How to Take a Sound Measurement Somebody Else Can Repeat
Why this matters
Six months after the reading, somebody will ask what it meant. It might be an inspector, an attorney, a customer, or you, deciding whether a treatment worked. At that point the number on its own is worth nothing, because nobody can reproduce it: not the distance, not the weighting, not what else was running, not whether the meter was reading true that day. A sound measurement is a record with a closed calibration loop around it, and the number is only one field in that record. Filling those fields takes a few minutes at the time and cannot be done afterwards at any price. Leaving them blank costs the whole visit, and you do not find out which visit it was until you need it.
What a record has to answer
Three questions, and every field earns its place by answering one of them.
Was the instrument telling the truth? That is the calibration loop, and it is the only field pair that can void everything else.
What quantity is this? Weighting, bandwidth, time basis, and whether any correction was applied or declined. A level missing any of these is not a measurement.
Where and under what conditions? Position, height, orientation, source operating state, what else was running, weather if outdoors. This is the half that makes the reading repeatable rather than merely labelled.
The fields, and what a blank one costs
Filled in the order you meet them, not in order of importance.
| Field | What a blank one costs |
|---|---|
| Date, start time, sample duration | Nobody can align the reading with a production log, a complaint log or a weather record |
| Who took it | No one to ask about anything the sheet does not say |
| Purpose of the measurement | The reader guesses which criterion it was meant to be judged against, and often guesses wrong |
| Instrument type, class, serial number, date of last laboratory verification | The result carries no defensible uncertainty, and an out-of-date verification is found by the other side, not by you |
| Calibrator class, output level and frequency, date of last laboratory verification | The calibration loop is open; the pre-check proves nothing |
| Pre-check reading | No evidence the instrument was true at the start |
| Frequency weighting | The number could be off by more than 20 dB from what the reader assumes |
| Time weighting or integration mode, and integration period | A fast maximum and an equivalent-continuous level get compared as if they were the same quantity |
| Bandwidth reported: overall, octave bands, or both | An overall figure cannot drive any treatment decision, and the reader cannot tell it was not meant to |
| Measurement range or threshold setting | Noise the applicable standard requires you to capture may have been excluded without leaving a trace |
| Reference point on the source, distance, microphone height, orientation | The reading cannot be repeated and no distance rule can be applied to it |
| Windscreen fitted, wind speed, air temperature | An outdoor reading with unrecorded wind is not defensible |
| Source operating state: staging, load, speed, cycle | Two readings of the same source taken at different loads get compared as if the source had not changed |
| Everything else running | The background is not reproducible, so no correction can be verified |
| Background reading, taken with the source off and nothing else changed | No attribution is possible; you have a total, not a source |
| Margin and whether a correction was applied or declined | The reader cannot check the arithmetic, including you |
| Post-check reading and drift | The set is unverified and can be challenged in full |
| Photograph of the microphone position | The only field that reliably answers a question the sheet did not anticipate |
The calibration loop, and why drift is not corrected
Check the instrument against an acoustic calibrator at the start of the set and again at the end, with the same calibrator, and record both readings rather than an adjustment. This is the field that makes a reading into data. 29 CFR 1910.95(d)(2)(ii) requires that instruments used to measure employee noise exposure be calibrated to ensure measurement accuracy, and a pre-check with no post-check does not establish that the instrument stayed accurate through the readings that matter.
Set an acceptance and apply it. A widely used field acceptance is 0.5 dB of drift between the pre-check and post-check. The test method, ordinance or specification your result will be judged under states its own acceptance, and that one governs.
Drift beyond acceptance voids the set. It is not corrected. The temptation is to split the difference and adjust the readings, and it is wrong: drift tells you the instrument moved, not when it moved or by how much at any given reading. Everything between the two checks is repeated.
Field calibration is not laboratory verification. The calibrator check confirms the instrument still agrees with a known reference at one frequency. It says nothing about the instrument's response at 63 Hz or at 8 kHz, which is what its periodic laboratory verification covers, and which is why the verification date is its own field.
The measurement sequence
- Write the purpose and the basis before the meter comes out. Purpose decides the criterion; the criterion decides the weighting, bandwidth and time basis. Choosing them afterwards to fit the number you got is the one failure this whole procedure exists to prevent.
- Set the range or threshold to cover what the applicable standard requires. For employee exposure work, 29 CFR 1910.95(d)(2)(i) requires that sampling include all continuous, intermittent and impulsive noise in the 80 dB to 130 dB range, so a meter set to start at 90 dBA has excluded noise the standard requires you to capture.
- Do the pre-check and record the displayed reading, not "OK".
- Set the position, and record it as you set it. Distance from a named reference point on the source, height above the floor or grade, microphone orientation per the incidence the instrument is specified for. Photograph it.
- Confirm the source state with whoever owns it, and write the state down. If it will change during the sample, either wait or record the change.
- Take the sample for the full stated period. A 15-minute equivalent-continuous level abandoned at 4 minutes is a 4-minute level.
- Take the background with the source off and nothing else altered, on the identical basis and at the identical position.
- Compute the margin and apply or decline the correction using the sibling card's zone table, and write down which you did.
- Do the post-check, record it, and compute drift against the acceptance.
- Write the result as a sentence carrying its full basis, not as a bare number.
The completed record
Purpose: screening to decide whether 29 CFR 1910.95(d)(1) monitoring is triggered at the press operator position.
Date and time: Tuesday, 10:05 to 10:20 local. Sample duration: 15 minutes. Taken by: shop lead.
Instrument: integrating sound level meter, Class 2, serial recorded on the sheet, last laboratory verification within the manufacturer's stated interval. Calibrator: acoustic calibrator, 94.0 dB at 1 kHz, laboratory verification current.
Pre-check: 94.0 dB displayed. Post-check: 93.8 dB displayed. Drift: 0.2 dB against a 0.5 dB acceptance, so the set stands. Had it read 93.3, drift would be 0.7 dB, the set would be void, and the whole visit would be repeated rather than adjusted.
Basis: A-weighted, overall, equivalent-continuous over 15 minutes (LAeq,15min) re 20 micropascals, with a C-weighted overall logged in parallel over the same period. Range: set to cover 80 dB to 130 dBA per 1910.95(d)(2)(i).
Position: microphone on a stand at the operator's normal head position, 1.55 m above the floor, operator stepped clear, orientation per the incidence the instrument is specified for. Stand set outside the press's operating envelope and clear of the point of operation, which stays guarded per 29 CFR 1910.212(a)(3)(ii); nothing was placed inside the guard and no guard was moved. Windscreen fitted. Photograph taken.
Source state: press running its normal production cycle at its shift rate, on the same job it ran all morning. Everything else running: two extraction fans on, welding bay idle, no forklift traffic during the sample.
Readings:
- Press running: LAeq,15min = 88.4 dB, LCeq,15min = 91.0 dB
- Press stopped, everything else identical: LAeq,15min = 79.4 dB
C minus A: 91.0 minus 88.4 = 2.6 dB, which says the content is mid-frequency dominated and raises no low-frequency flag. That screen belongs to the weighting card and is applied here as it states it.
Margin and correction, printed as its own line: 88.4 minus 79.4 = 9.0 dB, inside the zone where a background correction applies, and the correction at a 9.0 dB margin is 0.6 dB on the sibling card's table. Applied, the press's own contribution at that position is 87.8 dB LAeq,15min.
Which figure answers which question, because they are not the same figure. The exposure question is answered by the total at the operator position, 88.4 dB, because that is what reaches the operator's ear regardless of which machine made it. The attribution question, how much of that is the press, is answered by the corrected 87.8 dB. Reporting only the corrected figure as the exposure would understate it.
Forward check: recombining 87.8 with 79.4 by energy returns 88.4, which is the reading that was taken. The correction closes.
Result, written as a sentence: at the press operator position, 1.55 m above the floor, with the press running its normal production cycle and two extraction fans on, the A-weighted equivalent-continuous sound pressure level over 15 minutes was 88.4 dB re 20 micropascals, instrument drift 0.2 dB, background 79.4 dB on the same basis.
What follows from it, stated with its gate: this is a 15-minute screening sample and not an 8-hour time-weighted average, so it does not by itself establish an exposure. It is exactly the kind of information that indicates exposure may equal or exceed an 8-hour TWA of 85 dBA, which is what 29 CFR 1910.95(d)(1) makes the trigger for a monitoring program, so full-shift dosimetry is the next step rather than a conclusion drawn from this sheet. Hearing protection is worn in that bay by anyone present because you must raise your voice to be heard at arm's length there, and it changes what a person hears, not what the microphone on the stand reads.
References
- 29 CFR 1910.95, Occupational noise exposure, including the monitoring trigger at (d)(1), the sampling range at (d)(2)(i) and the calibration requirement at (d)(2)(ii)
- 29 CFR 1910.212(a)(3)(ii) for point-of-operation guarding where a microphone stand is set near a machine
- The instrument and calibrator manufacturer's documentation for verification intervals and specified microphone incidence
- The test method, ordinance or specification the result will be judged under, in the edition it names, which owns its own drift acceptance and correction rules
- See related: How to Add and Subtract Noise Levels Correctly; What a Sound Level Meter Class Actually Buys You