Concentration of reflected sound in a room treated with cube corner retroreflectors^a)

Acoustic retroreflection is the concept that sound could be returned to the incident direction, or the source, by a reflective surface or surfaces, regardless of the source position (within practical limits). Retroreflection is not necessarily remarkable in a room—arguably any simple rectangular room is retroreflective in a trivial sense because every image-source reflection is returned to the source—but the presence of numerous retroreflectors can lead to reflected sound concentrating onto the source. Such retroreflectors should be small enough to be numerous, but sufficiently large to not act as mere scatterers. Often used in optics and radar (e.g., Nilsen and Lu, 2004; Molebny , 2017) and sometimes in other branches of acoustics (e.g., Leysen , 2015; Ploix , 2020), arrays of cube corner retroreflectors (CCRs) can be scaled for sound wavelengths relevant to room acoustics. Through a full-scale physical construction and measurement experiment, complemented by wave-based simulations, this paper demonstrates how the acoustics of a non-trivially retroreflective room differ from usual expectations.

Acoustically retroreflective treatment is rarely used intentionally in buildings but can be found incidentally. Right-angle corners are common in architecture, and concentrations of sufficiently large concave trihedral corners are sometimes found on or in buildings. Certain building facades have been observed to be retroreflective (Crawford, 1991; Cabrera , 2018; Cabrera , 2020). Some Indian stepwells, within which many concave trihedral corners are simultaneously visible, have the potential to exhibit retroreflection (Cabrera , 2022). Retroreflection has been used intentionally in auditoria to support musicians on stage, e.g., shelves have been used around an orchestra shell (Toyota , 2021), and experimental on-stage CCRs have been assessed as beneficial (Tuominen , 2013). An extreme case of retroreflection for a musician is presented by Rapp (2022) who built a dome comprising 40 CCRs as an unusual semi-outdoor practice space. The present project may be the first involving intentional design of a predominantly retroreflective room.

The effect of retroreflective arrays in architecture is to concentrate reflected sound at the source position. This is usually not through coherent focusing because generally the retroreflectors are not at a constant distance from the source, and individual retroreflectors are assumed to be non-focusing. Instead, the sound concentration is from an energetic cluster of arrivals seen in the impulse response. In the prior studies of retroreflection in architecture cited, sound concentration from array retroreflection was mostly observed at high frequencies (2 kHz octave band and higher), with scattering at lower frequencies.

For retroreflector arrays comprised of non-focusing reflectors, the sound concentration effect comes from a balance of numerosity and reflector size. The finite size of individual reflectors introduces diffraction loss, leading to a pronounced decrease in effectiveness below a distance-dependent cut-off frequency (Cabrera , 2020). However, a single large retroreflector is unlikely to be of interest because, despite sending sound back to the source, it does not concentrate sound. Assuming incoherent summation, reflected sound concentration at the source should increase by 3 dB per doubling of the number of reflectors, notwithstanding diffraction loss. Realistically, some summation may be constructive rather than merely incoherent.

Recent work has introduced focusing acoustic retroreflector designs (Cabrera , 2023; Lu , 2023) in which retroreflectors are combined with geometric focusing or wave-based surface optimization methods to maximise sound concentration to a source arbitrarily located within a target zone. For CCRs, this involves warping the surfaces. Because focusing retroreflectors rely on geometric sound concentration, relying less on numerosity, they should be larger than non-focusing counterparts. However, the present study does not use focusing or surface optimization, because it predates the work on focusing retroreflectors, and because the reflectors in the present study are too small to achieve a worthwhile focusing benefit for speech (Cabrera , 2023).

For the room dimensions of the present study, $ΔL$ has a value of about 2 dB at 125 Hz, 1.1 dB at 250 Hz, 0.6 dB at 500 Hz, and is negligible at higher frequencies.

These calculations involve simplifications, which may limit application. At low frequencies, indicated by the Schroeder frequency (Schroeder, 1996), divergence from these predictions is expected. For the rooms considered in this paper, the Schroeder frequency is within the upper half of the 125 Hz octave band.

It may be hypothesized that a room with substantial acoustic retroreflection would deviate from diffuse reverberant field theory—rather than having homogenous energy density in the reflected sound field, reflected sound would be concentrated around the source.

Critical distance, $r c$⁠, is the distance at which the direct sound energy is equal to reverberant and could be calculated theoretically by equating Eq. (1) or Eq. (4) to $−10 log( r c 2)$ and rearranging. In a room with a diffuse reverberant field, the $r c$ has a broader meaning related to spectral and temporal diffusivity (Schroeder, 1996). However, if the reverberant energy is inhomogeneous—and instead concentrated towards the source—then its intercept with direct sound would approach the source. Therefore, it could be hypothesized that a predominantly retroreflective room would have its empirically determined critical distance foreshortened compared to the theoretical value.

Room acoustics is generally for people—and retroreflection pertains especially to the sound of one's own voice because reflected speech sound might concentrate onto a talker. However, the mouth and ears are neither collocated nor omnidirectional, and the head and body are scatterers in the sound field. Voice support (ST_V) takes these issues into account, being calculated from the reflected-to-direct energy of oral-binaural room impulse responses using a head and torso simulator (Pelegrín-García , 2012).

It may be hypothesized that a room with substantial acoustic retroreflection would yield greater voice support than predicted by Eq. (6). Voice support has been used in studies of voice regulation and speaking comfort. Higher levels of voice support from reverberation are associated with reduced vocal output and greater speaking comfort (Pelegrín-García and Brunskog, 2012), but early reflections may be more helpful than reverberation (Bottalico , 2016). Hence, a potential use for retroreflective treatment might be to maintain good voice support with relatively short reverberation time.

An approximately rectangular room 5.35 × 3.65 × 2.83 m (55.26 m³) was furnished and treated to be predominantly retroreflective using 156 explicit square trihedral CCRs with 0.35 m edge lengths (Table I). Each has the equivalent flat area (boresight shadow) of 0.21 m², or an actual surface area of 0.37 m². The 89 furniture and wall CCRs are from hard smooth materials with relatively high surface density (painted sheet steel, smooth painted 18 mm plywood). The 67 ceiling reflectors are made from plastic membranes each stretched in a steel wire frame. The floor is painted concrete and was not treated to be retroreflective. The room volume lost due to the treatment is 5.7 m³, yielding an adjusted room volume of 49.56 m³.

The room is an office, so the treatment was also designed from a practical useability standpoint (e.g., the room has a door and a floor; the sideboard was designed to store larger items, leaving some of that wall flat). The other furniture in the room, such as a small desk and chairs, is made from steel wire mesh to be reasonably acoustically transparent (but was removed for the detailed measurements reported in this paper). The room has no windows. As a legacy of former partitioning, a water pipe runs vertically near the room's center. A concrete beam spans the width of the room ceiling. Figure 1 provides views of the room.

To prioritize retroreflection over reverberation, 35 m² of 100 mm polyester wool sound absorptive material was installed against the end walls and ceiling. This material is largely behind the end wall cupboards and above the ceiling reflector array treatment, but some remains visible. The measured mid-frequency reverberation time of the unoccupied treated room is 0.26 s, with an increase to almost 0.4 s in the high frequency range (Fig. 2). The measurements were made with 12 distinct omnidirectional source-receiver position combinations, following ISO 3382-2 (ISO, 2008) to the extent possible within the room confines. The ceiling membrane panels likely contributed to the low-mid frequency absorption, while being more reflective in the high frequency range. The 125 Hz band values span a large range and are likely affected by acoustic coupling to an adjacent reverberant corridor via an air vent in the door, along with low modal density in that band due to the small room size.

Two miniature dodecahedral loudspeakers (0.1 m diameter, Dr Three type 3D-032, Tokyo, Japan) were used as the sources so that their respective source center could be near a receiver and to avoid unnecessary scattering by the sources. Impulse responses (IRs) were measured from each source to a linear array of 101 omnidirectional microphones (RØDE Lavalier, Sydney, NSW, Australia), which was successively displaced. The microphones were attached to a 10 mm square cross section steel bar, at 20 mm intervals. Displacement was achieved electromechanically using a stepper motor driving a ball screw, which controlled the linear displacement of the microphone array [Figs. 1(c) and 1(d)] in 10 mm increments, controlled by the measurement computer. Once the microphone bar had completed a full traverse, the measurement was repeated with the bar displaced laterally by 10 mm. This yielded a receiver plane of 202 × 188 = 37 976 positions in a 10 mm square grid (2.02 m × 1.88 m at 1.2 m height, and 202 × 187 = 37 774 at 1.5 m). Due to physical constraints, the sources were just above the receiver plane (rather than on the plane), with their centers offset by 0.06 m. IRs were also recorded from two additional microphones (Hottinger Brüel & Kjaer (HBK) type 4190, HKB, Naerum, Denmark) at fixed arbitrary positions in the room to check the consistency of the repeated test signal in the room. The test signal was a 5 s exponential sweep spanning 80–12 500 Hz. The audio interfaces, computer, and personnel were outside of the room, with the door closed. Figure 3 includes a diagram of the room and transducer locations.

A practical end truncation point of 150 ms was sufficient to accumulate all significant energy. The reference 1 m emitted energy was calculated by averaging squared direct sound pressure from receivers at 0.25 ± 0.002 m from the source, with −12 dB compensation for geometric dispersion.

The reference microphone recordings were also processed to yield IRs. Maximum absolute deviation from median energy was 0.13 dB, with a 0.04 dB interquartile range.

Numerical acoustic simulation was undertaken using the FDTD method. Three room models were simulated. The first was an attempt to replicate the physical room. The second simulation, referred to as “maximum CCR,” was of a room with a more extreme CCR array treatment, covering almost all the non-floor surface (Fig. 3). The third simulation was a simple rectangular room.

The acoustic simulation was done in PFFDTD (Hamilton, 2021a) on a desktop computer with 128 Gb system RAM and 2 x NVIDIA RTX A6000 GPUs (NVIDIA, Santa Clara, CA) with 48 Gb VRAM each. A 13-point face centered cubic scheme (Hamilton and Webb, 2013) was used to approximate the Laplacian in the second-order linear scalar wave equation and simulations were executed using double precision floating-point numbers. The PFFDTD solver implements boundary fitting (Bilbao , 2016) as well as frequency dependent boundary impedances (Hamilton , 2016). Fratoni (2022) discuss and demonstrate practical aspects of this type of simulation compared with a geometric acoustics approach. The time and spatial steps were approximately 10⁻⁵ s and 3.6 mm, respectively. A single simulation to yield a 400 ms impulse response took 20 min and 41 s using the described hardware.

The simulation of the real room had a reverberation time spectrum close to that measured (Fig. 2). However, because the other two simulations differed in form, matching the simulations' absorption was more useful than matching their reverberation time. The 125 Hz octave band was not matched because of the high uncertainty of its reverberation time measurements. The maximum CCR room had the same basic volume as the real room, but only 1.6 m³ was occupied by the CCR treatment (compared with 5.7 m³, mainly because the maximum CCR treatment did not incorporate cupboards). Therefore, a slightly longer reverberation time was needed to match the real room's absorption. The simple rectangular room was dimensioned 5.16 × 3.52 × 2.73 m, to be the same volume as the original room minus its furniture. A simulation duration of 400 ms was used to evaluate reverberation times (T₂₀) for tuning absorption in each scenario. Absorption coefficients were adjusted iteratively to match the measured equivalent absorption area according to Eq. (2). The final absorption coefficients are shown in Fig. 4.

Oral-binaural room impulse responses (OBRIRs) were measured using a head and torso simulator (HATS, HBK type 4128 °C, HBK, Naerum, Denmark). These measurements were made with the HATS's ears at two heights: 1.2 and 1.5 m. They were made at three positions in the room (Fig. 3). OBRIRs were measured at four orientations at Positions 1 and 2 and three orientations at Position 3, yielding 22 distinct measurements. The purpose of such measurements is to assess the extent to which a person's own voice is returned to their ears by the room environment. As described by Pelegrín-García (2012), this can be represented as room gain (G_RG—which is the increase in energy received at the ears compared to the free field, in dB) or voice support (ST_V—which is the ratio of room-reflected energy to the direct oral-binaural energy, in dB). Evaluated in octave bands, overall speech-weighted values were derived using the 125 Hz to 4 kHz bands following Pelegrín-García (2012).

The overall distribution of reflected energy levels in the three simulation scenarios and measurements (all source and receiver positions) is summarized in Fig. 5. Note that the total number of values for the measurements (151 500) is about half that for the simulations (316 092), and that the simulations include greater distances. Figure 6 shows spatial maps of the reflected energy levels over the receiver plane, with the area not covered by the measurement in black. Figure 7 gives an example of the spatial decay of reflected energy as a function of distance from the source (1.5 m height, source at Position 1). The direct sound is excluded from the data displayed in all these figures (but is shown for reference as a geometric calculation in Fig. 7).

The simulation of the real room yielded energy level distributions that are broadly similar to the measurements. Apart from the 125 Hz band, all spatially averaged energy means for the simulation are within 1 dB of the measurement, and medians are within 1 dB (Fig. 5). Quartiles, maxima, and minima also match within 1 dB at and above the 1 kHz band, and for A-speech weighting. These matching bands are of most interest for this study because the effect of retroreflection is mostly in them. Similarities between the real room and its simulation are also apparent in Fig. 7, including the spatial decay rate values and most of the critical distances in these bands.

Nonetheless, differences between the real room and its simulation occur in the energy returned to the source—apparent in the large circles in Fig. 5, as well as in the encircled zones of Fig. 6. In the real room, the source is a small dodecahedral loudspeaker, with the closest receiver 6 cm from the source center. The physical loudspeaker acts as a scatterer of reflected sound to an extent. On the other hand, in the simulation, the source and receiver are collocated, and the source is not a scatterer. These differences in scenario are likely to have influenced the differences in reflected energy at the source location. Even so, the measurement and simulation are in agreement that sound returned to the source is towards the maximum value of the receiver plane, especially in higher frequency bands.

Figure 5 shows the measured and simulated energy level distributions along with theoretical predictions. The Sabine theory predictions are about 1 dB greater than the Eyring predictions. In the case of the simple rectangular room simulation, the energetic means are about the same as the theoretical predictions (but slightly less at 125 Hz and 250 Hz, and slightly greater at 1 kHz bands).

Spatial variation in reflected energy is seen in low frequency bands, and is likely attributable to a sparse ensemble of room modes within those bands (evident similarly in the measurement and other simulations of Fig. 6). Spatial variation of reflected energy within the simple rectangular room is minimal for higher frequency octave bands, yielding small or zero spatial decay rates (Fig. 7). Hence the simple rectangular room reasonably conforms to established theory, showing a homogenous reverberant sound field that is controlled by the room's absorption.

The three CCR cases (the measurement and two simulations) show sound concentration on the source especially in the 2–8 kHz octave bands, also noticeable at 1 kHz and weakly at 500 Hz. Sound concentration is also evident for the A-speech-weighted results. Concentration is seen as an extension of the upper distribution tail in Fig. 5 in these bands, with most upper tail datapoints at or near the source. Reflected sound concentration on the source is also clear in the spatial maps of Fig. 6, and the spatial decays in Fig. 7.

For the measurement, the greatest reflected energy in the 4 kHz band at a source is 7 dB greater than the corresponding octave band maximum of the rectangular room simulation, and 11 dB greater than the mean expectation from diffuse field theory. In the most extreme cases for the maximum CCR simulation, the 4 and 8 kHz octave reflected energy levels at the source are 10 dB greater than the octave band maxima for the simple rectangular room, and 13 dB greater than the mean expectation from the diffuse field theory. Spatial decay rates of reflected energy for these bands in the maximum CCR simulation are greatest, at 2.2 dB per distance doubling. In these bands, the maximum CCR spatial maps (Fig. 6) show parallel ridges extending from the source position along the length and width of the receiver array. These ridges coincide with reflector array periodicity, and so are due to increased coherent summation. This is less apparent in the real room measurement or simulation, which have a less regular arrangement of CCR arrays.

The empirical critical distances that were found from the intercept between fitted spatial decay of reflected sound and the geometric direct sound are summarized in Fig. 8, along with the theoretical critical distances of each room. For the simple rectangular room, the results are essentially the same as theory in the 1–8 kHz octave bands. At low frequencies, the results are greater than theory, but it is obvious (in Fig. 7) that the low frequency results are not meaningful because of the wide range of reflected energy values influenced by room mode patterns. The three retroreflective room cases have empirical critical distances that are systematically shorter than the diffuse field theory in the high frequency bands. Foreshortening is greatest for the maximum CCR room simulation (critical distance is half that of theory in the 8 kHz band), but foreshortening is still evident in the treated room measurement and simulation.

The spatial decay rate of reflected energy grows linearly with frequency in the three retroreflective cases with a slope of 0.2–0.3 dB/octave, based on robust fitting (Fig. 9). The highest values are around 2 dB, considerably less than the spatial decay rate of direct sound (6 dB). When the direct sound is included, the three CCR rooms have greater spatial decay rates (about 3 dB) than the simple rectangular room (about 2 dB) in the high frequency octave bands.

The overall mean value of speech-weighted ST_V was −4.1 dB, with individual measurement values spanning the range from −5.6 to –2.6 dB (Fig. 10). Values tended to be greater at Positions 1 and 3 (closer to the CCR arrays at the ends of the room than Position 2). Slightly greater mean and maximum values were measured at 1.2 m than at 1.5 m height. The theoretical speech-weighted ST_V for the room is −9.0 dB, 4.9 dB less than the mean measured value. Hence the result implies that the reflected sound field from the HATS is concentrated onto its head, consistent with the intention of the CCR array treatment.

The measured ST_V values are mostly from the early-reflected sound. The mean speech-weighted voice support calculated from the first 50 ms of the impulse responses is −4.4 dB (0.3 dB less than the full impulse responses).

The octave band voice support values mostly increase with frequency (Fig. 11). Note that the 250 Hz band includes the frequency of the second vertical axial mode, which has a pressure antinode near 1.5 m—this is the likely cause of greater values at 1.5 m (also seen at 250 Hz in the omnidirectional measurement results, Fig. 6). Octave-band voice support is expected to increase over frequency in almost any room, so the theoretical diffuse field octave-band ST_V based on Eq. (6) is shown for comparison. Excess voice support (relative to theory) rises with frequency at 1.3 dB/oct, including in the octave bands in which retroreflection is evident. The theoretical model does not include the 8 kHz band.

Clearly the spatial distribution of reflected sound in the real room and simulations treated with CCRs differs from usual expectations in the high frequency bands. To achieve a theoretical speech-weighted voice support [based on Eq. (6)] equal to the measured value of −4.1 dB, the room's reverberation time would need to be 2.6 times its measured value, i.e., a mid-frequency reverberation time of 0.7 s (and about 1 s in the high frequency bands). For a room of its size, the result would be unusually reverberant. To achieve a theoretical diffuse field level equal to the level received at the source position in the physical room at high frequencies, the reverberation time adjustment would need to be more extreme, the 4 kHz band being the most extreme case. Realistically, in the 4 kHz band, the energy mean could be 4 dB below the maximum of a diffuse field (based on the distribution in the specular room simulation), in which case the 4 kHz band reverberation time of the real room would need to increase by a factor of 3.5 (to 1.3 s) based on Eq. (1). This would be akin to bathroom acoustic conditions, but with an unusually high frequency reverberation emphasis.

This room is not only an acoustic experiment—it is also an office. No subjective or behavioral experiments have been conducted in the room at the time of writing. Nevertheless, the authors have experienced the room in the period since completion in late 2020, and the following subjective comments come from that. The room does not seem reverberant and would be described as “dry.” Support of one's own voice is noticeably strong. The sound of one's own voice is “crisp,” with high frequency consonant phonemes noticeably emphasized. This high frequency emphasis is not disturbing and contributes to a feeling of acoustic support. Experience in the room is consistent with the notion that acoustic retroreflection could be used to enhance room acoustical quality for speaking.

The A-speech-weighted results show some sound concentration at the source, but much less than the high frequency octave bands. This reflects the fact that omnidirectional speech energy peaks in the 500 Hz band, A-speech weighting's power spectral centroid being 818 Hz. This raises the question of whether using speech energy is appropriate for a single number summary when the sound concentration phenomenon is mainly at higher frequencies. Intelligibility-oriented weightings emphasize higher frequencies, especially the 2 kHz octave band. While autophonic intelligibility is not an established concept, there does seem to be some benefit in having one's own speech clear and crisp (Appel and Beerends, 2002; Bottalico , 2016). Rapp (2021) found high-frequency voice support (boosted at frequencies >1 kHz only) was advantageous for speech quality and voice regulation, and found a relative disadvantage in low-frequency (<1 kHz) voice support.

The choice of 350 mm edge-length CCRs for the room treatment was based on experience with early prototypes, results of prior studies, and the pragmatics of space in the room and available furniture components. The maximum CCR simulation shows that greater sound concentration at the source could be achieved through more extensive surface treatment. There are many geometric variant possibilities, such as: different sized CCRs, pentagonal or triangular-faced CCRs, non-CCR retroreflectors, and focusing retroreflectors. For example, the effective width or diameter of the reflectors is about 0.5 m, but Cabrera (2023) found that greater A-speech-weighted sound concentration is achieved at a 2 m distance by 0.6 m width arrayed ideal square retroreflectors. Furthermore, focusing retroreflectors can yield stronger retroreflection than non-focusing ones, but should be about double the linear size of non-focusing equivalents for their advantage to be best realized (Cabrera , 2023). Therefore, it is likely that the retroreflected energy levels found in this study could be exceeded by alternative retroreflective treatment in a room of this size.

The CCR treatments used in this paper are unlikely to be applicable to general rooms, considering the large surface modulations and unusual angles. The protruding corners create a potential hazard for people when used on a wall, also creating a visually busy surface lacking an overall flat face. Protruding corners can be avoided by using tessellated CCRs with pentagonal faces instead of square (Kim and Lee, 2007), which may have more general architectural uses. An interesting alternative approach is to use focusing retroreflectors by warping a rectangular room's basic surfaces at corners and edges. This introduces the potential for unobtrusive treatment that yields retroreflective sound concentration at least in zones within a room. Hence, although the treatment examined in this study is impractical practical alternatives can be developed.

This paper is about a small room treated to be predominantly retroreflective in the high frequency range. Being small, the room does not have pressing acoustic problems, but retroreflective treatment could play a useful role in larger and more challenging rooms. Concentrating reflected sound onto a talker means that strong voice support can be achieved with short reverberation time. Reduced reverberation time can have positives, such as reduced background noise build-up, and greater speech clarity. It can also cause a greater spatial decay rate, leading to lower speech sound pressure levels for listeners. Perhaps this could be beneficial in environments wherein intelligible speech or speech build-up is unwanted—such as open-plan offices or eating establishments. Spatial decay rates observed in this study are much less than recommended for open-plan offices (ISO 3382-3), but that is partly because the room is small. In such spaces, spatial decay could be further enhanced by conventional means (barriers and absorption), and there may be voice regulation benefits from acoustic support. In medium-sized rooms for speech communication increased spatial decay of speech could be unwanted, and so may need to be mitigated by other reflections. Nevertheless, in classrooms, reverberation time recommendations range from 0.4 s or less to emphasize intelligibility (Mealings, 2016), to 0.7 to emphasize teacher voice support (Pelegrín-García , 2014). Achieving this voice support with a short reverberation time may benefit both talkers and listeners.

Surfaces in room acoustics are conventionally characterized in terms of specular reflection, scattering, and absorption, and widely used theories of reflected sound distribution are built on these. This paper demonstrates that a form of acoustic treatment which is predominantly retroreflective in the high frequency range leads to a reflected sound distribution that deviates substantially from usual expectations. The reflected sound energy concentrates at the source. For the room and treatment parameters of this study, this led to 5 dB excess voice support relative to diffuse reverberant field expectations. For a collocated (or minimally separated) source-receiver, the excess sound concentration was mostly in the 2–8 kHz bands, but still present in the 1 kHz band and weakly evident in the 500 Hz band. In the most extreme cases, the measured reflected sound returned to the source was 7 dB greater than the maximum of an equivalent simple rectangular room, or 10 dB for the maximally treated simulation. Unlike focusing rooms, there is no sweet spot because the reflected sound concentration follows the source. In room acoustics, sound concentration is usually an acoustician's nemesis, but retroreflective sound concentration may be an exception because there may be benefits in supporting talkers' sound, for which further research is needed.

Jonothan Holmes is supported by an Australian Government Department of Education, Skills and Employment Research Training Program scholarship. Microphones used for high spatial resolution measurement in this project were donated by RØDE (Sydney, NSW, Australia). The authors thank Edward Iverach and Rodney Watt for assistance in constructing the retroreflective treatment, and the Sydney School of Architecture, Design and Planning for making the room available. A preliminary investigation of the room and some variants thereof was done by Oliver Hutchison as part of the degree Master of Architectural Science (Audio and Acoustics).

The authors declare that they have no conflicts of interest to disclose.

The data that support the findings of this study are openly available in the Open Science Foundation repository “Room treated with cube-corner retroreflectors, Wilkinson Building 2020” at http://doi.org/10.17605/OSF.IO/N9FGE.