Background

Room Overview

Program

This room is an atrium used primarily as a multipurpose gathering space. At various hours of the day, there are students and faculty collaborating on course work, eating meals, or socializing at tables that are spread over most of the main floor. There is a cafe along one wall which operates from 8 am to 7 pm, producing sounds from espresso machines, refrigerators, and blenders, as well as people ordering and working. Noise also comes from nearby laboratories and foot traffic on the upper three floors. 

The space is also used for networking events, speaking events, and musical performances, most commonly featuring small groups of 1-3 musicians. There are several guidelines for arrangements of the space depending on the capacity of the event, ranging from 100-170 audience members. Two of these guidelines are included below; one for a theater/lecture event, and one for a banquet/reception event. 

Room Characterization

Methods

For each of these receiver locations above, an impulse response was generated by popping a balloon at the source location. The resulting sound was recorded at the receiver location by an omnidirectional microphone. Each measurement was taken with minimal people in the space and little background noise present, aside from the relatively quiet HVAC system. Receiver #1 was intended to represent an audience member close the the performer or another ensemble member, while Receiver #2 was meant to represent the listening experience of audience members near the back of the performance space.

Based on these results, we feel that the there is a lack of clarity in the space, particularly at higher frequencies. The main consequence is that it becomes relatively difficult to clearly hear the audio from a sound source, and becomes more difficult as one gets further away. Moreover, the lack of early reflections make it harder for musicians to hear each other, as well as for both musicians and speakers to project sound to the back of the space. Although the reverberation times are reasonable, the loudness and clarity seem to be good metrics to focus on improving.

Qualitative Results

During the characterization process, one of our group members played the violin at the identified source and receiver points, as well as by the seating area past Kindlevan Cafe. We observed that the room was reasonably reverberant, i.e., wet, as well as that the higher pitched notes seemed to not last quite as long as the low-frequency notes. Moreover, although the sound carried fairly well at short distances, it was tougher to hear the violin in further corners of the room.

In addition, after noticing a large number of parallel surfaces, we used sharp claps to identify the presence of a flutter echo, a phenomenon which we found to be prevalent in the space, particularly in the front entrance hallway.

Lastly, when surveying the room, a conversation was made with an AV technician setting up for a speaker event. After asking a few questions about the room’s characteristics, it was concluded that the room ‘was not very challenging to set up amplification in’ and that ‘the echoes dissipated when the room was filled with people’. This tracked with our own listening experience when observing a wet environment with noticeable reverberation and echo, as looking back at the quantitative results, the reverberation times were reasonable, but significantly lower at higher frequencies. That being said, the additional absorption provided by a crowd suggests that additional reflective surfaces would be beneficial to aid the sound in reaching the back of the audience.

Recommendations for Improvement

Acoustical Simulation

Based on our listening experience and quantitative values discussed above, we determined that the acoustics of SEC Atrium have significant room for improvement. More specifically, the clarity and loudness of the space is somewhat lacking, and the lack of reflections hinders the ability sound to travel throughout the space. Although this is the case for speaker events, it is particularly true for musical performances.  We have chosen to analyze both C50 and C80 to assess both use cases of the space; C50 will be used to analyze speech clarity, as it describes the first 50 ms of sound energy, in which the majority of speech information is present. C80 will be used to analyze musical clarity, as the human ear is slightly less sensitive to variations in clarity when it comes to musical performances. Thus, our goal is to slightly increase T15, decrease C50 and C80, and increase loudness. Moreover, we observed that the parallel walls in the front entrance hallway produced a noticeable flutter echo. As such, we considered two possible solutions to simulate in CATT Acoustic.

Clamshell: Cost Effective

• Portable acoustical orchestra shell included to add early reflections, increase loudness, and allow direct sound towards the audience

• Diffuse reflector added to the back of the brick supports in the front entrance hallway to remove the flutter echo

Overhead Reflectors: Best Performance

• Permanent glass reflectors suspended from the ceiling to allow sound to better travel throughout the listening area

• Portable acoustical orchestra shell included to add early reflections, increase loudness, and direct sound towards the audience

• Diffuse reflector added to the back of the brick supports in the front entrance hallway to remove the flutter echo

Simulation Results

As expected, the change in clarity was more significant than the change in reverberation times. The reverberation was largely unaffected because the added surface area to the atrium was small compared to the overall surface areas of the atrium (~2,500 ft^2 compared to 200,000 ft^2). By compairson, clarity is measured here by the ratio of early and late sound energy in the room.  Adding the reflectors helped produce more early reflections, thereby increasing clarity, and lowering the value of our quantitative parameter as desired. The clarity value for the clamshell leans slightly too low; the direct sound is relatively lower, and the reverberation time is slightly lower than desired as well. This could be detrimental with music, where some reverberation is desirable. However, this is an effect that would need to be corroborated with qualitative observation.

Auralization

Theoretical Context

An auralization is a binaural simulation of a source in a particular space at a given listening position. Is primary components are an impulse response, a convolution of said impulse response with anechoic audio, and a spatial rendering of the convolution to represent the source and receiver positions. 

The impulse response for an auralization may be either measured, as with our balloon pop and omnidirectional microphone, or simulated, as with our TUCT ray tracing simulations. This impulse response can then be combined with an anechoic recording of a source through a process known as convolution. Convolution represents the area of overlap under the combination of the anechoic recording and impulse response as the impulse response "slides across" the anechoic recording. It can be understood as modifying the anechoic recording based on the impulse response of the room-- frequencies that are more represented in the impulse response will then be louder in the convolved audio. It is important that anechoic audio, or audio recorded in a room with little to no reverberation, is used in this process so that the dynamics of the room in which the audio was recorded does not affect the auralization. Finally, it is important that when the convolved audio is presented, there is a spatial rendering of the space which the convolution is meant to depict. The visual depiction of the changes in a space allows the listener to contextualize what they are hearing within the larger project of designing the room, and allows them to keep track of which option they are listening to.

Auralizations are used widely in acoustical consulting as design and communication tools. In the process of designing a space, it can be useful to compare multiple simulations of design parameters in varying configurations to see how large of a difference is made between changes. Once options have been identified, it is then necessary to communicate the significance of these changes. The clients of a given project may not have a strong theoretical understanding of the impacts of these parameter changes. An auralization allows them to make informed choices based on the more intuitive way they experience sounds in the simulated space, in addition to other factors such as cost and aesthetics that they may be more familiar with. 

In the visualizations below, the source labelled A0 represents the speaker position, and sources B1, B2, B3, and B4 represent the chamber quartet.

Discussion

Comparison of Initial Analysis to Room Parameters

Although they are on a similar order of magnitude, our initial Sabine-calculated reverberation times are noticeably larger (+0.1 to +1 s, depending on frequency) than our measured reverberation times. This is almost certainly caused by inaccuracies with our estimations of room dimensions. Because we were unable to acquire dimensioned floor plans, we had to interpolate and estimate the overwhelming majority of areas/volumes. This likely led to compounded errors which led us to underestimate our reverberation times. Moreover, we opted to exclude a variety of surfaces such as tables, chairs, and counters, which could have had a larger impact than we had expected.

As for the qualitative room parameters, these lined up reasonably well with our listening experience in the space. Particularly, the abundance of flutter echoes caused by a plethora of parallel surfaces contributed to the larger reverberation times, causing the room to sound reverberant, wet, and have a noticeable echo. The calculated values of Clarity (C50/C80) were also undesirably high (~7 compared to a preferred of -4 to 0). While not a perfect representation of intelligibility of words, this could explain the regular use of amplification for vocal presenters.

Moreover, when playing violin in the space, we noticed that lower frequency notes seemed to last longer than their high frequency counterparts. This lines up with our numerical results, which demonstrate higher absorption at higher frequencies, as well as a bass ratio of approximately one.

That being said, we feel that the ability for a speaker or performer to fill the space of the large room is not well-described by these calculated parameters. As such, we would consider using the strength parameter to characterize the loudness of the space in order to better identify this room attribute.

Simulation Discussion

Qualitatively, listening to the different auralizations gave distinct and positive feedback. The inclusion of the orchestra shell greatly improved the intelligibility of the speaker, making their words more distinct and less overwhelmed with reverberation. It made the speaker sound warmer and the cellist became more present, the balance between the musicians greatly improved. The shell noticeably increased the loudness at the further position as well, regardless of the source. The resultant sound had a more intimate and less distant feel. The difference between the clamshell alone and the addition of overhead glass panels was less dramatic. We could still pick up differences that positively impacted the listening experience, but the change was far less dramatic. The musicality, the loudness, and this time the reverberation were slightly effected. The glass added slightly more reverberation to the instruments, but without the loss of intelligibility to the speaker. The reverberation sounded more like a grand hall for the quartet without sounding like an empty warehouse. Qualitatively, the clarity was slightly improved; the difference was more noticetable from the further receiver.

These results are also supported quantitatively. As desired, the reverberation times are very similar in all three cases, but the clarity values improved drastically. Looking at the data from the simulation results section above, the unmodified room had C50 values ranging from 2 to 4 dB; adding the clamshell lowered these values into the range of -11 to -19 dB. The additional overhead reflectors increased the C50 values slightly, but still resulted in an improvement from the base case, with values ranging between -4 and -14 dB. Although the clarity drop is more profound at lower frequencies, the high frequency drop relative to the unchanged model was significant. The values for C80 follow a similar pattern, although the changes are less dramatic. This demonstrates quantitatively that both speech and musical clarity should be improved by our proposed changes. Overall, these results mesh well with the qualitative listening experience. The clamshell made a dramatic difference improving the sound quality of the space, both in terms of loudness and sound clarity, while the glass reflectors in conjunction with the clamshell made a difference which was tangible, though not nearly as dramatic. The addition of early reflections in the clamshell caused a sharp decline in C50 and C80, while the glass reflectors added later reflections to modify this impact. In short, by significantly lowering the C50 and C80 values at every frequency, the data shows that our modifications successfully hit our metric of improving the clarity of the SEC Atrium.

Based on the results of our simulations and auralizations, we feel that our modifications absolutely achieve the goals that we set out to accomplish. Qualitatively, the sound sources become louder, warmer, and clearer, while the quantitative results demonstrate a decrease in clarity (a good thing in our case) with similar reverberation times. That being said, it is worth mentioning that the clamshell had a significantly bigger impact than the glass clouds. Given the fact that the installation of these glass panels would be much more expensive and time-consuming than the portable clamshell, it probably would not be worth it to install them. As such, we would recommend the first design solution, which consists of the clamshell and not the reflectors.

Regarding the feasibility of our suggested changes, the ones in the cost effective scenario are very realistic. The addition of diffuse reflectors would be inexpensive and easy to install, particularly so considering the actual surface area that they would cover is relatively small. Moreover, the acoustic orchestra shell would be reasonably priced, easy to install and move, and the device's portability would prevent it from inhibiting the SEC Atrium's use as a multipurpose space. In contrast, the addition of permanent glass panels would not be quite as feasible. The main challenges would be the large price tag, difficulty of installation, and permanence of the design solution. While beneficial for performances, the reflectors might not be ideal for the room's everyday use. 

Generally speaking, we feel that the auralization progress was a success, as it very effectively demonstrated the acoustical differences between the three different versions of the room model, including with different sound sources and from different listening locations. As such, it allowed us to identify the qualitative benefits and drawbacks of different modifications to the design. However, this process was not without its limitations.

Firstly, the auralizations were listened to using headphones, which cannot perfectly simulate spatial audio. As such, we had to deal with problems of head tracking and in-head localization. Head tracking refers to the fact that if the user turns their head, it will not change the listening experience the way that it would if they were listening in person. In addition, in-head localization describes the discrepancy between an in-person listening experience and a headphones-based listening experience, where the sound will appear to the headphones user to be coming from inside their own head, as opposed to in front of or around them. Given the resources available to us, these issues were fairly inevitable.

Moreover, our CATT Acoustic simulations utilized cone tracing and not wave modeling. This was done due to the fact that wave modeling takes significantly longer to run while only providing marginally better results. With a room as large and complex as the SEC Atrium, we felt that this trade-off was reasonable. Furthermore, our model included some geometric simplifications, such as the omission of tables and chairs, which were instead accounted for in our scattering coefficients. Lastly, our absorption and scattering coefficients are not perfect, as many assumptions had to be made about the materials in the room. We believe the impacts of these choices to be negligible, and that our representation of the space is still useful, given the closeness of our measured and simulated parameters.

If we had more time to improve our process, we would have taken the time to allow CAT5T to run more computationally-intensive algorithms, such as a wave-based method. This would likely yield more accurate results, as well as better visualize the impacts of our efforts to mitigate the flutter echo. In addition, we would have performed more a detailed analysis of the materials in the room, instead of making educated guesses regarding the use of materials and their absorption and scattering coefficients.