Diffusion in Concert Hall Design

1. Project overview: what, where, who, and why

In late 2023, the City of Marrowgate approved a renovation of the 1970s-era Northbank Civic Hall, a 1,050-seat multipurpose venue used for orchestral concerts, touring jazz acts, spoken-word events, and occasional amplified pop shows. The renovation goal was simple on paper: “make it sound like a concert hall again” without sacrificing speech clarity or increasing noise complaints from neighboring residential blocks.

SonusGearFlow was brought in as the project documentarian and technical advisor alongside the acoustics consultant (Alder & Finch Acoustics), the architect (Benton+Kline), and the integration contractor (Harbor Signal). The client team included the city’s project manager, the hall’s technical director, and a programming committee representing the symphony, a university music department, and touring promoters.

The venue’s most persistent complaint was “dead in the center, harsh on the sides.” Measurements and listening tests confirmed it: strong early reflections in the 1–4 kHz region from parallel wall segments, audible flutter echo in the upper balcony vomitories, and a lack of lateral energy in the stalls that made unamplified strings feel small. The guiding “why” for this project was to reintroduce supportive, evenly distributed reflections—especially lateral reflections—while controlling specular, seat-to-seat variability. Diffusion became a primary tool, not as a decorative add-on, but as a deliberate mechanism for steering and scattering energy.

Renovation construction ran from January through October 2024, with acoustic commissioning and opening performances scheduled for November 2024.

2. Challenges and requirements at the outset

The existing room was a shallow fan shape: 32 m long, 24 m wide at the rear, and 14 m at the proscenium. Ceiling height varied from 10 m above the stage to 14 m over the rear stalls, with a heavy plaster ceiling and a series of large, flat soffits housing HVAC. Walls were painted CMU with intermittent wood panel inserts installed in the 1990s as a cosmetic refresh.

Key constraints emerged early:

• Reverberation target conflict: The symphony wanted ~1.9–2.1 s mid-frequency RT for orchestral repertoire; the speech users wanted under 1.4 s for conferences. The city would not fund motorized banners across all surfaces, so diffusion had to work with limited variable absorption.
• Budget and scope limits: The acoustic treatment budget cap was $420k, including materials, rigging modifications, access equipment, and labor. Ceiling replacement was out of scope.
• Noise and mechanical restrictions: NC-20 was required for classical programming. Any added structures could not compromise airflow or create rattles.
• Stage support: Musicians reported poor ensemble due to weak early energy and “late slap” from the rear wall. The stage shell was undersized and largely absorptive.
• Safety and maintainability: Treatments must meet Class A fire rating, be serviceable from existing catwalks, and avoid collecting dust in unreachable cavities.

The most pressing acoustic defects were not a simple “too live/too dead” issue. The hall had unevenness: clarity and brightness differed dramatically by seat, and certain notes “pinged” due to specular reflections. Diffusion was selected to reduce these seat-to-seat anomalies without pushing RT beyond acceptable limits.

3. Approach and methodology chosen

The team used a three-part methodology:

1. Baseline measurement and mapping: Spatial measurements to quantify variability and locate reflection hot spots.
2. Modeling and iteration: A hybrid workflow: geometric acoustic modeling (CATT-Acoustic) for early reflection and energy distribution, with targeted hand calculations for diffuser bandwidth and depth limitations.
3. Prototype-and-verify: Build and test a limited set of diffusers on site (one balcony bay and one sidewall section) before committing to full fabrication.

Baseline measurements were performed over three nights to avoid outside noise: one empty-hall session and two with 600 seats occupied using seat-absorber proxies (fabric-wrapped mineral wool pads placed on seats). Tools included an NTi XL2 with M4260 mic, a dodecahedron source (NTi DS3), and Room EQ Wizard for spot checks. Impulse responses were captured at 24 audience positions and 6 on-stage positions. The consultant also used a B&K 2250 for verification.

The initial targets were set as ranges rather than single numbers: RT60 (500 Hz–1 kHz) of 1.8–2.0 s for orchestral mode and 1.4–1.6 s for speech mode, C80 between -1 and +3 dB for orchestral seats, and STI ≥ 0.55 for speech with the installed PA. Most importantly, seat-to-seat standard deviation for C80 in the stalls needed to drop by at least 30% relative to baseline.

4. Step-by-step execution narrative

Week 1–3: Baseline findings and problem localization

The impulse responses showed two recurring issues. First, strong early reflections at 18–28 ms from the sidewalls were highly specular, producing comb filtering and harshness for seats under the balcony edges. Second, late energy arrived unevenly, with a pronounced reflection cluster around 95–120 ms coming from the rear wall and upper balcony face, audible as a “slap” on stage and in rear-center seats.

The modeling team traced the 18–28 ms reflections to long, uninterrupted wall segments between 3 m and 7 m above the floor—exactly where older wood panels were mounted in a mostly flat plane. The late slap was linked to the rear wall’s shallow concavity and the balcony fascia acting like a reflector.

Week 4–6: Define diffuser strategy and bandwidth

Diffusion was split into three zones:

• Sidewall diffusion (stalls): Reduce specular reflections and increase lateral energy consistency from ~700 Hz upward.
• Rear wall diffusion: Break up the 95–120 ms reflection cluster without absorbing so much that the hall loses warmth.
• Stage enclosure upgrades: Replace absorptive elements with reflective/diffusive components to improve ensemble support and early energy to the audience.

Depth constraints were real. The architect allowed a maximum of 180 mm build-out on sidewalls to preserve egress width and sightlines, and 250 mm on the rear wall. These depths informed the diffuser choices: deep, low-frequency QRD arrays were not feasible on sidewalls, so the design emphasized mid/high diffusion paired with selective absorption elsewhere for mode control.

Week 7–10: Prototype installation and listening tests

Harbor Signal fabricated two prototypes:

• A 1.2 m × 2.4 m 2D “skyline” diffuser, 150 mm max depth, built from flame-rated MDF blocks with a matte polyurethane finish.
• A 1D QRD panel, 200 mm depth, 1.2 m × 2.4 m, using a 7-well sequence intended to diffuse from roughly 600 Hz to 4 kHz.

The prototypes were mounted in a sidewall bay near row M and on the rear wall above the center aisle. The team ran A/B impulse response tests and conducted listening sessions with a string quartet and a single spoken-word performer using the existing PA (d&b audiotechnik Y-Series flown L/R with a center fill).

The 2D skyline delivered a noticeable reduction in “zing” for the side seats but did not materially improve warmth. The 1D QRD improved lateral image and reduced the sense of slap when placed on the rear wall. Based on this, the final design used a combination: 1D QRD arrays where a dominant reflection path needed to be broken up directionally, and 2D diffusion where scattering in multiple planes improved seat-to-seat uniformity.

Week 11–22: Full fabrication and coordinated installation

Fabrication began in parallel with HVAC rework. A key coordination step was ensuring diffuser backs were sealed to prevent them acting as unintended Helmholtz resonators. The team also specified neoprene isolation washers on mounting brackets to prevent buzzing under high SPL touring acts.

Installation was sequenced to avoid damaging finished surfaces: rear wall first (largest panels, highest lift requirements), then upper balcony face, then sidewalls, then stage shell components. All diffusers were labeled and installed per a location map tied to the model, because rotating a 1D QRD by 90 degrees changes the diffusion axis and can reintroduce specular problems.

Week 23–26: Commissioning, tuning, and acceptance testing

After construction, the team repeated the 24-position measurement grid. In addition, they performed a “seat variance sweep” across three rows (E, M, U) capturing impulse responses every two seats to quantify variability. This data was used to confirm that diffusion reduced seat-to-seat spread, not just average values.

5. Technical decisions and trade-offs made

Three decisions drove most of the trade-offs:

Diffusion vs. absorption for problem areas

The initial instinct from non-technical stakeholders was to “treat harshness” with absorbers. The model showed that adding broadband absorption on the sidewalls would reduce the hall’s lateral energy and perceived envelopment, making the room clearer but smaller. Instead, the team used diffusion to break up the specular paths and reserved absorption for targeted low-mid buildup under balconies.

1D QRD vs. 2D skyline selection

QRD panels were chosen for the rear wall and balcony face where a strong, coherent reflection path had to be redirected and scattered horizontally. The final rear wall system used 18 panels (each 1.2 m × 2.4 m), 220 mm depth, arranged in a staggered pattern. Sidewalls used 2D skyline-style arrays with 120–160 mm depths in 30 panel bays, because multi-axis scattering helped reduce tonal shifts for off-center seats.

Depth limits and diffuser bandwidth

With 150–220 mm depths, diffusion below ~500–700 Hz is limited. Rather than forcing deeper diffusers (and losing seats or aisle width), the design leaned on other methods for low-frequency control: sealing wall cavities, adding 50 mm mineral wool above the balcony ceiling clouds (hidden), and improving stage shell reflectivity to increase early energy rather than relying on late low-frequency reverberation.

Material selection was also a trade-off: hardwood would have been more durable, but cost and lead time pushed the team to MDF with a fire-rated coating. To address durability, corners were reinforced with hardwood edging on panels installed within reach of audience traffic.

6. Results and outcomes with specific details

Acceptance testing in October 2024 showed measurable improvements aligned with user feedback:

• RT60 (occupied proxy): 1.92 s at 1 kHz (up from 1.68 s baseline) with orchestral mode curtains retracted; 1.52 s with side and rear absorptive curtains deployed (new curtains added only at rear corners and side upper sections—limited but effective).
• C80 (stalls average): improved from -2.5 dB baseline to +0.6 dB average in orchestral mode, with less “spikiness” in frequency response at off-center seats.
• Seat-to-seat variability: standard deviation of C80 across the stalls measurement sweep decreased by 37% (from 2.7 dB to 1.7 dB), meeting the uniformity goal.
• Stage support (STearly proxy): musicians reported improved ensemble, corroborated by a 2–3 dB increase in early energy at on-stage mic positions between 20–80 ms, attributed to the upgraded stage shell and rear-wall diffusion reducing discrete slap.
• Speech performance: STI improved from 0.49 to 0.60 average using the existing d&b Y-Series system after retuning EQ (less aggressive 2–4 kHz cuts were needed because harsh reflections were reduced rather than absorbed).

Subjectively, the most consistent comment from visiting engineers was that the room became “predictable.” Touring A1s noted they could push vocal level without encountering sudden feedback nodes tied to narrow reflection paths. The symphony’s conductor reported improved “bloom” in the stalls without losing articulation in fast passages.

The project remained within the $420k acoustic treatment budget. Diffuser fabrication and installation accounted for approximately $265k of that figure; curtains and hidden absorption accounted for ~$95k; measurement, commissioning, and contingency made up the rest. The diffuser portion required 11 installation days with a two-person lift crew and one carpenter, plus two night shifts to avoid conflicts with other trades.

7. Lessons learned and what could be done differently

Two issues surfaced during commissioning:

• Orientation control: Despite labeling, three rear-wall QRD panels were installed rotated 90 degrees. The effect was subtle in casual listening but measurable in the impulse responses at rear-center seats. A final “panel orientation walkdown” checklist would have caught this earlier.
• Paint finish and high-frequency behavior: The initial finish spec used a slightly textured coating for scuff resistance. That texture introduced a small increase in HF scattering on already diffusive surfaces, which was not harmful but complicated correlation between modeled and measured HF energy. In future projects, finish texture should be specified with acoustic behavior in mind, not only durability.

If the team could revise the scope, the first upgrade would be a modest increase in allowable diffuser depth on the rear wall (from 220 mm to ~300 mm) to extend effective diffusion closer to 400–500 Hz. The second would be additional variable absorption for speech-heavy events—specifically motorized banners above the upper sidewalls—because the current curtain deployment met targets but required manual changeover and staff time.

8. Takeaways applicable to other projects

• Use diffusion to fix consistency, not just “liveness.” In many halls, the biggest win is reducing seat-to-seat tonal swings caused by specular early reflections. Measure variability explicitly; don’t rely on averages.
• Prototype in the actual room. Even a single bay of diffusion can reveal whether you need 1D redirection or 2D scattering. The cost of prototypes is small compared to fabricating the wrong treatment at scale.
• Depth limits define your bandwidth. If you can’t build deep, don’t pretend you’re fixing low-frequency problems with shallow diffusers. Pair diffusion with hidden absorption, sealing, and stage support improvements to achieve a balanced result.
• Coordinate details that cause noise. Panels that buzz under touring SPL can ruin an otherwise excellent acoustic design. Specify isolation washers, rigid backing, and sealed cavities as part of the diffuser package.
• Installation orientation matters. Treat diffuser layout like loudspeaker aiming: document it, label it, and verify it. Small mistakes can reintroduce the very reflection path you’re trying to break.
• Model early reflections, then verify with impulse responses. Diffusion is most effective when targeted at specific time windows (e.g., 15–35 ms sidewall reflections or 80–130 ms rear-wall slap). Use measurements to confirm you changed the time structure, not just the look of the room.

The Northbank Civic Hall renovation demonstrated that diffusion is not a cosmetic upgrade—it’s an engineered solution for controlling reflection quality and spatial uniformity. For project managers, the core management insight was that diffuser work touches architecture, fire code, rigging, and finish schedules. For engineers, the key technical insight was to treat diffusion as a directional tool with measurable outcomes, backed by prototypes and a verification plan.