Reverberation Time Prediction Tools Comparison

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

In February, a municipal arts council in Portland, Oregon approved a renovation of a 1970s community hall into a flexible performance and lecture space. The room—used for chamber music one night and amplified town-hall meetings the next—had a long history of complaints: “boomy speech,” “muddy music,” and “it’s loud even when nothing is amplified.” The council wanted a solution that would be defensible to stakeholders and predictable enough to avoid change orders late in construction.

Our team consisted of a project manager from the general contractor, a lead audio engineer (my role), an acoustical consultant, and the venue’s facilities manager. The brief was specific: achieve RT60 of 1.2–1.4 s (mid-band) in “music mode,” and 0.9–1.1 s in “speech mode,” using retractable elements if needed. The renovation also included a new PA system and stage lighting, but reverberation control was the schedule-critical path because it affected ceiling work and wall finishes.

The case study focuses on the tool comparison that guided the design: we evaluated three prediction approaches—simple statistical calculators (Sabine/Eyring), a dedicated statistical acoustics program (with octave-band modeling), and a hybrid ray-tracing workflow—then reconciled them against site measurements.

2) Challenges and requirements at the outset

The hall’s geometry and finishes were the core challenges. The room was approximately 24.6 m (L) × 14.8 m (W) × 6.1 m (H), with a shallow stage at one end and a low-slope ceiling that peaked over the seating area. The calculated volume was ~2,220 m³, large enough that mid-frequency RT would materially affect speech intelligibility and gain-before-feedback.

• Existing surfaces: painted CMU sidewalls, hardwood floor, gypsum ceiling on metal deck, and a rear wall of glass block.
• Use patterns: 60–240 occupants depending on configuration; mixed seating (chairs, risers) changed absorption significantly.
• Noise constraints: HVAC retrofit targeted NC-25 for classical events; any additional duct lining had to be coordinated with MEP.
• Budget: $95k allocated for acoustic treatment and installation (not including MEP noise work).
• Schedule: 14 weeks from design kick-off to substantial completion; ceiling procurement lead times were a risk.
• Architectural constraints: limited tolerance for “studio look.” Treatment had to be paintable or fabric-wrapped in approved colors, and the rear wall could not be fully covered due to an egress corridor and a display area.

Tool selection mattered because stakeholders wanted confidence in predicted RT outcomes before ordering large quantities of absorptive ceiling and wall panels. The contractor also needed quantities early to lock procurement.

3) Approach and methodology chosen

We structured the work in three parallel tracks:

1. Baseline measurement: Measure existing RT and background noise to calibrate predictions.
2. Prediction tool comparison: Model the same “as-is” and “proposed” conditions in multiple tools, then compare sensitivity to assumptions (occupancy, absorption coefficients, scattering).
3. Design convergence: Use the most reliable method per frequency band and project stage: quick calculators for early sizing, octave-band statistical modeling for procurement quantities, and ray-tracing checks for spatial issues (rear-wall flutter, stage support).

The tool set was deliberately mixed:

• Sabine/Eyring spreadsheets: internal templates built around ISO 354 absorption coefficients and typical occupancy values.
• Statistical acoustics software: an octave-band RT calculator that supports Eyring, Millington-Sette, and user-defined absorption by surface and audience.
• Hybrid ray-tracing: a simplified 3D model used for early reflection paths and sanity-checking RT trends above 500 Hz (where scattering and directional effects can matter).

The project goal was not “which tool is best” in the abstract, but “which tool is best for this room, this timeline, and this procurement risk.”

4) Step-by-step execution narrative

Week 1–2: Site survey and baseline measurements

We started with a half-day site survey to confirm dimensions, surface types, and any hidden cavities. Then we ran baseline acoustical measurements during an off-hours window to minimize HVAC and street noise.

Measurement setup (typical for this scale of room):

• Measurement method: impulse response via sine sweep
• Audio interface: RME Babyface Pro FS
• Measurement mic: Earthworks M30 (calibrated)
• Source: QSC K12.2 on a stand at ~1.8 m height (representative talker height at stage edge)
• Software: Room EQ Wizard for capture and T20/T30 extraction
• Mic positions: 8 positions across seating area, plus 2 near rear wall

The measured mid-band RT60 (500 Hz–1 kHz) averaged 1.85 s with a spread of ±0.15 s across positions. Low frequencies were longer: ~2.2 s at 125 Hz. Clarity (C50) for speech was consistently negative in the rear third of the room, aligning with user complaints.

Week 2–3: “As-is” modeling in three tool categories

We built an “as-is” surface schedule: floor area, wall area by material, ceiling area, stage front, and glazing. For absorption coefficients, we used manufacturer/standard references (painted CMU, sealed hardwood, gypsum board, glass) and assumed 10% seating absorption because chairs were typically stacked along sidewalls in the empty condition used for measurement.

In the Sabine spreadsheet, the “as-is” RT prediction landed at ~1.65 s at 1 kHz, underestimating measured RT by ~0.2 s. Eyring moved closer (~1.75 s) because average absorption was low enough that Eyring’s correction mattered.

The statistical software model (octave-band) predicted 1.78 s at 1 kHz and 2.05 s at 125 Hz. The ray-tracing model produced ~1.7–1.8 s mid-band depending on the scattering coefficients assigned to the ceiling and side walls, but it was less stable below 250 Hz (expected for simplified geometric acoustics).

The most important early conclusion was practical: our inputs were plausible and the models were within 10–15% of measured RT, good enough to proceed—provided we calibrated to measurement rather than trusting defaults.

Week 3–5: Treatment concepts and iterative predictions

We developed three concept packages:

1. Ceiling-heavy absorption: replace ~60% of the ceiling with NRC 0.80 mineral fiber panels, minimal wall treatment.
2. Balanced ceiling + wall: ~35% ceiling absorption plus broadband wall panels at rear and sidewall first-reflection zones.
3. Variable acoustics: moderate fixed absorption plus retractable heavy curtains along the rear and upper sidewalls for “speech mode.”

Each concept was run through the spreadsheet first for quick quantity estimates, then through the statistical model for octave-band RT, and finally checked in ray-tracing for reflection behavior (rear-wall flutter, stage-to-rear slapback).

Week 5–7: Design freeze and procurement quantities

We chose the variable acoustics concept because it met both RT targets without making the room too dead for chamber music. Design freeze occurred at the end of week 7 to allow ordering ceiling materials and curtain track hardware.

Week 10–12: Installation coordination and field changes

During demolition, we discovered a plenum cavity above a section of the ceiling that was acting like a lightly damped absorber at low-mid frequencies. Removing old ductwork changed the effective absorption slightly. We updated the model with the revised ceiling construction and adjusted the wall panel quantity by +12 m² to stay on target.

Week 13–14: Post-install verification

We repeated the measurement protocol with the new finishes and with curtains in both open and closed positions. We also ran a quick STI-PA check with the installed system set to a conservative EQ (no “intelligibility boost” tricks until we confirmed the room behavior).

5) Technical decisions and trade-offs made

The biggest decision was how to allocate absorption to control mid-band RT while not creating an overly absorptive high end. Mineral fiber ceilings can pull 2–4 kHz down quickly, which can reduce “air” for acoustic music and make amplified systems sound dull unless EQ compensates.

We used the models to push absorption where it gave the most predictable benefit:

• Ceiling absorption: limited to ~280 m² of NRC 0.80 tiles (about 45% of the net ceiling area). This provided broad mid-high control without overcorrecting.
• Wall broadband panels: ~54 m² of 50 mm fiberglass panels (fabric-wrapped, tested coefficients) placed on the rear wall and rear sidewalls to reduce slapback and improve C50 in the back third.
• Variable curtains: 38 linear meters of 600 g/m² velour on track, deployed for speech mode. Curtains were chosen because their absorption is more effective above ~250 Hz, complementing the panels.
• Low-frequency control: we avoided heavy tuned trapping because budget and wall depth were limited. Instead, we accepted a slightly longer 125 Hz RT and focused on controlling 250 Hz–2 kHz, where speech and most music clarity lives in this venue.

Trade-off: the ray-tracing model suggested modest diffusion on the rear wall could reduce flutter without adding much absorption. But adding diffusion increased fabrication complexity and cost. We instead used panel placement with small air gaps (25–50 mm) behind some panels to improve low-mid absorption efficiency, a simpler approach that fit the schedule.

Another trade-off was audience absorption uncertainty. The hall sometimes hosts seated lectures (high absorption) and sometimes standing receptions (lower absorption per person and different distribution). We modeled three occupancy cases (0%, 50%, 100%) and ensured the fixed treatment didn’t create RT60 < 0.8 s when fully occupied with curtains closed.

6) Results and outcomes with specific details

Post-install measurements were taken with 120 chairs deployed (typical event setup), HVAC at normal operation, and curtains tested in two positions.

Measured RT60 (average across 8 positions):

• Music mode (curtains open): 125 Hz: 1.95 s, 250 Hz: 1.55 s, 500 Hz: 1.32 s, 1 kHz: 1.26 s, 2 kHz: 1.18 s
• Speech mode (curtains closed): 125 Hz: 1.85 s, 250 Hz: 1.35 s, 500 Hz: 1.05 s, 1 kHz: 0.98 s, 2 kHz: 0.92 s

That landed inside the target bands for mid frequencies. The low end remained longer than mid-band, but it was improved by ~0.25–0.35 s compared to baseline due to the added porous absorption and slight changes in ceiling construction.

Speech clarity improvements: C50 increased from roughly -2 to -4 dB in the rear seating area to +1 to +3 dB in speech mode. STI-PA (measured with the installed loudspeakers at typical level) improved from 0.46–0.52 pre-renovation to 0.62–0.68 in speech mode after treatment and system tuning.

Tool comparison outcomes (what matched reality):

• Sabine-only spreadsheets were fastest for ballpark sizing but consistently optimistic in the “as-is” case and slightly pessimistic once we added significant absorption (because it’s sensitive to average absorption assumptions).
• Eyring-based statistical modeling tracked measured mid-band RT most closely after calibration: within ±0.08 s at 500 Hz–1 kHz for both curtain states.
• Ray-tracing was most useful for identifying the rear-wall slap and stage-to-rear reflection path. It was less reliable for absolute RT below 250 Hz, but it helped justify rear-wall treatment locations to the architect and contractor with visual reflection plots.

From a project management standpoint, we stayed within the $95k acoustic treatment budget (final: $91.4k including labor) and met the schedule without expedited shipping by freezing ceiling quantities in week 7.

7) Lessons learned and what could be done differently

Three lessons stood out.

1. Calibrate early or don’t trust any model. The first measurement day paid for itself. Once the “as-is” model matched within ~10%, we could compare tools meaningfully. Without calibration, the discussion would have been about whose coefficients were “right,” not what the room would do.
2. Audience absorption is the biggest swing factor in multipurpose rooms. We could have done a second baseline measurement with chairs deployed and a small audience present to reduce uncertainty. Even 30–40 people can shift 1 kHz RT noticeably in a room this size.
3. Use ray-tracing for placement decisions, not just RT numbers. The most valuable ray-tracing output was not RT; it was showing the reflection path that created perceived “slap” at the back. That directly informed where we spent wall-panel budget.

If we did it again, we would budget an extra half day for source directivity testing. Using a point-source loudspeaker is common for RT, but the venue’s real sources (speech at a lectern, small ensemble, PA) have different directivity. A basic set of measurements with a directional source (or at least multiple source positions) would better bracket expected variance.

8) Takeaways applicable to other projects

For audio engineers and project managers comparing reverberation prediction tools, these were the practical takeaways from the job:

• Start with a spreadsheet, but don’t end there. Sabine/Eyring calculators are ideal for early budgeting and quantity checks. Use them to move fast, then validate with octave-band statistical modeling before procurement.
• Pick the “right” tool per question. If the question is “How many square meters of absorption do we need?”, statistical tools are efficient. If the question is “Where is the slap coming from?”, ray-tracing earns its keep.
• Model multiple occupancy states explicitly. Create at least three cases (empty, typical, full) and ensure you don’t accidentally design a room that only works in one condition.
• Document coefficients and assumptions. The fastest way to derail a project is to lose track of which absorption coefficients were used, at what mounting condition (direct mount vs air gap), and under which standards. Keeping a single surface schedule tied to drawings prevented confusion across architect/GC/consultant.
• Plan for field discoveries. Older buildings hide cavities, leaks, and construction changes that shift acoustics. A contingency (we carried ~8% of treatment budget) made it possible to adjust panel quantities without a contract fight.
• Define success in measurable terms. RT60 targets are helpful, but pairing them with clarity metrics (C50 for speech, or EDT where relevant) produces a more usable room and a more persuasive closeout report.

The end result was a room that could host chamber music without feeling overdamped, while still supporting intelligible speech without aggressive system processing. More importantly, the team had a repeatable workflow: measure, calibrate, compare tools, and then commit quantities with enough confidence to protect the schedule.