THD+N Total Harmonic Distortion + Noise

What THD+N is and what it measures

THD+N stands for Total Harmonic Distortion plus Noise. It captures the harmonic distortion products together with all other non-fundamental energy in the output — noise, hum, interference, intermodulation — expressed as a single ratio relative to the fundamental signal. The result is usually reported either as a percentage of the fundamental or as a dB ratio: 0.1% corresponds to -60 dB, 0.001% to -100 dB, and 0.0003% to roughly -110 dB.

The +N matters because the figure folds in everything that isn't the fundamental — ground-loop hum, high-frequency interference, intermodulation between hum tones and the signal — not only the discrete harmonic spectrum, which makes it more comparable across devices than THD alone. The THD ratio is therefore always lower than the THD+N ratio for the same device under the same conditions. There is also a historical reason the two are bundled. In the pre-FFT era, analyzers could not isolate harmonics from noise; they could only notch out the fundamental and read whatever was left. The combined number stuck as a single, easily-stated figure of merit even though modern FFT instruments can separate the two.

How THD+N is computed

THD+N is computed by passing the device output through a sharp notch filter tuned to the fundamental, applying a bandwidth-limiting filter, then taking the RMS of what remains and dividing by the RMS of the unfiltered signal. The notched residual is harmonics plus noise plus everything else; the denominator is the full original signal.

The integration bandwidth is critical to the result and must be reported alongside the number. For audio the conventional window is 20 Hz to 20 kHz, and AES17 specifies bandwidth-limiting the residual to 20 kHz or the upper band-edge frequency, whichever is lower. A wider window admits more noise and pushes the figure up — integrating over 100 kHz instead of 20 kHz can materially worsen a quoted result even though the device hasn't changed. The stimulus is a sine wave at a defined frequency and amplitude. AES17 specifies 997 Hz for digital and digital-plus-analog systems, while IEC 60268 specifies 1 kHz for analog-only systems; 997 Hz is used because it is not an integer divisor of common sample rates, so the test tone lands on different converter codes from sample to sample and exposes converter behavior an exact 1 kHz tone could mask.

THD+N and SINAD describe the same underlying linearity, just inverted: for a given input frequency and amplitude they are reciprocal as long as the noise bandwidth is the same for both. A THD+N of 0.1% (-60 dB) corresponds to a SINAD of +60 dB; 0.001% (-100 dB) corresponds to +100 dB.

What good THD+N numbers look like

As a reference point for what 'transparent' THD+N looks like in modern solid-state amplification, Benchmark's AHB2 power amplifier is rated at less than 0.0003% THD+N at full rated power across the 20 Hz to 20 kHz audio band. Benchmark explicitly contrasts this with the much looser 1% THD+N rating point that has long been the manufacturer convention for stating amplifier output power.

A THD+N versus level sweep typically shows a falling curve at low signal levels — where the residual is mostly noise and any fixed noise floor shrinks as a percentage of the rising fundamental — bottoms out at the device's most linear operating region, then rises sharply as the device approaches clipping. This shape, not a single 1 kHz number, is what tells you how the device behaves across its dynamic range.

Whether a given THD+N figure is audible depends on whether the distortion components fall below the absolute hearing threshold or are masked by the music itself. If the residual sits below either limit, the listener cannot distinguish the distorted signal from the original. This is why two devices with the same THD+N can sound different — different harmonic distribution, different masking by the program material — and why very low numbers (well below ~0.001%) typically deliver no further audible benefit on music playback.

Common pitfalls in THD+N reporting

A single THD+N figure is meaningless without its measurement context. Audio Precision recommends a complete spec be written, for example, as 'THD+N less than 0.01%, 1 Vrms, 20 Hz – 20 kHz, unity gain, 20 kHz BW' — stimulus level, frequency range, gain, and measurement bandwidth all stated. THD+N changes with frequency (higher fundamentals push some harmonics out of the measurement band), with output level (the curve-shape behavior above), and for amplifiers with load impedance.

An amplifier's THD+N at 1 W, well inside the linear region, is typically far better than its THD+N near rated output, where the device is approaching clipping. Manufacturers can legitimately quote either point — but a low number quoted at 1 W with no rated-power figure, or vice versa, is not enough to characterize the amplifier. The only honest picture is the full level-vs-distortion sweep, which is why third-party measurement sites publish curves rather than single numbers.

THD+N is also not interchangeable with intermodulation distortion (IMD), which uses a two-tone stimulus and reveals different non-linearities, and it is not the same as the noise floor measured with no signal present. THD+N folds the noise floor into a single metric only with a signal applied, and it cannot replace IMD testing for catching distortion mechanisms — like sum-and-difference products — that a single-tone test will miss.