The Low SPL module measures room background noise levels below the self-noise floor of individual microphones by using two-microphone cross-correlation.
Each microphone generates its own electronic self-noise, which is independent and uncorrelated between the two mics. The actual room noise, however, arrives at both microphones as a correlated signal. By computing the cross-power between the two channels and averaging over time, the uncorrelated self-noise cancels out, revealing the true acoustic noise floor beneath it.
This technique can measure noise levels 10–15 dB below the noise floor of the individual microphones — making it possible to characterize extremely quiet rooms with measurement-grade microphones.
Three traces are shown across the octave or third-octave bands:
Red — Mic 1: Individual SPL measured by microphone 1 (includes mic self-noise)
Blue — Mic 2: Individual SPL measured by microphone 2 (includes mic self-noise)
Green — XCorr: Cross-correlated result — the true room noise with self-noise rejected. This should read lower than either individual microphone in bands where self-noise dominates.
Tap or drag on the graph to place a cursor and read per-band dB values. Pinch to zoom the dB scale, or use a two-finger drag to shift the range up or down.
Use the Octave / 1/3 Octave control to switch between 10-band octave resolution and 31-band third-octave resolution. Third-octave mode provides finer spectral detail and always uses the CPSD algorithm (since the IIR filter bank only covers octave bands). Noise rating curves (NC, NCB, NR, PNC, RC) are only available in octave mode, as these standards are defined at octave-band center frequencies.
Two processing methods are available, selectable in Settings:
Computes the FFT of both channels using a 16384-point Hann-windowed transform (~2.9 Hz resolution at 48 kHz). The complex cross-spectrum Sxy = conj(X) · Y is computed for each frequency bin and averaged across multiple overlapping segments. Because mic self-noise is phase-random between the two channels, the complex averaging causes self-noise components to cancel toward zero, while the coherent room noise accumulates constructively.
The per-band SPL is obtained by summing the cross-spectral magnitude within each octave or third-octave band’s frequency range and applying Parseval normalization. This method is generally more effective at noise rejection and provides the most accurate sub-noise-floor measurements.
Applies IIR octave-band filters to each channel independently in the time domain, then accumulates the cross-product (L×R) of the filtered samples. The individual mic powers (L² and R²) are also accumulated for the red and blue traces. This method runs in real-time in the audio callback and updates continuously.
Because Filter×Corr averages the real-valued product rather than the complex cross-spectrum, it does not reject self-noise as efficiently as CPSD — you may see the green trace reading a few dB higher than with CPSD. However, it provides a useful quick check and responds immediately. This method is only available in octave mode.
Place both microphones as close together as possible (within a few centimeters) so that the room noise arrives coherently at both capsules. At high frequencies, even small separations can reduce coherence and lower the measured cross-correlation level. Pointing both microphones in the same direction helps ensure matched frequency response.
Two microphones of the same make and model will have similar self-noise characteristics and frequency response. This is not strictly required — the cross-correlation technique works with any two microphones — but matched mics give the cleanest results because the individual mic traces will track each other closely.
Longer averaging reduces variance and gives more stable readings. The CPSD method accumulates FFT segments over time; increasing the “FFT Averages” setting in Settings will use more segments per computation, giving a smoother result at the cost of slower updates. For very low noise floors, 30–50 averages or more may be needed.
Calibrate both channels before measuring. The cross-correlated result uses the average of both channels’ calibration offsets. If channel 2 has a different sensitivity, adjust its trim in the Mux settings.
Any vibration, handling noise, or electrical interference that is correlated between the two channels will NOT be rejected — it will appear as part of the measured noise floor. Use shock mounts, keep cables apart from power lines, and avoid touching the mic stands during measurement.
Several standardized noise rating curve families can be overlaid on the graph via Settings. These curves define maximum acceptable noise levels per octave band for different environments.
Noise curves are only available in octave band mode. All of these standards — NC, NCB, NR, PNC, and RC — are defined at octave-band center frequencies (31.5 Hz through 8 kHz). The rating is determined by comparing measured octave-band levels against the curve values at those specific frequencies. Third-octave data cannot be directly compared to these curves because the narrower bandwidth of each third-octave band results in lower per-band energy (approximately 5 dB less than the corresponding octave band), which would produce misleadingly low ratings.
The most widely used system in North America (ANSI/ASA S12.2). NC curves are defined across 8 octave bands (63 Hz – 8 kHz). The NC rating of a room is determined by the highest curve that any measured band touches or exceeds. The “limiting band” (LB) shown on the display identifies which frequency band determined the rating.
Typical NC targets:
| NC-15 to NC-20 | Concert halls, recording studios |
| NC-20 to NC-25 | Home cinemas (THX reference) |
| NC-25 to NC-30 | Bedrooms, private offices, churches |
| NC-30 to NC-35 | Living rooms, conference rooms |
| NC-35 to NC-40 | Restaurants, retail, open offices |
| NC-40 to NC-50 | Lobbies, corridors, workshops |
NCB (ANSI/ASA S12.2) is a balanced revision of NC that uses the same curve shapes but extends the evaluation to include 31.5 Hz. NCB was developed to address “rumble” and “hiss” complaints that the original NC system did not adequately capture. The NCB rating is computed identically to NC but evaluated across all 9 octave bands (31.5 Hz – 8 kHz).
The international equivalent of NC, defined by ISO 1996. NR curves are used predominantly in Europe and other regions outside North America. The curve shapes differ slightly from NC, generally being more restrictive at low frequencies. NR ratings are specified for the same types of environments as NC.
PNC is a refinement of NC developed by Beranek in 1971, adding the 31.5 Hz band and adjusting curve shapes for better correlation with subjective annoyance. PNC curves are somewhat more restrictive than NC at low and high frequencies. While less commonly specified today, PNC remains useful for critical listening environments.
RC curves (ANSI/ASA S12.2) evaluate noise character as well as level. Unlike NC, which only determines a single number, RC also classifies the spectral balance of the noise as neutral (N), rumbly (R), or hissy (H). RC is often preferred for HVAC noise evaluation because it identifies not just how loud the noise is, but whether it will be perceived as comfortable or annoying.
Background noise control is critical for home cinema and screening room performance. Low-level details in movie soundtracks — quiet dialogue, atmospheric effects, spatial cues — are masked by room noise. The following standards and recommendations apply:
THX specifies NC-25 (NCB-25) or lower for certified home cinemas and private screening rooms. This is the target for achieving the full dynamic range of movie soundtracks and hearing subtle surround effects. Commercial THX-certified cinemas target NC-30.
SMPTE recommends NCB-25 or better for critical screening rooms and dubbing stages. This recommended practice addresses both steady-state HVAC noise and transient intrusions. The NCB rating is preferred over NC because it includes the 31.5 Hz band, which is important for detecting low-frequency rumble from HVAC systems that can interfere with LFE channel reproduction.
Dolby recommends NC-25 or lower for Atmos-equipped home theaters to ensure that height channel and object-based audio details are not masked. Since overhead and surround speakers typically operate at lower levels than the main channels, background noise has a proportionally greater masking effect on these elements.
The HAA recommends NCB-25 as the target for dedicated home theaters. Their training materials emphasize that HVAC noise, equipment fans (projector, AV rack ventilation), and electrical noise (dimmers, transformers) are the most common sources of excessive background noise in residential installations. The HAA certification program requires noise floor measurements as part of the system commissioning process.
Common sources of background noise in home cinemas include:
Use the NC/NCB overlay to identify which octave bands exceed the target curve. The limiting band (LB) tells you where to focus your noise reduction efforts.
