Part  I: Room characterization and Measurements

Methods

In order to characterize the acoustic properties of the room, we made both qualitative and quantitative assesments. We listened to each others voices from different locations in the room, as well as a harmonica being played from stage. We took measurements on both the physical dimensions and acoustical impulse response of the room. We used an omnidirectional microphone as a reciever and baloons as an impulse source to measure an impulse response recorded from two different locations in the space. One of us stood in the center of the stage with the balloon and the other sat in the center of the audience area with the microphone. We recorded 2 balloon pops from one microphone location, and then the reciever moved to another location in the audience, this time off to stage left and higher up. We ran into some issues with mechanical background noise as well as the sound of the balloon hitting the floor after being popped, but tried to mitigate these issues by moving away from the source of the noise and recording multiple pops.

In order to measure the physical dimenions of the space we used a tape measurer to measure the smaller components and then used the geometrical aspects of the room to extrapolate the larger values such as total diameter and height.  For example, we measured one of the secions of the walls to be about 2.51 meters long, and there are 26 sections making up the circular shape of the room, so we could approximate the circumference of the room as 66 meters.

Measurements

We plotted energy decay below and calculated T20, EDT, C50, and C80 values from our data. We recorded and analyzed data from two locations in the room. If we compare the graphs from the closer location with those from the further location from the sound source, we can see that the data from the closer location included more realistic values for reverberation time when looking at the raw unfiltered data. This is most likely due to the HVAC and balloon popping sounds that are included in our recording, and the length of the sample we used. 

Data Filtering and Noise Reduction

The raw data seems unrealistic. Over 7 seconds of T20 is just impossible. Upon reviewing the Schroeder decay curve, two methods were taken to reduce the effect of noise on T20 and EDT issues.

The first method is to use a modified t20 calculation method by setting the threshold at -18dB instead of -25dB. The linear (log scale) decay region is short enough due to the presence of noise that setting the threshold at -25dB is not going to be capable of capturing the characteristics of the decay. However, it seems like that the T20 value is still a bit too high due to the energy being added to the entire spectrum, which messes up the C50 and C80 calculations as well.

Another way was to use the noise cancelling algorithm provided by Adobe Audition, with the specs that are listed above. While the signal to noise comparison shows that the filtered signal is not far off from the original at higher frequencies, there are still of a couple of more peaks that are visible in the FFT diagram of the noise print. It is likely that the enclosed space with a lower ceiling could be a source of standing wave with the frequencies between 100Hz and 200Hz, as it correlates to the fundamental standing wave generated by parallel surfaces that are approximately 2-3 meters apart. This may affect the calculation as the signal to noise ratio is much lower at these peaks at lower frequency bands. 

Qualitative Results

Balch Arena has a relatively short reverberation time, which makes for a dry and dampened auditory experience. Sources and listening experiences in different parts of the space have different results, as the materials used in the treatments and furnishing of the room are not symmetric. There is a lot of noise from HVAC and lighting in the upper third of the space. We were able to hear more acoustical support when the sound source was closer to the bare wall rather than the side of the area covered by thick curtains, which we anticipated. We found ourselves having to speak loudly to hear when communicating from the stage area to the main level, and similarly between these levels and the upper catwalk. From attending shows in this space we remembered most using microphones, and with a full audience we would anticipate even more absorption during a live performance since the seats are only somewhat absorptive when empty. The space is much larger than it appears when looking all the way up to the ceiling above the HVAC pipes, so the sounds of our voices at a normal volume level have to travel much farther in this direction, for example speaking from the main level up to the catwalk, and reflect off surfaces that produce mainly diffuse reflections on this upper level.

When thinking about clarity, we noticed it varied dramatically depending on the direction of the sound source. We also noticed by comparison to lower and higher pitched sounds, mid frequencies around the vocal range seemed to project relatively clearly. This is supported by the C50 and C80 values at both locations, because the lower octave bands yield less clarity while the greater octave bands yield greater clarity before dropping off at the highest bands. This makes sense given that most performances are vocal and speaking performances in this space.

The musical listening experience of a single instrument would be different from a larger ensemble, but is still a good representative of the room acoustics. An instrument with a lower volume inside this dry room feels very little support and could not be heard very well in this situation. Articulation is clear, but the overall volume is quite low.

Discussion

Our first impression of Balch Arena was its surprisingly dry acoustic response considering the size of the space. When listening to a speaker on stage from the audience section, the sound of their voice is not supported by the room, you can mostly just hear the direct sound with very little mid and high frequency reverberations. We did however notice low-frequency sounds tended to linger and accumulate in the room, making it difficult at times to distinguish individual words or sounds.This qualitative assessment lines up well with our measurements. With an average T20 of 0.765 seconds across the six octave bands measured, the sound energy in the room decays at a rapid pace throughout a broad band of frequencies. At lower frequencies we calculated relatively high reverberation measurements. For each calculation at both microphone locations the highest T20 value was found at the 125 Hz octave band, with a bass ratio at the closer position of a whopping 1.96.  The T20 measurements at the closer position could be a result of the enormously large space on the top of the room as well as the lack of materials that are absorptive on the lower frequencies in comparison to the higher frequencies. The initial test at the further position still showed an insane T20 value at lower frequency ranges. Our best hypothesis to explain this phenomenon is the fact that there was loud low frequency background noise present during the recording at the farther position. This caused a high ambient sound energy in the low octave bands, influencing our T20 predictions. Since EDT and T13 is not as affected by ambient noise, the EDT and T13 values were more reasonable.

As for clarity, it was very easy to pick out discrete sounds from accross the room. When talking to a group-mate in the balcony from the stage, the group-mate could understand what was being said without trouble. This was supported by the C50 and C80 values, with an average of 6.4 and 10.4 dB respectively from the first listening position.

Compared to the measured T20/T13, the calculated sabine reverberation time is longer by a margin of larger than 10% for all frequency ranges. Some room of error could include measuring error, room shape estimation, scattering by the lighting section and volume calculation error. The room was not a square-shaped room, making it much harder to measure. Our take was rather simple by measureing the point to point distances, which may create quite some error when counting in the curvatures. The room is also not a perfect square or an oval. Our estimation was to calculate mean distances throughout different locations for approximating the overall shape to be circle. The top section of the room is a huge empty space in our calculation, where we made the assumption that the lighting section is acoustically transparent. However, the mesh and large amount of lighting may have created a lot more scattering than we anticipated. This could be reflected in the simulations as well as scattering in increased. The last error that may occur is the volume calculation. The arena is extremely tall, and there was not a simply way of measuring the height as it's not only unreachable but extremely dark up top. Our estimation of the total height is based on normal floor heights, which could lead to a major source of error.

Having a high bass ratio in a performance space is generally not desirable as it can negatively impact the overall sound quality of the space. A high bass ratio means that there is an excessive amount of low-frequency sound in relation to the mid and high-frequency sounds. This can result in a boomy and muddled sound, where individual notes or sounds are difficult to distinguish. It can also make it harder for performers and presenters to hear themselves clearly, which can negatively affect their performance. Additionally, a high bass ratio can make it more difficult for sound engineers to achieve a balanced mix, as the low-frequency sounds can overpower the other sounds in the mix. Therefore, it is important to strive for a balanced frequency response in performance spaces to ensure optimal sound quality for all types of performances. We would reccomend methods to decrease low frequency reverberations such as helmholtz absorbers and more elements to scatter low frequency sound.

In conclusion, while the Balch Arena has a dry acoustic environment that can be suitable for certain types of performances, its high bass ratio and relatively high ambient noise levels may limit its versatility and potential for creating optimal listening conditions for a wide range of performances. Therefore, there is room for improvement in the form of balancing the frequency response and reducing extraneous noise to enhance the overall acoustic quality of the theater.

Part 2: Auralization

Shown below are comparisons of the T20 and clarity values between our three models:

Auralizations

In acoustics, auralizations are a way of simulating the way that sound behaves in a physical space. They involve a virtual rendering of the space in three dimensions, as well as a computer algorithm that models the way sound waves propagate in the virtual space. The model uses the way the waves interact with objects, walls, floors and ceilings, as well as the material properties and absorption coefficients assigned to those materials. 

Once the sound wave propagation is modeled, the software generates an audio file of the simulated response.  The audio file can then be played back, usually accompanied by a visual representation of  the space. Auralizations are used in architectural acoustics to simulate and propose changes to the design of concert halls, recording studios, and other rooms.

We used the software CATT acoustic to create auralizations of a person speaking in the space as well as a jazz band playing in the orchestra pit. Jazz band is a very nice example of modern genre music with the involvement of a more presence rhythm section as well as amplified bass. Male speech voice was chosen due to the wider range of frequencies involved. Both sound sources are ideally placed in the middle of the stage. The two reciever positions are: in the audience section right in front of the performance area, and higher up in the audience section on the left side.

For the directivity of the sources we used the "talker" directivity for the male voice and the "wide cardioid" for the jazz band. 

Key: Model1 is OG room, Model2 is helmholtz, Model3 is balcony seating, last number is reciever position

Discussion

Based on the resulting T20 and clarity values of the simulations, it is clear that the adding the helmholtz absorbers has a large effect on the frequency response of the room. The low frequency reverberation is greatly reduced while the mid frequencies are more supported. This is also clearly observable in the auralizations as well.

Although adding the Helmholtz absorbers did block some of the reverberation in lower frequencies, the overall clarity of the sound is still somewhat muddy for the jazz band convolution. This is probably due to the absence of early reflections. 

It is reasonable to propose additional seating in the balcony since the structure is already present. This may pose safety risks and require changes to the evacuation plan of the arena, but would be worth exploring since the space has faced issues with exceeding capacity and turning away audience members for popular shows. Adding Helmholtz absorbers to all of the walls currently covered by curtains is somewhat less realistic since they can be expensive to manufacture for custom frequency blocking. This would also take up some additional space in the walkway around the arena, which could make the space more crowded.

All and all, adding both the Helmholtz resonator and seating on the balcony makes the most ideal improvement by adding clarity and mid-range frequency strength in the auralizations which are associated to the "Model 3" of the auralization clips.The auralization was based on a stereo track (jazz band) and a mono track (male voice). The simulation of the male voice clip rendered completely normal, but the convolved audio stereo track of the jazz band actually sounds really skewed to the left. Upon checking the IR, it appears that the IR on the left channel is a lot larger in amplitude than that on the right channel. While in the room, this was not such a big issue, but could be on a pair of in-ear monitors possibly due to the limitations of two seperate channels and the directivity of the simulation. This issue is less presented when wearing the over-head headphones. This could also be a problem associated with earphone tunings.

While these auralizations attempt to simulate how the room would sound with the proposed changes, they include many assumptions and estimations. For example the area of the seated balcony is modeled as absorptive seating, although there would likely be some reflective materials in the structure of the balcony if it were to be constructed. Another assumption that is carried through all of our measurements and simulations is the estimation of the upper volume of the arena. As we mention earlier, it was difficult to take measurements and see above the HVAC and lighting equipment, so we used our best judgement to fill in these gaps.

If we were to iterate on this design or propose further improvements to the space, it would be beneficial for us to obtain a floor plan or more precise measurements. We would also want to broaden our understanding of scattering, which would allow us to make better estimations for the upper part of the arena since the lighting section likely has a lot of scattering resulting from the reflective surfaces that we omitted in our model. Finally, combining the elements that improved our auralizations into one refined improved design would require features that increase early reflections and decrease low frequency reverberation. In these initial simulations we focused on two distinct improvements in order to see how they changed the acoustic response of the space independently, but combining these would give a better result as far as acoustic measurements and listening quality for the audience.