How to Conduct an Acoustic RT60 Survey

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

In February, we were brought in to document and execute an RT60 survey for the main multipurpose hall at Easton Civic Arts Center, a mid-sized municipal venue in the Pacific Northwest. The room hosts everything from town halls and spoken-word events to small orchestra rehearsals and amplified community concerts. The client team included the city’s facilities project manager (owner’s rep), a local AV integrator responsible for the upcoming sound system refresh, and a consulting architect managing a minor interior renovation (new seating and wall finishes).

The immediate driver was a familiar one: intelligibility complaints for speech events, especially from the rear seating. Staff were compensating with higher SPL, which increased feedback risk and listener fatigue. The city wanted an evidence-based acoustic baseline before authorizing any treatment, and the integrator wanted RT60 data to support system design choices (coverage, EQ strategy, and whether to budget for steerable columns).

The goal wasn’t to write a full acoustic report with modeling and prescriptions; it was to run a practical, repeatable RT60 survey that could guide decisions and establish “before” metrics for post-renovation verification.

2) Challenges and requirements at the outset

The hall is a rectangular shoebox-style room with a shallow stage, balcony overhang, and a mix of materials that evolved over decades: painted CMU side walls, a lightly textured drywall ceiling with HVAC diffusers, carpet in aisles only (the seating area is mostly hard floor under removable seating), and wood wainscot on the rear wall. The volume is approximately 6,800 m³ (measured from drawings and spot-verified on site), and the room seats 520 depending on configuration.

A few constraints made the survey more complicated than a textbook RT60 measurement:

• Limited access window: We had one 6-hour block on a Monday morning (08:00–14:00) between weekend events and an afternoon school rehearsal.
• Operational noise: The building required the HVAC to remain in “occupied” mode for life-safety ventilation. Background noise sat around 38–45 dBA depending on diffuser proximity.
• Variable room setup: Seating was in a “flat floor” arrangement with portable chairs, and the city wanted data that could translate to other configurations (banquet, lecture).
• Stakeholder expectations: The integrator wanted a single RT number, while the owner’s rep needed frequency-dependent values to justify treatment scope. We had to set expectations early: RT is band-limited, position-dependent, and best reported as a curve (e.g., octave bands), not a single headline.

We established success criteria: deliver octave-band RT60 (T20/T30 where valid) at representative locations, document measurement conditions, and provide an interpretation against common targets (e.g., speech-focused multipurpose hall: ~0.9–1.2 s midband occupied). We also agreed to include early decay time (EDT) where reliable, because EDT correlates strongly with perceived “liveness,” especially for speech.

3) Approach and methodology chosen

For a room of this size, we chose an impulse-response-based method using exponential sine sweep (ESS) rather than interrupted noise. ESS gives robust impulse responses in the presence of moderate background noise and makes it easier to repeat measurements quickly at multiple positions.

Standards guidance (ISO 3382-1 for performance spaces) informed the plan, but we kept it practical: a grid of microphone positions across the audience plane, a few positions under the balcony, and at least two source locations representing typical loudspeaker placement (center stage and stage left). We aimed for 12 mic positions and 2 source positions, yielding 24 IR captures, plus duplicates at a few points to check repeatability.

Equipment choices were based on reliability, calibration, and speed:

• Measurement software: Room EQ Wizard (REW) for capture and analysis, because it supports ESS, multi-measure workflows, and exports RT parameters by band.
• Interface: RME Babyface Pro FS (stable clocking, predictable levels).
• Measurement microphone: Earthworks M30 (omni, flat response), with an iSEMcon EMX-7150 as backup. Both were individually calibrated; we used the M30 for all primary data for consistency.
• Acoustic calibrator: 1 kHz / 94 dB calibrator to confirm sensitivity and verify the signal chain.
• Source: dodecahedron loudspeaker (NTi DS3) driven by a Crown XLS 1502 amplifier. A dodeca source approximates omnidirectional radiation, which is helpful when you’re characterizing the room rather than a specific PA loudspeaker.
• SPL metering: handheld Class 2 meter for quick checks and a logging app to track background noise trends during the session.

We also agreed on data hygiene: document temperature/humidity (affects air absorption slightly at HF), note HVAC state, and photograph mic and source locations with tape marks so the survey could be replicated after renovation.

4) Step-by-step execution narrative

Pre-site prep (Day -2 to Day -1)

Two days before the site visit, we reviewed architectural PDFs and plotted likely mic positions: three rows front/mid/rear, four seats across per row (L/C/R plus a side position near the wall), plus two positions under the balcony. We converted that into a one-page field sheet with a table: mic ID, coordinates relative to centerline, height, and notes. We pre-labeled gaffer tape flags (M1–M12 and S1–S2) to reduce on-site friction.

The day before, we verified the measurement chain in the shop: mic calibration file loaded into REW, loopback timing reference tested, and sweep level established to avoid clipping at the interface input while still achieving strong SNR in the tail. We packed an extra mic stand, sandbags, and a 30 m XLR to avoid last-minute routing problems.

On site: setup and calibration (08:00–09:15)

We arrived at 08:00 with the hall empty and the HVAC running. First task: measure and log background noise at three points (front center, mid rear, under balcony). We saw 41 dBA at mid-room and 45 dBA under the balcony, dominated by low-frequency airflow and some midband hiss from diffusers.

We placed the dodeca source at S1: center stage, 1.5 m above stage floor, about 1 m upstage of the proscenium line. Source height matters; too low and you overemphasize floor reflections, too high and you bias toward ceiling paths. We chose 1.5 m to approximate a talker height and a typical PA acoustic center.

Next, we set the microphone height at 1.2 m above finished floor for audience positions (seated ear height). Under the balcony, we kept the same height to isolate the effect of the overhang.

Calibration was straightforward: 94 dB at 1 kHz to confirm mic sensitivity, then we ran a short sweep to set output gain. We targeted a measurement level around 80–85 dBA (C-weighted) at mid-room, which provided enough headroom above background without being disruptive to adjacent offices.

Capturing impulse responses (09:15–12:00)

We captured sweeps at mic positions M1–M12 with the source at S1. Each capture included two sweeps averaged to improve SNR and reject transient noise (a door closing, HVAC cycling). We kept sweep length at 10 seconds with a 2-second pre-delay and sufficient post-capture to ensure the decay tail was recorded.

A practical detail that saved time: we routed the sweep and record trigger from a single laptop, and we disciplined ourselves to move only one variable at a time. The source stayed fixed for the first pass; only mic position changed. We used tape marks on the floor to keep positions repeatable and photographed each setup.

Under the balcony, we ran into the first real-world annoyance: a persistent 63 Hz rumble from HVAC. The tail at low frequencies was being masked, making T30 unreliable below 125 Hz at some locations. Rather than forcing a number, we flagged those bands as “insufficient decay range” and leaned on T20/EDT where the decay curve had enough usable dynamic range.

Second source location and duplicates (12:00–13:15)

After a quick check-in with the project manager, we moved the dodeca to S2: stage left, 2 m from the side wall and 1.5 m high. This was meant to simulate a common loudspeaker cluster or a presenter off-center. We re-ran M1–M8 (front and mid positions plus under-balcony), focusing on the areas where intelligibility complaints were strongest.

We also repeated two mic positions (M4 mid-center, M11 under-balcony center) from the first pass. Repeatability came in within about 0.05–0.08 s in midbands, which was acceptable given the active HVAC and minor positional tolerances.

On-site review and sanity checks (13:15–14:00)

Before teardown, we did a quick review of decay curves in REW. We looked for:

• Clean impulse responses without clipping or obvious anomalies
• Reasonable decay linearity in the Schroeder-integrated curve
• Sufficient decay range (ideally 35 dB for T30; minimum ~25 dB for T20)
• Outliers indicating a bad capture (e.g., a chair scrape during a sweep)

We re-shot one position where a sudden HVAC hiss spike occurred mid-sweep. That five-minute correction saved an hour of debate later.

5) Technical decisions and trade-offs made

Several decisions involved trade-offs that are worth calling out because they come up on nearly every RT60 survey:

• HVAC on vs. off: Turning HVAC off would improve low-frequency decay visibility, but the building required occupied ventilation. We chose realism over ideal conditions and documented background levels. The data represents how the room behaves during actual events.
• Dodeca source vs. using the installed PA: Using the PA might reflect real event conditions, but it introduces loudspeaker directivity and DSP processing as variables. The dodeca kept the survey focused on the room’s reverberation characteristics. We noted that system tuning can’t “fix” excessive midband RT.
• Number of positions vs. schedule: ISO-style coverage could suggest more positions, but within six hours we prioritized a representative grid and duplicates for repeatability. Twelve mic positions gave enough spatial insight to identify under-balcony behavior and rear-wall effects.
• T20/T30/EDT reporting: In noisy bands, insisting on T30 produces misleading results. We reported T30 where the decay range supported it, otherwise T20 with clear labeling, plus EDT for perceptual context.
• Unoccupied room: Measurements were taken without an audience, which typically yields longer RT than occupied conditions. Rather than guessing, we used seat absorption estimates for a rough “occupied projection” and recommended a follow-up verification during a rehearsal with at least 60% occupancy if the city wanted high confidence.

6) Results and outcomes with specific details

We processed data the next day and delivered a 9-page case-study style memo with tables and plots. Key findings (source at S1, average across audience positions excluding under-balcony unless noted):

• Midband RT60 (T30) averaged: 1.62 s at 500 Hz, 1.58 s at 1 kHz, 1.46 s at 2 kHz.
• Low-frequency RT: 2.05 s at 125 Hz and 2.28 s at 63 Hz (where measurable), indicating modal buildup and limited LF absorption.
• High-frequency RT: 1.18 s at 4 kHz and 0.98 s at 8 kHz, suggesting some HF loss from air absorption and seat materials, but not enough midband control for speech.
• Under-balcony condition: EDT at 1 kHz under the balcony averaged 1.35 s (shorter than open seating), but clarity complaints persisted there due to reduced early lateral energy and a pronounced 160–250 Hz “warmth” buildup. RT alone did not explain the perceived muffling; the impulse responses showed a strong early reflection cluster off the balcony face at ~25–35 ms.
• Spatial variation: Rear positions near the back wall exhibited a 0.15–0.25 s longer decay around 500 Hz compared to mid-room, consistent with a reflective rear wall and limited diffusion.

We compared the midband RT to a practical target for a multipurpose hall prioritizing speech: approximately 1.0–1.2 s occupied (500 Hz–2 kHz). Even allowing for audience absorption, the room was likely to remain above target unless additional midband absorption was added.

The project manager needed specifics for budgeting, so we translated the results into actionable implications:

• Any PA upgrade would need more direct-to-reverberant ratio (tight coverage, controlled vertical dispersion) to maintain intelligibility in a 1.5–1.6 s room.
• Acoustic treatment should prioritize 250 Hz–1 kHz absorption on upper side walls and rear wall to reduce midband RT without making the room acoustically “dead” at high frequencies.
• Under-balcony improvements likely needed a combination of balcony soffit absorption and careful loudspeaker aiming/delay fills; RT reduction alone wouldn’t address early reflection timing.

Timeline-wise, we completed field work in one morning, delivered preliminary plots within 48 hours, and a finalized memo in five business days after incorporating the integrator’s questions about “single-number” reporting (we included an average of 500 Hz and 1 kHz for quick reference, clearly labeled as an average, not a universal truth).

7) Lessons learned and what could be done differently

A few takeaways stood out from the process:

• Background noise dictates what you can claim at low frequencies. We could measure trends at 63–125 Hz, but confidence was lower under the balcony. Next time, we’d coordinate a brief HVAC “quiet window” (even 20 minutes) if the building can allow it, specifically to improve LF decay range.
• RT60 doesn’t capture everything listeners complain about. The under-balcony area showed shorter EDT but still sounded muddy. Including C50/C80 or STI estimates (with appropriate caveats) would have strengthened the story. We had the impulse responses to compute them; we simply didn’t scope it initially.
• Room setup matters more than people expect. Portable seating and stage curtains were partially deployed during measurement. We documented it, but a more controlled baseline (curtains in a known position, consistent seating layout) would reduce ambiguity when comparing “before vs. after.”
• Repeatability checks pay off. Those duplicate measurements gave stakeholders confidence and helped us identify one corrupted capture immediately.

If we were rerunning this survey with the same time window, we’d add two more mic positions near side walls to better characterize lateral reflection differences, and we’d explicitly compute EDT, T20/T30, and C50 at minimum.

8) Takeaways applicable to other projects

For audio engineers and project managers planning their own RT60 survey, the core lessons from this hall translate well:

• Define the decision you’re supporting. If the goal is treatment budgeting, you need banded RT values and spatial variation. If the goal is commissioning verification, you need repeatable positions and a clearly defined room configuration.
• Use an impulse-response method when time is tight. ESS with a dodeca source allows fast, repeatable captures and gives you more metrics than RT alone if you decide to extend the analysis later.
• Pick mic and source heights intentionally. 1.2 m mic height for seated audience and ~1.5 m source height for talker/PA proxy avoids skewing early reflections.
• Log background noise and don’t overreport low-frequency confidence. If you can’t achieve enough decay range, label it. A “precise” LF RT number derived from noisy data will mislead system and treatment decisions.
• Measure multiple positions and at least two source locations. One position can be an anecdote. A small grid reveals patterns: rear-wall buildup, balcony effects, and side-wall asymmetry.
• Deliver results in a format stakeholders can use. Provide plots for engineers, a small summary table for project managers, and an interpretation tied to the venue’s program (speech vs. music). Include measurement conditions so the survey is defensible.

An RT60 survey isn’t complicated in theory, but it becomes meaningful only when it’s executed with repeatability, documented context, and an honest view of limitations. In this case, a single morning of disciplined measurements gave the city a baseline, gave the integrator design constraints grounded in reality, and prevented the common mistake of treating RT as a one-number problem with a one-product solution.