Fibson homePlan my room
Menu

How to measure your room's reverberation time with a phone

Measure reverberation time with a phone app: T20 or T30, the test sound, background noise and ISO 3382-2 positions, and how far to trust the result.

A tall felt organ pipe standing upright, with a saffron pocket watch hanging from its mouth on a thin charcoal chain.
Reader question
Can we measure RT60 ourselves with a smartphone app, and how far can we trust the number?

Yes, you can measure a room’s reverberation time with a phone, and the number is good enough to compare a room before and after treatment, if you measure it the way the standard describes. Use a loud test sound, preferably pink noise from a powered speaker switched on and off, read T20 in octave bands, measure at two or more phone positions in a quiet, empty room, and average. Treat the result as a survey-level estimate: ISO 3382-2 assumes accuracy better than 10 percent in octave bands for a survey made with proper instruments, and a phone adds uncertainty from its microphone and its source.

The app listings that rank for this question explain their buttons but not the method. This article supplies it: what an app actually measures, how to pick between T20 and T30, how to make the test sound, where to stand, how far to trust the reading, and how to compare it with a Sabine estimate from the reverberation time calculator. What value your room should reach is a separate question, answered in what reverberation time your room should have.

What reverberation time is, and what a phone app measures

Reverberation time is the time a sound takes to fade by 60 decibels after its source stops. ISO 3382-2, the international method for ordinary rooms such as offices, restaurants and homes, defines it in exactly those terms, as the decay of the average sound energy in the room after the source is switched off.

The 60 decibels go back to the first person to measure it. In 1895 Harvard asked Wallace Clement Sabine to fix the lecture room of its new Fogg Art Museum, and he did it by timing how long the note of an organ pipe stayed audible after the pipe was shut off. In his Collected Papers on Acoustics he defines reverberation with an initial intensity of “one million times minimum audible intensity.” A millionfold drop in intensity is 60 decibels.

His instruments were an organ pipe an octave above middle C, an electro-pneumatic valve and a chronograph that could time intervals from half a second to ten seconds. The observer’s only duty, he wrote, was to make the record “when the sound ceased to be audible.” In the empty lecture room the sound lasted 5.62 seconds; with cushions borrowed from Harvard’s Sanders Theatre covering all 436 seats, it lasted 2.03 seconds.

Sabine also showed why one reading is not enough. He repeated the empty-room measurement every second or third night for two months and reported a mean of 5.57 seconds, a figure he obtained again one and three years later. Repeat readings and an average are still the core of the method.

A phone app does the same job with a microphone instead of an ear. It plays or listens for a test sound, splits what it hears into octave bands, and tracks how fast the level falls in each band once the sound stops. The decay is rarely measured all the way down to 60 decibels below the start, because ordinary rooms are too noisy for that. Instead, the app fits a straight line to part of the decay and extends it.

Laboratory tests work this way too. Intertek’s summary of the ASTM C423 lab test describes measuring “the time it takes for the sound level to drop 25 decibels,” repeated 10 times at each of five microphone locations. The method is sound; what changes with a phone is the quality of the source, the microphone and the room’s background noise.

T20 or T30: which number to read

Read T20 first. It is the reverberation time worked out from the first 20 decibels of usable decay, and it is the range the standard prefers for ordinary rooms.

The definitions are precise. Under clause 3.5 of ISO 3382-2, T20 is taken from the stretch of the decay between 5 and 25 dB below the starting level, and T30 from 5 to 35 dB below. In both cases the line fitted to that stretch is extended to a full 60 dB drop, and the first 5 dB after the sound stops is left out of the fit.

How T20 and T30 are read from a decay curveA fictional decay curve. The sound level is steady until the source is switched off at 0.2 seconds, then falls along a slightly ragged straight line until it reaches the background noise 45 dB below the start, at 1.7 seconds. The T20 range runs from 5 to 25 dB below the start; the T30 range runs from 5 to 35 dB below. A dashed line fitted to the decay and extended to 60 dB below the start reaches it 2.0 seconds after the source stops. Test sound switched off T20 range: 5 to 25 dB down T30 range extends to 35 dB down Background noise, 45 dB down Fit extended to 60 dB down 2.0 s after switch-off 0 20 40 60 0 0.5 1.0 1.5 2.0 2.5 3.0 Time (seconds) dB below starting level
Fictional decay drawn for illustration: a room with a 2.0 s reverberation time and background noise 45 dB below the starting level. Definitions of T20 and T30 from ISO 3382-2:2008, clause 3.5, checked November 2026.

The introduction to ISO 3382-2 gives three reasons for preferring the 20 dB range. The way people judge reverberation is related to the early part of the decay; the early decay is the right part for estimating a room’s steady sound level; and in field measurements “it is often difficult or impossible to get a evaluation range of more than 20 dB.”

That last reason decides most phone measurements. The standard adds that a 20 dB evaluation range “requires a signal-to-noise level of at least 35 dB.” One way to read the figure: the 5 dB skipped at the start, the 20 dB used, and a further margin that keeps the end of the fit clear of the background noise. A T30 uses 10 dB more of the decay, so it needs roughly 45 dB.

When T20 and T30 disagree

If the decay is a straight line, both ranges see the same slope, so T20 and T30 agree. When they differ noticeably, the decay is bent. The standard’s Annex B adds two measures, one for non-linearity and one for curvature, so that a bent decay can be flagged as less reliable.

A bent decay often has a physical cause you can find. A likely one is a room open to another space that decays at a different rate: a dining room open to a tall bar area, an office open to a stairwell, a living room open to a double-height hall. If your T30 is clearly longer than your T20, look for the second space before you blame the app.

Choosing the test sound: interrupted noise or an impulse

ISO 3382-2 accepts two ways of producing a decay, and phone apps use one or the other. The interrupted noise method records the decay after broadband noise is switched off. The integrated impulse response method records the room’s response to a very short sound and integrates it backward in time to recover the decay.

For the impulse method, the standard’s definitions note that true impulses cannot be produced, but that short transient sounds, such as a gunshot, can come close enough. They also allow an alternative: play a known signal such as a sine sweep or a maximum-length sequence, then convert the recording into an impulse response. An app that plays a sweep through a speaker and computes the decay from the recording is taking this route.

For interrupted noise, one long-running iOS app is a good example of how the method is meant to work. Its App Store listing says it pulses pink noise on and off, filters the incoming signal into octave bands from 63 Hz to 4 kHz, and times the decay in each band. It also tells the user to connect the phone’s output to an amplifier and speaker rather than rely on the phone itself.

Here is how the common test sounds compare. The judgment column reflects the standard’s two requirements: enough level to clear the background noise in every band, and no overload anywhere in the chain.

Test sound How it is made Strengths Watch for
Interrupted pink noise App plays noise through a powered speaker, then cuts it Defined in the standard; easy to repeat and to average several decays Needs a speaker loud enough to clear background noise in every band
Sine sweep App plays a sweep through a speaker and computes the impulse response One run covers every band; the app does the conversion Needs an app that supports it; the speaker still has to reach every band
Balloon burst A large balloon popped at the source position No speaker needed; loud Hard to repeat exactly; can overload a phone held close
Hand clap or wood block Struck at the source position Always available May be too weak to give a full decay in every band

Two rules from the standard apply to all of them. Under clause 4.2.1 the source must be loud enough to produce the required decay range without the background noise contaminating it, and under 4.2.2.3 no overload is allowed at any stage of the measuring equipment. For a phone, overload means holding it too close to the balloon or the speaker. Stand back at least a few feet and check that the app’s input meter does not hit its ceiling.

Protect your hearing while you do this. Pink noise loud enough to clear a busy room’s background is unpleasant at close range, so start the noise, step away and keep the run short.

Background noise sets the limit

Background noise is the limit most phone measurements hit first. The decay can only be read where it stands clear of the room’s own noise: refrigeration, ventilation, street traffic, an ice machine, the hum of a drinks cooler.

The numbers to aim for come from the standard and from the app makers. ISO 3382-2 asks for at least 35 dB between the test sound and the background for a T20. The interrupted noise app quoted above asks for “at least 25dB of headroom between the noise floor of the environment, and the level of the pink noise for T20 analysis.” Aim for the standard’s figure; the app’s figure is the minimum it needs to show a result at all.

The listing is candid about what happens when the room wins: “you may not be able to overcome the ambient noise level in the room to get a good reading in every band.” Check each band separately; a reading can be valid at 1,000 Hz and missing or unreliable at 125 Hz.

Here is how those figures play out in one hypothetical room. Suppose the app shows the test noise 40 dB above the background at 500 Hz and only 28 dB above it at 125 Hz:

  • 500 Hz, T20: 40 dB clears the standard’s 35 dB, so the reading is usable.
  • 500 Hz, T30: a 30 dB range needs roughly 45 dB, so treat the T30 as marginal.
  • 125 Hz, either range: 28 dB is above the app’s 25 dB floor, so it will show a number, but it is below the standard’s 35 dB. Report it, and mark it as unreliable.

How to buy more headroom

  • Measure out of hours. An empty restaurant early in the morning is usually quieter than mid-afternoon, with fewer deliveries and less kitchen equipment running.
  • Switch off what you can. Music, fans, portable heaters, dishwashers and ice machines. Note anything you could not switch off.
  • Close doors and windows, unless the room is normally used open to the next space; then note that too.
  • Turn the source up and leave the phone where it is. A louder speaker adds headroom; moving the phone closer to the speaker breaks the minimum-distance rule described below.

Measure the room empty

People absorb sound, so a full room rings less than an empty one. Clause 4.1 of ISO 3382-2 says reverberation time should be measured with the room unoccupied, and treats a room as unoccupied when no more than two people are present. Measure empty, with the furniture in place, and write down what was there.

For most rooms you can ignore the air itself. The standard says the effect of air absorption is small, and temperature and humidity need not be measured, when the reverberation time is under 1.5 seconds at 2,000 Hz and under 0.8 seconds at 4,000 Hz. In a large, hard room that rings longer than that, note the temperature and humidity with your readings.

Where to put the source and the phone

One position is not a measurement. Reverberation time varies across a room, and the standard sets minimum numbers of positions for each of its three accuracy levels. Clause 4.3, Table 1, gives these minimums:

Level Source and phone combinations Source positions Phone positions Decays per position (interrupted noise) Nominal accuracy
Survey 2 1 or more 2 or more 1 Better than 10% in octave bands
Engineering 6 2 or more 2 or more 2 Better than 5% in octave bands, 10% in one-third octaves
Precision 12 2 or more 3 or more 3 Better than 2.5% in octave bands, 5% in one-third octaves

The survey level is the one written for a phone-sized job. The standard describes it as suitable for assessing a room’s absorption for noise control, measured in octave bands only. The engineering level is the one meant for checking a building against a reverberation specification, and the standard asks for more positions than the minimum in rooms with complex shapes. The published sample of clause 4 is the French edition; the figures above are translated from it.

Spacing rules, in feet

The standard’s spacing rules are short and practical:

  • Phone positions at least 2 m apart, about 6.5 ft, which is half a wavelength at the frequencies usually measured.
  • At least 1 m from any surface, about 3.3 ft, including the floor. Hold the phone at chest height in open floor, not against a wall or on a table.
  • No symmetrical positions. Do not mirror the phone positions across the middle of the room.
  • Sources where the room is normally used. A podium, the head of a table, the bar. In small rooms such as homes, or where there is no obvious position, put one source position in a corner.
  • Not too close to the source. The minimum distance is 2 × √(V ÷ (c × T)), where V is the volume, c the speed of sound and T the reverberation time you expect.

The last rule keeps the direct sound from swamping the decay, and it works in any consistent units. For a room of 13,200 cu ft expected to ring for about 2.3 seconds, with sound traveling 1,125 ft per second, it gives 2 × √(13,200 ÷ 2,587.5) = 2 × 2.26 = about 4.5 ft.

Source and phone positions in a 40 by 30 foot dining roomPlan of a fictional 40 by 30 foot dining room with storefront glass on the left wall and a bar on the right wall. Source S1 stands near the bar and source S2 near a window table, each with a dashed circle 4.5 feet in radius around it. Phone positions M1, M2 and M3 are spread unevenly across the room, all at least 3.3 feet from the walls, at least 6.5 feet from each other and outside both circles. Two sources and three phone positions make six combinations. S1 S2 M1 M2 M3 3.3 ft from walls 40 ft 30 ft, storefront glass Bar Source (speaker) Phone position 4.5 ft from source 2 sources × 3 phone positions = 6 combinations
Fictional example: positions sketched for the example dining room, which is not a Fibson project. Spacing rules from ISO 3382-2:2008, clause 4.3.1; the 4.5 ft minimum source distance comes from its Equation (1) with a volume of 13,200 cu ft and an expected reverberation time of 2.3 s.

Two sources and three phone positions give six combinations, the engineering-level count, for very little extra effort. If you have only one speaker, move it between runs. The standard also allows taking a new phone position for each decay, as long as the total number of decays is reached.

Step by step: a survey measurement with a phone

The procedure below follows the survey level of ISO 3382-2, adapted to the equipment most people have: a phone, a measurement app and a powered speaker. Allow an hour or so for a restaurant dining room or an office floor.

  1. Pick the app and the method. Use an app that reports T20, and T30 where it can, in octave bands from at least 125 Hz to 4,000 Hz. Note its name and version.
  2. Set up the source. A powered speaker on a stand at mouth height, connected to the phone or to a second device, playing the app’s noise or sweep. If you only have a balloon, use the largest you can and pop it at the source position every time.
  3. Quiet the room. Switch off music and fans, close doors and windows, and write down anything still running. Clear the room to two people or fewer.
  4. Mark the positions. Two source positions and three phone positions, following the spacing rules, marked with tape on the floor so you can return to them after treatment.
  5. Check headroom before measuring. With the test sound on, the app should show at least 35 dB above the background in each band you care about. If a band falls short, turn up the source or accept that band as unreliable.
  6. Measure each combination. At each phone position, take at least one decay per source position, and two if you have time. Hold the phone at chest height, screen up, away from your body, and stay still and silent.
  7. Record every reading, not only the average: band by band, position by position, T20 and T30 where shown. Screenshots help.
  8. Average band by band. Average the T20 values at each octave band across all combinations. Do not average across bands except to report a single mid-frequency figure, and say which bands it covers.
  9. Note the conditions: date and time, furnishing, what was running, the source used and the app’s settings.

One extra habit pays off. Before the real run, take a single test decay at the first position and look at the bands. If 125 Hz or 250 Hz shows no value, the source is not loud enough there, and no amount of repetition will fix it.

Which bands to report

Report every octave band you measured, normally 125, 250, 500, 1,000, 2,000 and 4,000 Hz. A room can be fine at 2,000 Hz and still ring at 250 Hz, and the panels that fix one band do less for the other.

If someone asks for a single number, give it with its bands attached. Rating systems differ here: the reverberation table in the LEED v4.1 acoustic performance credit, for example, is stated “at 500 Hz, 1000 Hz, and 2000 Hz,” while the estimates on this site use the 500 to 1,000 Hz range. A “mid-frequency RT60” without its bands cannot be compared with anything.

Common problems and what they mean

Most odd readings have ordinary causes. The table below lists the ones a phone measurement runs into most, with the fix for each.

What you see Likely cause What to do
No value in the 125 or 250 Hz band Not enough level above the background in that band Turn up the source, quiet the room, or report the band as missing
T30 clearly longer than T20 A bent decay: a second space, or background noise reaching the end of the fit Look for open doorways and voids; check headroom for T30
One position far from the others Phone too close to the source, a wall or a corner, or someone moving nearby Check the spacing rules and measure that position again
Readings jump between runs at one position A noise event during the decay: a truck, a compressor starting, a door Discard runs with obvious interruptions and average several decays
Readings shorter than the room sounds People in the room, or doors open to a larger soft space Measure empty, with doors as they are normally used, and note both
The input meter hits its ceiling Overload, which the standard does not allow at any stage Move the phone back or lower the source level
Values close to the app’s stated lower limit The instrument’s own averaging limits what it can time Treat them with caution; the listing quoted here gives 50 ms as its limit

Write down every problem you saw next to the readings it affected. A reading marked “truck passed during run 2” is still useful; an unexplained outlier quietly averaged in is not.

How far to trust the number

Trust a phone reading most in the middle bands, as a comparison between two dates in the same room, and least as an absolute figure at 125 Hz. That ranking follows from what the evidence says about phones, and from what the standards say about rooms.

What the research says about phone microphones

Most published testing of phone apps concerns sound level, not decay. In a 2014 study, Kardous and Shaw at NIOSH evaluated more than 130 iOS apps and found only 10 that met their selection criteria. Two of those had mean differences of 0.07 dB (unweighted) and minus 0.52 dB (A-weighted) from reference values, and two more were within plus or minus 2 dB. Their 2016 follow-up found that external calibrated microphones brought the same apps within 1 dB of the reference.

Reverberation time depends on the slope of the decay, not its absolute level. A constant calibration error shifts the whole curve up or down and leaves the slope unchanged, which is why an uncalibrated phone can still time a decay. What matters more is the microphone’s own noise floor and how evenly it responds across the bands. The app quoted above applies “a compensation network” to the built-in microphone’s response and lets you calibrate against a sound level meter.

For decay measurements specifically, one study is directly useful. Spratford, Walker and McCreery used an iOS device, an app and an external microphone to measure 164 classrooms. They concluded: “As a screening tool, the application was shown to be effective in reliably measuring reverberation and classroom noise levels.” Screening is the right word for most phone work.

What the standards say about real rooms

Real rooms do not behave like test chambers. The significance section of ASTM C423 warns that when measurements are made in a building “in which the size and shape of the room are not under the operator’s control, the approximation to a diffuse sound field is not likely to be very close.” That is one reason the standard asks for several positions, and for more of them in rooms with complex shapes.

The standard’s own accuracy figures set the scale. A survey with compliant instruments is assumed to be better than 10 percent in octave bands. Differences smaller than that between two of your readings, or between your reading and an estimate, tell you very little.

The instrument’s own speed

A decay can only be timed if the meter reacts faster than the sound fades. Clause 4.2.2.2 of ISO 3382-2 sets the averaging time of an exponential-averaging meter at less than one thirtieth of the reverberation time being measured, and of a linear-averaging meter at less than one twelfth. For a well-treated meeting room ringing for 0.4 seconds, that means averaging in under about 13 milliseconds.

Few app listings say how their averaging works. The interrupted noise app quoted above gives a range of 50 milliseconds to 4 seconds, which tells you where its makers think its limits are. In a short, dead room, readings near the bottom of any app’s range deserve the least trust.

A practical ranking

Use of the reading How much to trust it
Before and after treatment, same phone, same positions, 500 to 2,000 Hz High: most errors repeat and cancel out
Absolute value at 500 to 2,000 Hz, averaged over 4 to 6 combinations Moderate: a survey-level estimate
Comparing your room with someone else’s measurement Low: different equipment and methods
Absolute value at 125 Hz from a phone and a small source Low: often limited by headroom
Proof of compliance with a specification or rating Not suitable on its own

The last row deserves a note. A field method meant for other purposes does not stretch to compliance either: ASTM E2235, an ASTM standard for measuring decay rates in sound insulation tests, says it “shall not be used” when decay rate is to satisfy a criterion directly, giving as the example a room that must not be so reverberant that speech becomes hard to understand. ISO 3382-2 assigns that job to its engineering level.

Comparing a reading with a Sabine estimate

A measured value and a calculated estimate check each other. The estimate comes from the room’s size and finishes; the measurement comes from the room itself. When they agree within the accuracy of the measurement, the estimate can be trusted to predict the effect of panels.

Sabine’s formula in feet is RT60 = 0.049 × V ÷ A, where V is the volume in cubic feet and A is the total absorption in sabins. The HyperPhysics reference at Georgia State University gives it as 0.049 seconds per foot for dimensions in feet and 0.16 per meter in metric. Turned around, a measured reverberation time tells you how much absorption the room really has: A = 0.049 × V ÷ RT60.

Example (fictional). The example room is a 60-seat dining room, 40 by 30 ft with an 11 ft ceiling, so its volume is 13,200 cu ft and the top line of the formula is 0.049 × 13,200 = 646.8. It is not a Fibson project. Its surface inventory, unoccupied and furnished, totals 276.4 sabins at mid frequencies, so the estimate is 646.8 ÷ 276.4 = 2.34 s.

Suppose the owner measures it at the six combinations in the plan above and gets these T20 readings, in seconds:

Combination 500 Hz 1,000 Hz
S1 to M1 2.2 2.0
S1 to M2 2.1 2.0
S1 to M3 2.3 2.1
S2 to M1 2.1 1.9
S2 to M2 2.2 2.0
S2 to M3 2.2 2.1
Average 2.18 2.02

The mid-frequency average is (2.18 + 2.02) ÷ 2 = 2.10 s, about 10 percent below the estimate. Working backward, the room holds 646.8 ÷ 2.10 = 308.0 sabins, 31.6 more than the inventory counted, perhaps upholstered bar stools or a curtain left off the list. The panel plan adds 386.4 sabins, so the corrected forecast after treatment is 646.8 ÷ (308.0 + 386.4) = 646.8 ÷ 694.4 = 0.93 s, against 0.98 s from the inventory alone.

A 10 percent gap is about the size of survey-level uncertainty, so in this case the measurement mainly confirms the estimate. The correction still helps: it tells the planner the room starts slightly better than the inventory suggests, and the forecast moves with it.

Gaps much larger than 10 percent mean something is missing. Common causes include an open doorway or stair into another space, a surface recorded as the wrong material, heavy curtains or upholstered booths left off the list, and a measurement taken with people in the room. Find the cause before trusting either number.

To run the estimate side yourself, enter the room in the reverberation time calculator, which reports both the Sabine and Eyring results. The panel quantity calculator then works from a starting value to the absorption needed for a target.

Measuring before and after treatment

A before-and-after pair is the most reliable thing a phone can give you, because most of its errors repeat. The trick is to change nothing but the panels.

  • Same positions. Leave the tape marks down, or photograph them with a tape measure in frame.
  • Same source and settings. The same speaker, volume, app, version and evaluation range.
  • Same room state. The same furniture, the same doors open or closed, the same equipment running, and the room unoccupied both times.
  • All the panels. Measure after every panel is up, not halfway, unless you want a reading per stage.
  • Band by band. Compare each octave band separately. Panels change the middle and upper bands most, and a change of a few hundredths of a second at 125 Hz is within the noise of the method.

Expect the measured change to differ somewhat from the forecast. Lab coefficients come from test specimens in a reverberation room, and ASTM C423 itself says they “should be used with caution” because real areas are larger and real rooms are rarely diffuse. If the after reading lands within about 10 percent of the forecast, the plan did what it was designed to do.

Some programs ask for a measured result rather than an estimate. The legacy WELL v1 reverberation feature, for example, sets maximums of 0.6 seconds for conference rooms and 0.5 seconds for open offices and lists a performance test as its verification method. Where a rating or a contract depends on the number, have it measured at the engineering level by someone equipped to do so; a phone reading is the screening step before that.

What to send with a room plan request

A phone measurement does not replace the room’s dimensions and finishes; it adds a check on them. If you have one, send it with the rest of the room details described in what you send and what you receive.

The reason is simple. A measured reverberation time tells you how much absorption the room has in total, but not where it is. A panel plan has to know which surfaces carry the problem, a hard ceiling over the tables or a long glass wall, and only the surface inventory and photos show that.

Send these, in this order of usefulness:

  1. The band-by-band readings at each position, T20 and T30 where shown, from 125 Hz to 4,000 Hz. Averages alone hide the spread.
  2. A sketch of the positions with rough distances, like the plan above. The room measurement worksheet has space for the room’s dimensions; add the positions to the sketch.
  3. The method: app name and version, interrupted noise or impulse, the source used and whether the app showed enough headroom in each band.
  4. The room state: date and time, furnished or not, people present, doors and windows, and anything that could not be switched off.
  5. Screenshots of the app’s result screens.

A room acoustic assessment starts from the surface inventory; a reading like yours gives that estimate something to be checked against, as in the Sabine comparison above. Fibson designs and makes acoustic panels to order and plans each room before anything is ordered. If you would like the estimate, the target and the panel plan worked out for your room, send us the room with your readings attached to the description.

Working on a room like this?

Send the size, a few photos and what happens in the space. We reply with questions or an outline of the panel plan.

Plan my room