a vibration mat or other resilient layers. Consequently, their insulation abilities are limited, providing average results for airborne sound similar to the simple floating floor cases 1–4. For impact sound insulation, they perform poorly, even worse than the bare CLT when there is no parquet floor, comparing the descriptors $R_w$ and $L_{n,w}$ without correction spectra (Table 2). Such a technical solution is inefficient.

Summarizing practical applications, the cases of floors 6 and 7 stand out as optimal among the dry solutions (floors 1–7), for their $R_w\left(C\right)$ and $L_{n,w}\left(C_I\right)$ values. Two plasterboard layers on top of two vibration mats seem to have sufficient performance without extended total thickness. This is important as an easy economic setup during the construction stage. Then, floor 9 stands out from wet solutions, having a 60 mm concrete cast and parquet floor on top, with the best overall performance for airborne and impact sound insulation. The presented findings can be considered for further laboratory measurements, building construction options, and industrial applications.

The linear regression comparison offers a few insights about the shifting of the trends in each case and the importance of the slope values. The impact sound model of the reference floor also demonstrated the necessity of using the full spectrum from 50 Hz to acquire a representative descriptor, which has already been suggested [12,16,17,18].

This study has a few limitations to be addressed. It utilizes full configurations, as combined materials, hence specific material behavior is not analyzed extensively. Secondly, it employs a single CLT floor component, without testing other similar samples for deviations that are reported to be significant [21,32]. Further, laboratory measurements record the insulation performance of the tested sample, while in situ measurements take into account flanking transmission and the interactions of the surrounding structure [13,34,38]. Consequently, laboratory measurements can be indicative, but not representative of a setup in a realized building.

Finally, the issue of structure classification can be mentioned, although standards utilize in situ measurements, not laboratory values, in order to account for flanking transmission in structures. The Swedish regulations have minimum requirements (acceptable Class C/BBR) of apparent (in situ) acoustic descriptors $L{’}_{nT,w,50}$ = 56 dB and $D_{nT,w,50}$ = 52 dB [44] for new buildings. The latest ISO 19488:2021 indicates acceptable thresholds of $L{’}_{nT,w,50}$ = 58 dB and $D_{nT,w,50}$ = 50 dB [45] (Class D). Only floor cases 8–10 of this study have descriptor values close to the above requirements, thus they could provide acceptable solutions assuming negligible flanking transmission in a real setup. However, such a hypothesis can be only validated with field measurements, which is not the case in this study.

5. Conclusions

This article presents standardized laboratory measurements of sound insulation for cross-laminated timber (CLT) floors with additional layers in 12 configurations. Higher thickness and total mass in a floor structure lead to enhanced sound insulation in both aspects of airborne and impact sound, though there are deviations within similar cases due to various materials.

The results indicate that CLT floor configurations with resilient layers and floating floors can provide total improvements of up to 22 dB for airborne sound and 32 dB for impact sound descriptors ($R_w$, $R_{w, 100}$, $R_{w, 50}$, $L_{n,w}$, $L_{n,w,100}$, and $L_{n,w,50}$) compared with the reference five-ply CLT component. A supplementary suspended ceiling can increase this performance even further.

A wooden parquet floor as a top covering was found to provide arbitrary improvements of 0–2 dB for airborne sound descriptors, but without a clear trend of influence in certain frequency bands. All cases tested without a parquet floor behave somewhat differently, especially for impact sound. Still, every impact sound descriptor is improved by 1–2 dB when there is a wooden parquet floor in floating floor setups. In the dry solutions with a concrete screed layer, the improvement reached up to 7 dB (floating floors only).

The findings are encouraging for solving problems of detrimental impact sound propagation in the low frequency range in lightweight buildings made of CLT components. However, the acoustic performance of the presented cases in a realized building may differ owing to flanking transmission via other components. Future studies should address comparisons between laboratory and in situ measurements for CLT floors, which is a costly operation. The mechanical properties of the floor sample and materials need to be measured and connected further to sound insulation quantities.