IEC 61012: Measuring Audible Sound in Ultrasonic Environments with U-Weighting Filters

IEC 61012:1990 — Filters for the measurement of audible sound in the presence of ultrasound / Reading time: ~7 minutes

Walk into any modern factory and you will encounter ultrasonic equipment everywhere — ultrasonic cleaning baths, plastic welding machines, high-speed dental drills, food cutting systems. These devices pump intense acoustic energy into the air at frequencies above 20 kHz. The problem? The audible noise they also produce — the hiss of cavitation, the whine of bearings — is what matters for worker hearing protection, and your sound level meter may be lying to you about how loud it really is. IEC 61012 exists to fix exactly this: it defines the U-weighting filter, a precision low-pass filter that strips away ultrasonic contamination so that you can measure audible sound truthfully.

1. How Ultrasonic Energy Corrupts Audible Sound Measurements

1.1 The Frequency Response Gap in IEC 651 Sound Level Meters

Standard sound level meters built to IEC 651 (now superseded by IEC 61672, but the principle holds) specify frequency response characteristics only up to 20,000 Hz. This was not an oversight — when the standard was drafted, ultrasonic industrial noise was a niche concern. Above 20 kHz, there is no standardized roll-off, no prescribed tolerance, and no guarantee whatsoever about how a given instrument will behave.

Yet ultrasonic sources in industry are pervasive, as Table 1 illustrates. When the measurement microphone’s frequency response extends into the ultrasonic range — a desirable trait for broadband acoustic analysis, but a liability for audible-only measurements — the ultrasonic energy is faithfully transduced into the electrical domain, passes through the preamplifier, and adds to the RMS-integrated reading. The result is an inflated dBA or dB SPL value that does not represent the sound a human ear actually perceives.

1.2 The Physics of Ultrasonic Contamination

The mechanism is straightforward in signal-processing terms. A measurement microphone produces a voltage V_mic(f) proportional to the acoustic pressure at frequency f. The sound level meter applies a weighting filter H_A(f) and integrates the squared magnitude over its bandwidth to compute the displayed sound pressure level:

L_pA = 20 log₁₀ [ ∫ |V_mic(f) · H_A(f)|² df ]^1/2 / p_ref

When f > 20 kHz and the microphone still has significant sensitivity, the ultrasonic contribution to the integral inflates L_pA. The error is not a simple additive constant — it depends on the relative spectral distribution of audible and ultrasonic energy, the specific microphone frequency response above 20 kHz, and any nonlinearities (saturation, intermodulation) that strong ultrasonic signals may induce in the preamplifier stage.

2. The U-Weighting Filter: Design Principles and Technical Specifications

2.1 Nomenclature and Combination with A-Weighting

IEC 61012 designates this specialized low-pass filter as “U-weighting.” When cascaded with the A-weighting characteristic of a sound level meter, the combined frequency weighting is called “AU-weighting.” The nominal AU-weighting response (in decibels) is the algebraic sum of the A- and U-weighting relative responses at each frequency.

The U-weighting characteristic is flat (0 dB) across the entire audible band from 10 Hz to 10 kHz, begins a gentle roll-off at 12.5 kHz (-2.8 dB), and then enters a steep attenuation region beyond 16 kHz. Table 2 summarizes the key frequency points:

At the three ultrasonic reference frequencies, the AU-weighting (A + U cascade) delivers staggering rejection: -50.0 dB at 25 kHz, -65.4 dB at 31.5 kHz, and -81.1 dB at 40 kHz. To put this in perspective, a 120 dB ultrasonic field at 40 kHz would contribute less than 39 dB to the AU-weighted reading — buried well below the noise floor of any practical acoustic measurement.

2.2 The Six-Pole Filter Architecture

The U-weighting frequency response is realized by a 6-pole filter whose pole locations are specified in IEC 61012 Table 2. This is not an arbitrary filter design; the pole placement reflects careful optimization for the unique requirements of electroacoustic measurement:

Several design insights emerge from this pole configuration:

• The two high-Q complex pole pairs (Poles 5/6 at f_n ≈ 12.5 kHz, Q ≈ 2.14) create the sharp “knee” in the transition band, enabling rapid attenuation onset while preserving passband flatness up to 10 kHz. With a Q above 2, these poles contribute noticeable peaking that must be carefully managed in real implementations.
• The two real poles at -12,200 Hz provide unconditional far-frequency suppression, preventing any response “bounce-back” above 30 kHz that might otherwise occur in an all-pole filter.
• All poles have negative real parts — guaranteeing BIBO stability regardless of component tolerances.

2.3 AU-Weighting: Double Protection for Occupational Measurements

The AU-weighting provides two layers of ultrasonic defense: A-weighting attenuates the highest audible octaves (matching human ear sensitivity), while U-weighting brutally suppresses everything above 20 kHz. The cascade achieves what neither filter can do alone. For the occupational hygienist, AU-weighted measurements are directly defensible in regulatory contexts because they represent the best available estimate of what a worker’s ear actually perceives in an ultrasonic-noisy environment.

3. Practical Applications and Engineering Methodology

3.1 Ultrasonic Cleaning Bath Noise Assessment

Ultrasonic cleaners are the textbook application for IEC 61012. The cavitation process in the cleaning fluid generates broadband acoustic emission: a fundamental tone at the transducer frequency (typically 20-40 kHz) with harmonics extending upward, plus audible-band noise from bubble collapse, fluid surface agitation, and structural vibration of the tank. The audible component — the “frying bacon” hiss — is what occupational health regulations care about, but it is acoustically buried within a far more intense ultrasonic field.

The recommended measurement chain is:

Measurement Microphone → Preamplifier → U-Weighting Filter → A-Weighting (SLM internal) → RMS Detector → L_pAU Display

Note that the U-weighting filter may be an external accessory or integrated into the sound level meter. External filters require impedance matching: the filter’s nominal input impedance must be compatible with the preamplifier’s output impedance, and its output impedance must drive the SLM’s main amplifier input without loading errors.

3.2 Occupational Noise Compliance: Why U-Weighting Matters Legally

Occupational exposure limits are expressed in dBA or equivalent continuous A-weighted sound level (L_Aeq). When ultrasonic sources are present in the workplace, measurements without U-weighting present three legal vulnerabilities:

1. Over-estimation: Inflated dBA readings may trigger costly engineering controls that are not actually required by the true audible noise level.
2. Under-estimation: Some sound level meters exhibit nonlinear behavior (saturation, intermodulation distortion) when exposed to intense ultrasound, paradoxically depressing the displayed reading below the true audible level.
3. Inter-laboratory non-reproducibility: Different SLM models have different ultrasonic frequency responses, producing irreconcilable results from the same measurement location.

AU-weighting per IEC 61012 resolves these issues by providing a standardized, reproducible basis for audible sound measurement in ultrasonic fields. For regulatory reporting, always document that AU-weighting was used and cite the specific filter model employed.

3.3 Common Measurement Setup Errors

From field experience, these are the most frequent mistakes engineers make when measuring audible sound near ultrasonic sources:

1. Relying on A-weighting alone without a U-filter: A-weighting provides approximately -9.3 dB at 20 kHz per IEC 651, and is unspecified above that. It is wholly inadequate for rejecting strong ultrasonic fields from industrial equipment.
2. Using a wideband measurement microphone without a U-filter: High-quality condenser microphones with flat response to 40-100 kHz are excellent for spectral analysis but are disastrous for audible-only occupational measurements unless paired with U-weighting.
3. Neglecting insertion loss compensation: External U-weighting filters have a specified insertion loss at 1 kHz (stated by the manufacturer per IEC 61012 Clause 4.2.4). Failing to add this loss back to the measured level biases the entire passband downward.
4. Filter input overdrive: Although the U-filter attenuates ultrasound at its output, strong ultrasonic signals still appear at the filter input. If the maximum input voltage rating is exceeded, the input stage clips, generating audible-band harmonics that the filter passes through to the detector.
5. Operating outside rated environmental conditions: The filter’s insertion gain/loss can vary by up to ±0.5 dB over the specified temperature range (-10°C to +50°C per Clause 3.2). In precision measurements, apply the manufacturer’s temperature correction curve.

4. Engineering Insights: Bridging the Standard to Real-World Practice

4.1 The U-Weighting Filter as an Acoustic Anti-Aliasing Filter

From a signal processing perspective, the U-weighting filter is fundamentally an acoustic-domain anti-aliasing filter. In a digital sampling system, the anti-aliasing filter removes frequency components above the Nyquist frequency before the ADC, preventing them from folding back into the baseband as aliases. Similarly, the U-weighting filter removes frequency components above the human auditory system’s “Nyquist frequency” — roughly 20 kHz — before they reach the detector and corrupt the reading. The mathematical analogy is exact, and this insight is valuable when designing digital sound level meters: the analog anti-aliasing filter required before the ADC and the U-weighting filter can share design principles, though they serve different purposes and should not be conflated.

4.2 Measurement Uncertainty Budget for AU-Weighted Measurements

When preparing an uncertainty budget for AU-weighted sound level measurements, the following contributors should be evaluated:

• Filter tolerance: ±1 to ±3 dB depending on frequency (per IEC 61012 Table 1)
• Temperature effect: ±0.5 dB within the specified temperature range
• Humidity effect: ±0.5 dB over 30% to 90% RH
• Electromagnetic field susceptibility: Tested at 80 A/m, 50/60 Hz; orient for maximum indication
• Insertion loss uncertainty: Dependent on source and load impedance matching accuracy
• Quantization noise: Additional contribution in digital filter implementations

Aggregating these contributions using the root-sum-square method (assuming independence), the expanded uncertainty (k=2, approximately 95% confidence) for AU-weighted measurements in typical industrial environments is typically in the range of ±2.5 dB to ±4.0 dB, with the dominant term being the filter tolerance at ultrasonic frequencies. This is a practical reference value for determining whether a measured difference between two noise assessments is statistically significant.