Background

Room Overview

This room is called the Fisher Performance Room and is located in the Granoff Music Center. Its floor is assumed to be coated in linoleum with a surface area of around 2030 ft^2. The walls (around 4100 ft^2 in total) are drywall and are around 22 feet tall. There are also curtains in front of some of the walls on the sides and back of the room. These can be retracted to expose wide slats on the wall, giving it a more varied surface, or they can be rolled out to cover these walls. There is also a large window in a back corner of the room, with another curtain that can be positioned in front of it or retracted at the sides.

On one side of the room, near the sets of hardwood doors on either side of the front wall, there is a lowered, sloped ceiling with space used for storage underneath. This alcove is about 8 feet tall at its highest point and 6.5 feet deep. Above the main wall where this alcove lies, there are 7 plaster wedges in alternating directions made to provide a more varied surface for reflections and for decoration. The ceiling is flat with 8 absorptive panels hanging below an air gap of one or two feet. These panels take up a significant portion of the ceiling space as they are each around 10x10 feet. In the room, we placed 50 chairs arranged how they might be for a performance. There are 5 rows of 10 chairs each, with a 3-ft aisle running down the center. They are somewhat upholstered but are not very thick.

Engineering Drawing of Fisher Hall with Appropriate Dimensions

Hardwood Doors Alcove Under Wedges

Plaster Walls Plaster Wedges Dampening Curtains

Linoleum Floor Tiles Room Set Up with Array of Upholstered Chairs

Program

Similarly to Distler Performance Hall, Fisher is used as a rehearsal and performance space for various ensembles located on campus. The ensembles which typically use Fisher tend to be louder than the ensembles which primarily rehearse upstairs in Distler. While Distler is used by Tufts Symphony Orchestra, Tufts Chamber Orchestra, and several choirs, Fisher is the typical rehearsal space of groups such as Tufts Wind Ensemble, Tufts Pep Band, Tufts Jazz Orchestra, and several other faculty-run groups. Fisher also serves as a rehearsal room for clubs such as the Tufts Freshman 15 (a student-run jazz ensemble) and Torn Ticket II, whose pit orchestra uses the space. The various ensembles which use Fischer range in size from approximately 12 to over 60 members. Interestingly, one of the largest ensembles to use Fischer as a rehearsal space is the Tufts Youth Philharmonic Orchestra, though this group typically hosts its performances in Distler.

Room Characterization

Methods

To mimic a concert setting, the room was arranged to have 50 chairs facing the front of the room. The sound sources were played in the front of the room where performers usually are, and the microphone was situated in two different spaces. The first was in the middle of where the audience is seated, and the second is in the far left corner, close to the piano in the back corner. We chose these positions to see how the sound would vary based on the distance from the source, as well as the surrounding material (curtains, walls, audience members, chairs, etc). At both of these positions, we used an omnidirectional microphone to record 3 sound sources: a balloon popping, a brief song on the piano, and a person talking. During data collection, we also listened for any audible differences in the sounds of the sources, though it was difficult to distinguish smaller differences, especially for the balloon popping due to its shorter nature.

To take this a step further, we also decided to test the effect of the configuration of the curtains in the room. We took one set of measurements at both locations with all the room curtains drawn closed, then another set with all the curtains opened (bare walls as exposed as possible). There were two sets of curtains in the room: the ground-level curtains that could be pulled by hand, and the higher window curtains that had to be controlled by a panel in the room alcove. We rearranged the ground-level curtains manually to expose or hide the patterned wall behind.

For Sabine calculations, we assumed the room looks like it does in the CAD model and that there are no other objects in the room. In reality, there are racks full of chairs and stands as well as several percussion instruments in the small inlet under the wall with the wedges. There are also two pianos in the room. These objects are omitted for simplicity. We are also making assumptions about the sizes and distances of the wall shapes and objects that are too far from the ground to accurately measure with our equipment. Finally, these material constants are based on assumptions on what material each object and component is actually made of in the room, which we could not know without blueprints. 

For room auralization, our process involved Google Sketchup, CATT-Acoustic, and CATT's TUCT package. The room model was translated from OnShape to Google Sketchup, and materials were assigned to every interior surface. Changes were made to the chair models, considering the groups of them as two large square prism masses in the middle of the room instead of having each chair be an individual object. This allowed for quicker simulations while producing satisfactory results. Sources and receivers were placed according to the diagrams shown above, and CATT was used to run simulations using simulated sound sources (male speaker and jazz band). Using TUCT, WAV files were created using the simulation, and they were convolved to yield simulated listening experiences.

Quantitative Results

Below are the charts and tables for the EDT and T20 of the sound impulse calculated across various frequencies. The lower frequencies (125 Hz and 250 Hz) tend to exhibit longer EDT and T20 values, and while this is largely in part due to the room being less absorbent of said frequencies, it could also be due to a low signal-to-noise ratio.

Overall, the lower frequencies exhibited higher T20 values. We suspect this may be due to the microphone experiencing difficulty in distinguishing the low-frequency impulse of the balloon from the ambient background noise of the room, as the T20 values actually seemed somewhat dependent on the length of the entire audio clip. As such, we did our best to crop the audio to encompass just the sound impulse. 

With regards to microphone positioning, the microphone detected mostly longer reverberation times when placed on the side of the room. In contrast, placing the microphone in the center of the audience resulted in shorter reverberation times. 

In general, outside of the outlier low-frequency values, reverberation seems to be the longest at around 500 Hz, which roughly matches the average concert frequency of around 440 Hz. This means that music will reverberate for longer than undesirable sounds such as walking (1-4 Hz) or talking (~250 Hz).

Qualitative Results

Below are the Sabine reverberation time calculations for Fisher Hallin both curtained and exposed configuration, with the appropriate T20 values below them.

The reverberation times for all octave bands are noticeably different with the curtains retracted and with them covering large sections of the walls. This makes sense because the curtains are some of the largest absorptive materials in the room. It is also noticeable that the reverberation time generally decreases for higher frequencies.

For the critical listening exercise, one member of the group played piano from the location of the impulsive source while the other group members listened from the two different measurement locations in the room (at the center of the audience and near the back corner of the room). The biggest difference that was noticed was when the configuration of the room was changed using the curtains hanging in front of the reflective walls and windows. When the curtains were drawn over the walls, the thick and heavy fabric with waves was extremely effective in absorbing a lot of the sound and dampened most of the reverberation that would have been heard otherwise. As a result, the sounds were very clear, which aligns with the quantitative data as the reverberation time was much shorter when the curtains were covering the walls and windows. When the curtains were dragged back and the reflective surfaces were exposed, the sounds of the piano became much less clear as each instance of sound remained in the air and reverberated for much longer, as can also be seen with the measured RT values for this room configuration. However, there was a small change that could be noticed in the listeners’ experienced envelopment. While the curtains allowed for a clearer sound, as there were fewer surfaces to reflect off of, the space was not as well-equipped to envelop the users and give them the sense of immersion that was felt when the walls were exposed. Envelopment is aided by a diffusively reflective surface, which the ridges and tabbed areas of the walls served well to provide when the curtains were taken away. In this orientation, the listeners felt more immersed in the listening experience.

While the change in the location of the listener (the two measurement locations) didn’t drastically change the listening experience as the room is small enough that most areas receive similar sound qualities, there were some changes in directionality to note. In the middle of the room, the listener experiences a relatively even reception of sound as there are very few objects in the immediate surroundings. Even though the chairs were surrounding them, they are lower than ear level and do very little to change where the sound is perceived to be coming from. However, when the listener is in the back corner of the room with walls to their back and right side, the perceived sound is noticeably from the front and left. There is a slight dampening effect for the sound heard from the right ear. However, when the curtains are drawn and the wall is exposed, this effect is less prominent. In this configuration, there are also certain areas of the room (different from the measurement locations) that experience more flutter echo than others. This occurred, as expected, in the areas between two parallel surfaces. For example, the location of the sound source was between the window wall and the door wall, equidistant from both, and apparently far enough apart that the listener heard the sound echoing off of them. Stepping closer to the center of the room, this effect is much less, which makes sense because the presence of the alcove means the lack of a parallel wall system for the sounds to reflect off of and create the flutter echo effect.

The sound of a voice speaking in the room had very similar differences when the curtains were open vs. drawn and when the listener was traveling around different areas of the space. A similar test was done, where the speaker positioned themselves at the location of the impulse sound and spoke continuously as other group members observed from the microphone positions as well as other locations around the room. The sound carried more but lacked clarity when the wall was exposed and echoes were present, which made it more difficult for distinct words to be heard clearly, but the reverberation time was not nearly long enough to make the words indistinguishable. When the curtains were covering the walls, the voice was dampened considerably, but it was easier to make out certain words – there was more clarity. From the speaker's perspective, yelling in the area that seemed to have induced a flutter echo was an interesting experience. You could hear your own voice repeat and echo for a short amount of time before it died out.

Discussion

Comparison Between Measurements and Listening Experience

For this comparison, we chose to use the T20 values from the microphone placed in the middle of the room, as shown in the section above. The Sabine values fall roughly within 10% of the T20 values. To get this close to the T20 values, various materials were tried and adjusted until optimization plateaued. This was tricky because Sabine values were scattered both above and below the T20 values, not just in one direction. We suspect that this may be due to less-than-ideal sound recording quality. 

The difference in the reverberation time between the closed and open curtain configurations aligns with our qualitative experience. The curtains noticeably dampened the sound traveling from the source quicker than the reflective surfaces of the walls and windows did. The clarity values did not seem to change much between the microphone positions - rather they seemed to be most affected by the curtain configurations. The clarity values when curtains are open are much lower than when the curtains are closed. This aligns with our qualitative experience, as while the source sounded louder with the curtains open, the sound was not as easily decipherable, whereas the closed curtains led to a more dampened but distinct sound.

The qualitative parameters derived from the data do not cover some of the perceptual attributes of the listening experience such as timbre, loudness, balance, spatial impression, and extraneous sounds. The focus of our data analysis was on the clarity and reverberance qualities of the sound experience by a listener. These perceptual attributes were definitely noticeable and experienced when we were going about our measurement and observation process, however. And while certain values we derived (i.e. longer reverberation time in the curtain open configuration) made sense with some of the perceptual attributes (i.e. presence of flutter echo in the curtain open configuration), these two could not always necessarily be linked and a connection drawn between the two without the in-person experience to contextualize the data.

To better correlate with timbre, we would need to detect the range of frequencies of the sound source, as well as the presence of overtones and undertones. Balance refers to the distribution of energy across the range of frequencies. To better correlate with this attribute, we would need to visualize the amplitudes across the frequencies. Spatial impression depends on the size of the space, the width of the space, envelopment, and responsiveness. While the size and width were determined in this experiment, the envelopment and responsiveness were not. In order to measure envelopment, we would have to construct a binaural impulse response from the sound source, then measure the fluctuations in the interaural time differences. These calculations were beyond the scope of our project. Extraneous sounds caused by echo, image shift, and background noise also depend on further analysis not carried out for the purposes of our experiment. Flutter echoes were strongly present in the room and were caused by the presence of distanced parallel walls, but the measurement and analytic distinction would require observation in the peaks of the resulting energy graph.

Recommendations for Improvement

One improvement that could be made to Fisher would be to install more absorptive materials for low frequencies. In both sets of data, the 125-Hz octave band yields the largest reverberation time. This is the closest band to the range of human speech. Because this functions as a practice room primarily, speech is an important factor. To make it less reverberant in low frequencies, a new material could be installed that has a high absorption constant in those lower ranges to make it easier to hear other people across the room. One way to implement this could be to change the wedges on the wall from plaster to another material with higher absorption coefficients in the low ranges.

Another interesting observation is about the shape of the room itself. There is a large alcove on one side of the room which can act as a large sound absorber because of all the absorptive equipment and the general shape, which looks like it might reflect sound back into itself somewhat and trap waves from reflecting back out into the room. This could create an uneven experience for listeners if there is a performance in one of the longer ends of the room, since there are chairs placed directly in front of it. One suggestion, which is probably the most efficient solution, would be to place another large curtain over the opening of the alcove. If made of the same material as the curtain directly opposite this space in the room, it would even out the effects so listeners on opposite sides of the room have more similar experiences.

Experimental Process

During the gathering of the reverberation time data, a balloon was popped at each arrangement of the room set-up (curtains drawn or closed, two different microphone placements). The instantaneous sound - suddenly interrupted - allowed for a clear measurement of how long it took for the resulting reverb to recede. However, each balloon was not blown up to a standard size. Because of this variance, there is a chance the measured reverberation times are not the most comparable, as the microphone is not strong enough to pick up all of the lower decibel range as the sound fades. Perhaps having a standardized balloon size or an otherwise uniform impulse sound source that provides dependably consistent acoustical properties would have resulted in more accurate data results.

There is also the fact that reverberation time varies with the frequency of a sound. The sound of a balloon popping doesn’t cover the lower end of the frequency spectrum, as can be seen with all of our 125 Hz data. The 63 Hz data wasn’t even considered in this experimentation because the resulting values would not be of any use. It is difficult to find an instantaneous sound source that covers the entire frequency spectrum, short of condensing the sine sweep sound into a few milliseconds, but it would be ideal for better characterizing the room. Having a better idea of lower frequencies would allow users of the room to determine if it is a good space for their purposes.

Something interesting and unexpected that occurred during our recording process was the behavior of the balloon once it was popped. In Distler, the floors are carpeted, so we observed just the sound of the balloon popping before the carpet absorbed the sounds of the pieces falling to the ground. However, in Fisher, the floor tiles are made of linoleum, and the resulting sound of the exploded wet pieces of the balloon made very distinct and noticeable slapping sounds as they hit the floor. These sounds were quieter than the initial sound of the balloon popping, but were still distinct and picked up by the microphone. Because we were recording with the curtains open, the flutter echo lasted almost as long as it took for the pieces to hit the floor. Luckily this unexpected phenomenon did not interfere with the recording and we didn’t have to do a retake, but it was very interesting to note. For further experimentations and future measurements, it would be best to lay down a sound absorptive sheet below the balloon to prevent any possible interference.

Though the configuration of the room for the IR measurements was done as close as possible to when Fisher hosts performances, there was still one major difference between the simulation and reality: audience presence. It was not feasible to obtain a mock audience of 50 for the experimentation purposes, so it was hoped that the upholstered chairs would suffice. To get the measured and calculated equations as close to reality as possible, however, having people sit on the chairs would be ideal. During performances, there is also usually a wooden platform on which the performers sit with their instruments, which would also provide a different surface for sound to reflect off of. Particularly for larger instruments that sit on the platform themselves, this would change the quality of their sound and could be interesting to take into account, along with the effects of having an instrumental ensemble onstage that emits but also absorbs sounds themselves.

The microphone we used for the experiment was omnidirectional, which served extremely well for our intents and purposes. However, we didn’t have the appendable tri-pod that comes with it, which meant that we had to resort to whatever we had on hand to stabilize the microphone. The end result was a backpack placed on a chair with the microphone mounted on top. This was a secure, but not exactly level solution – the microphone was not perfectly perpendicular to the ground and was prone to shifting around as we made adjustments to the room set-up between each recording. As well as omni-directional microphones work, they do require some effort in ensuring the orientation is optimized for the space. An improvement for future measurements could be using a more fixed perch for the microphone, or obtaining the accompanying tripod.

The space in Fisher is unique. Other than the main rectangular space that composes most of the room, there is a small alcove on the door side of the room where instruments, chairs, music stands and other miscellaneous objects are stored. The geometry of this space presents some interesting acoustical characteristics, but the objects inside do a lot to change the sound heard as well. These objects were not taken into account in the Sabine equations, but future measurements could see them included. There are also two pianos in the room that may also play a part in the way the sound is heard by the audience, but none of these extraneous objects are present during performances, so they were left out of the equation.

As for the analysis and quantitative data comparison, measurements of the room that were made were as accurate as could be within reason. The horizontal dimensions were all taken as close to reality as possible, but the height of the space as well as the windows near the ceiling were not accessible and thus had to be estimated using surrounding objects and architecture. As we couldn’t obtain the blueprints of Fisher Performance Room and the materials of each surface can only be guessed at best, there is definitely room for improvement in the Sabine calculations which we compared with our measured data.

Source

Sound absorption values for room materials: http://heyizhou.net/notes/absorption-coefficients 

Acoustical Simulation

Model Summary

Simulation Results

We've recorded changes in T20, T30, and C50 between the different versions of the room. T20 and T30 were both noted to see if there were any differences between them, even slightly. As we can see in these charts representing reverberation time over different octave bands, the T20 and T30 values are very consistent between rooms for bands between 500Hz and 4000Hz. They differ more at lower frequencies. This is probably because the room has a lot of large geometry and takes time to reflect the sound back from those surfaces, causing a slight inconsistency between reflections measured earlier versus slightly later. Throughout the curtain-and-wedges version, T20 and T30 appear identical. In both measures, the newer versions of the room have significantly lower reverberance time than the original (0.1s-0.2s less consistently), which was the intended effect from adding the curtain. These values are also all lower than the measured data (except where they are very close at 2000Hz and 4000Hz) and the original calculated RT, which further proves that these changes would make the difference we expect. It seems that changing the material of the wedges didn't do what we expected and may not have had an effect on the RT, but it sounds different in the audio, so we also measured another parameter to try and explain this.

Clarity was chosen because when listening to the auralizations of the space, adding more absorptive material in low frequencies made individual sounds much more clear and distinguishable. So, we wanted to look at the data to see if it was consistent with what we heard. C50 represents a ratio between early (before 50ms) and late (after 50ms) sound reflections and is characterized by how easily different sounds can be distinguished from each other. In the graph, it is clear that the alterations to the room did indeed make sound in the room clearer, and at all frequencies, than the original. However, neither of the new versions are consistently the highest, which explains why both sound clearer than the original room design but between them the changes are more subtle.

Auralization

Auralization is the process of rendering and simulating the way a sound source would produce a sound and how the sound field would behave within the given constraints of a room. These constraints include dimensions and material properties as well as the locations of the source and receiver of the sound.  Auralizations can be very helpful after spaces are designed and before they are built. They can help sound architects simulate desired types of sound sources (e.g. jazz band, vocalist, lecturer, etc...) and analyze rooms or halls before construction even begins to ensure that the resulting sound fields and qualitative listening experience are as desired by the future users of the space. With the architectural layout and knowledge of what materials are being planned to be used, a CATT model of the room could be run and the results analyzed to ensure the resulting space would provide satisfactory concert hall acoustics characteristics. Specifically, if the intended purpose of the space requires a certain RT (lower for recording studios or higher for performance spaces) the auralization and simulation can help project what the intended room configurations would present. They can also use this system as we are now, testing different materials, dimensions, or additions to the room that may sound the way sound behaves to see if the room can be improved in any way after it has already been built. Auralizations are also excellent tools for determining what sound fields will look like when sourced from a specific location and received at another. Specifying these locations would allow sound designers to ensure things sound as desired from the location of the audience listening to an ensemble on stage, for example. This way, sound quality can be optimized for desired audience locations. There are three primary components of an auralization: the impulse response, the convolution of IR with anechoic audio, and the spatial rendering of the convolution.

An impulse response is created when a sound affects a room, whether it is played through speakers or generated in person. The impulse used to measure IR time is typically short and instantaneously cut off, just as swiftly as it is played to excite the room quickly. The most common method of gathering IR data is to pop a balloon in the desired space to measure the short but strong sound energy it emits. Clapping one's hands once creates a similar effect. Impulses aren't always this short, however. Often, sine sweeps are used instead, played at full power throughout all frequencies, which allows for full control of the frequency spectrum on the basis of time. The results from measuring the IR of a room measured in this way allows for further analysis and quantitative reasoning to determine the acoustical characteristics of a room. Different results are acquired when the source and receiver types are changed -- balloons, for example, create sound energy that spreads evenly in every direction, whereas instruments or people speaking is very directionally biased. The type of microphone used also affects the numerical values that result. Auralizations take these impulse responses and use them to simulate reverberation. Using a wave-based method, as opposed to the more time-efficient geometric method, provides a more accurate representation of the movement of sound through space. Considering sound as a ray simplifies its movement which is still accurate enough to render a realistic auralization. Wave-based methods are much more costly, both in time and in money, and so it is not necessarily worth using for general room acoustics. Instead, it may be applied in more minute settings, such as small parts or strings on an instrument. However, if a wave-based method were to be used to simulate a room impulse response, it may be for rooms where lower-frequency sounds are more important as rays are accurate to high frequency but not so much on the lower end. Perhaps when designing a room that is ideally sound-proofed against noises from urban environments which produce a lot of low-frequency background noise, using a wave-based method to analyze the room and its acoustical properties would prove useful. Mostly, for these simulations, we are interested in the parts of the IR data that display the sound waves/particle behavior represented in the direct sound, first-order reflections, and the following consequential higher-order reflections.

The second component of auralization is convolution, which is achieved by inputting anechoic audio files to be convolved with the aforementioned IR data from the desired space. Taking the input signal and the impulse response, convolution multiplies the frequency spectra, combining the two audio sources so that similar frequencies are proportionally played up and ones that are not shared are played down.

The last step of auralization is spatial rendering, which was done in CATT for this project. Simulating the physical space in which sound is to be played allows for the program to predict what certain sounds will sound like played and heard from different positions throughout the room. The virtualized model can be done well before the space is actually built, which is what makes it so useful for the collaboration of acousticians and architects. Placing the source and receiver in different places around the room and making various changes to architecture and materials can help developers identify necessary changes in the design, which is exactly what we're trying to do with Fisher Performance Hall.  

For the purposes of our project and the analysis of Fisher, the IR data simulated in the TUCT program was convolved with anechoic jazz recordings in MATLAB to produce what the band would sound like in our area of interest. 

Convolved Audio Files: Jazz Band and Male Speech in the Middle & Back of Fisher

Discussion

Impact

Between the different convolved audio files, the most noticeable difference was an increase in clarity, particularly in low frequencies, between the original design and the inclusion of only the curtain covering the storage alcove. This change occurred in both the jazz band and male speech auralizations and at both the middle and back of the room, though is most pronounced for the male speech in the middle of the room. This also correlates with an increase in dryness between the original design and storage alcove curtain, which is helpful in understanding low-frequency speech as well as jazz, where fast-paced instrumental lines can become muddled with too much reverberation. Additionally, these observations correlate with differences in T20, T30, and C50 between the different versions of the room in different positions as well.

The difference between the revised design with only the storage alcove curtain and the modified wedges is far more subtle than between the original design and the storage alcove curtain, and actually trends in the opposite direction as the initial change. The modified room design featuring the storage alcove curtain and modified wedges is actually a bit more muddled in low frequencies than the version with just the alcove curtain, though is still far clearer than the original room design. This trend is also true for both the jazz band and male speech, and in both of the two positions in the room which were studied.

Degree to which modified design achieved goals

The purpose of these design changes was to reduce the reverberation time for lower frequencies. Upon review of the recent simulations, the T20 for the design changes are much lower for lower frequencies compared to what they were before. The original T20 value of the room with the preexisting curtains drawn was around 1.12 seconds at 125 Hz and 0.91 seconds at 250 Hz. With the revised changes, the simulated T20 values dropped across all frequencies, but especially across low frequencies.

The addition of the alcove curtain reduced the T20 value by around 0.17 seconds for frequencies above 250 Hz. The T20 for 125 Hz and 250 Hz were lowered 0.42 seconds and 0.31 seconds respectively. This definitely reflects the acoustical effect we sought, and brings the T20 for all frequencies into a much closer range. 

Strangely enough, the modification done to the wedges seems to undo a bit of the new curtain's work. The T20 values for 125 Hz and 250 Hz are 0.37 seconds and 0.25 seconds lower than original, which is less reduction than the previous 0.42 and 0.31.  We believe that this may be due to the plaster of the original wedges being more absorbent of lower frequencies than our new plywood wedges. Due to how high they are off the ground, we were not able to confirm the material of the wedges in our visit to Fisher Hall - we used our best judgement and decided that they were made of some form of plaster. It is unknown what is within each wedge - perhaps the absorbent material inside is specifically arranged to absorb primarily lower frequencies.

Feasibility for recommended changes

Our first design change was the implement a curtain in front of the storage alcove, as detailed previously. This would not be too hard to implement, and would not require much permanent alterations to any parts of the room. The installation process would be relatively simple - a curtain guide rail would be screwed in above the alcove, and a curtain would be hung from it. The guide rail would be the only permanent alteration to the room due to it being screwed on (if it were removed, screw holes would show) - the curtain can always be drawn to the side or removed completely. As this alcove is much shorter than the rest of the room, the curtain would be significantly lighter than the curtains lining the other walls of the room.

Our second design change would be a bit more time-intensive to implement, but would not change the shape of the room at all. To replace the wedges with plywood filled with absorbent material, the previous wedges would first have to be removed. Depending on how they were originally fastened, this may be easy or difficult. If they were bolted/screwed in, removal would simply involve unscrewing them. If they were more permanently affixed, more effort would be required. The new wedges can be assembled separately prior to mounting. The new wedges would be attached using the same method as before, and after installation, the room geometry would be the exact same as before.

The monetary cost of these modifications would actually be relatively low, as curtains can be found at decent prices, and plywood and absorptive materials are extremely cheap.

Overall, these recommended changes are extremely feasible to implement, and would not take much effort to enact.

Successes and limitations of the auralization process

Overall, given the relatively limited scope of the design changes and relatively simple design of the room, the auralization process was very well-optimized. Simulation time was kept relatively low, and comparing the jazz band auralization to experience listening to an actual jazz band in the room, it is apparent that the auralization is relatively accurate to the actual room. Additionally, due to the optimization of the model of the room, simulation times were kept very low, allowing for quick computation and analysis of auralization results. 

The most significant limitation present in the auralization process is the lack of certainty to how the auralizations will be experienced. Since a physical auralization space which uses speakers to simulate the room is assumed to not be viable in this study, the next best thing would be for anyone listening to the auralization to use headphones with good sound quality, and possibly a noise-cancelling function. While litening with such a pair of headphones, all of the discussed differences between versions of the room were apparent, but for those listening on other devices, such as earbuds, this may not be as apparent. In all cases, however, the binaural nature of headphones and earbuds means that the sound source used in the auralization will localize inside the head of the listener, rather than in front of them as would be the case in the actual room. Additionally, due to the small scope of this study, we were limited to only two positions in the room, though with the well-optimized model, it wouldn't be very difficult to produce additional auralizations with receivers at more locations.

Improvements for future iterations

There are a few ways that the auralization process could be improved for future iterations. 

First, the room materials could be determined more accurately. We were limited to what we could see from the ground. If we were to do a more detailed model of the room, we would need to access the materials list of the room, or use ladders to go up to some unreachable parts (lowered ceiling panels, wall wedges) to determine the materials.

Another way to improve the accuracy of the auralization would be to model the chairs accurately. Currently, the 50 chairs are modeled as two large blocks. While the difference may not be drastic, modeling each chair individually may improve the auralization's accuracy.

Finally, as mentioned above, a more accurate model of the room's acoustics could be determined if more receivers were set up throughout the room. This would help with learning how the room sounds from all positions, rather than just the two positions we used.