Speech Intelligibility Measurement: RT60, Noise and STIPA

Speech intelligibility is decided by three measurable things: reverberation time, background noise and the ratio of direct to reflected sound at the listener. Measure all three on site, then summarise them with a STIPA reading at several seats. A room that sounds loud but fails these checks needs treatment or speaker changes, not more amplifier power.

This note is written by Singapore-based AV and IPTV integrator Prestige Solutions for AV consultants and hotel engineering teams who commission speech reinforcement in foyers, ballrooms and function rooms with our professional audio service. It gives worked RT60 and critical-distance examples, a pass-band table for STI, and the order of tests we use on site. Numbers are planning values as of 2026; the project specification and the instrument manual remain the final reference.

What does STI actually measure?

The Speech Transmission Index (STI) is a single number from 0 to 1 that describes how well speech modulation survives the path from talker to listener. It is defined in IEC 60268-16 and is the usual way to turn "I cannot understand the announcement" into a figure a contract can reference.

The bands above are the commonly published qualitative labels. The target for a given project comes from the specification or the consultant, and any emergency voice alarm requirement is a separate regulatory matter for the fire protection consultant to confirm; this note covers speech reinforcement and paging only.

How do you estimate RT60 before measuring?

Use the Sabine formula for a first estimate: RT60 = 0.161 x V / A, where V is room volume in cubic metres and A is total absorption in square metres. It is a planning tool for diffuse rooms, so confirm with a measurement, but it shows quickly whether a room is a problem.

The examples are arithmetic, not measurements from any site. They show why a glass-and-stone atrium cannot be rescued by speaker choice alone: at 4.4 seconds each syllable overlaps the next three or four, and raising the level only raises the reverberant field with it.

Why does critical distance matter more than speaker power?

Critical distance is the distance from a speaker at which direct sound and reverberant sound are equal in level. Beyond it, extra volume does not improve clarity. It is estimated as Dc = 0.057 x sqrt(Q x V / RT60), with Q the speaker directivity factor.

Take the 960 cubic metre function room with a speaker of Q = 5. At RT60 1.4 s the critical distance is about 3.3 m; after treatment to 0.7 s it grows to about 4.7 m. Listeners seated beyond those distances are mostly hearing reflections. This is why distributed ceiling speakers, placed so every seat is within or near critical distance of one, outperform a pair of large speakers at the front. The spacing logic for those positions is in the ceiling speaker spacing table for meeting rooms and foyers, and the trade-offs of concentrated arrays are in the line array versus point-source comparison for ballrooms.

How much background noise can speech tolerate?

Background noise lowers STI directly because it fills the gaps between speech modulations. As a planning rule, keep speech at least 15 dB above the ambient noise at the listener; a smaller margin pushes STI down quickly, even when reverberation is well controlled.

Singapore indoor spaces run air conditioning almost continuously, so the noise reading must be taken with the air conditioning in its normal operating state. A test done with the air handlers off will flatter the result and fail when the venue opens.

What is the on-site test order?

Run the tests in the order below. Each step explains the result of the next, so skipping ahead leaves a low STI with no cause attached.

1. Record the room state. Note door and curtain positions, partition settings, HVAC mode and audience simulation. Photograph the layout so the result can be repeated.
2. Measure ambient noise. Use a Class 1 or Class 2 sound level meter, A-weighted, over at least 30 seconds at each test seat, with the system muted.
3. Measure RT60. Use a swept sine or MLS impulse-response method at ear height of about 1.2 m, and report T20 or T30 in octave bands from 125 Hz to 4 kHz, not only a single broadband figure.
4. Set system level. Play pink noise through the speech chain and set the target speech level at the seats, checking that no loudspeaker zone clips before the target is reached.
5. Take STIPA readings. Use the test signal through the real microphone path or a line input, one reading per seat, at least 5 to 9 seats across the audience area including the farthest corner and a seat directly under a ceiling speaker.
6. Repeat with doors and partitions in the alternate setting if the room is used both ways.

A single STIPA reading takes about 15 seconds. Repeat each seat twice; if two readings differ by more than about 0.03, check for movement, noise bursts or a loose connection before trusting either one.

Which fixes follow which result?

The failure pattern tells you where to act. Treat the measurement as a diagnostic table rather than a pass or fail stamp.

Delay and equalisation come last. They polish a system that is already close; they cannot correct a room that fails the three physical tests above. Related control and routing work that sits alongside this tuning is listed on the Prestige Solutions home page.

FAQ

Is one STI reading in the middle of the room enough?

No. A centre seat is usually the best-covered position. A report needs readings across the audience area, including the farthest seat, a seat under a ceiling speaker and any seat near a reflective wall or partition. State the number of seats measured next to the result, since an average hides a failing corner.

Does a higher STIPA score mean the system sounds better?

Not necessarily. STI measures how well speech modulation is preserved, not tonal quality or music reproduction. A system can score well and still sound harsh through poor equalisation. Use STIPA to confirm speech clarity and use listening tests with recorded speech to judge tone and fatigue.

When should an acoustic consultant be brought in?

Bring one in when the estimated RT60 is above about 2 seconds for a speech-critical space, when the room has a curved or domed surface that focuses reflections, or when the venue plans a treatment budget. Speaker and processor changes can only work within the limits the room sets, and a consultant can specify absorption before ceilings are closed.

Recommended next step

Contact Prestige Solutions to review your site drawings and current operating pattern. Call +65 8010 2337 or +65 9825 8218, message us on WhatsApp (+65 8010 2337), or email sales@prestigesolutions.com.sg. You can also browse the full product range before the site walk.