Harmonic distortion is evaluated by injecting a sine‑wave signal — typically, as here, at 1kHz — and measuring the amplitude of the harmonics generated at multiples of that frequency.
Low distortion is often a marker of quality in audio equipment. We explain how to make sense of standard distortion specifications.
The previous instalment of this series defined distortion (see Part 2), for the purpose of this treatment, as “any signal component added by elements of the signal path in response to the intended audio. In the absence of an audio input, distortion is zero.” Though there are others, two forms of distortion dominate in audio systems, and appear most often on equipment data sheets. These are harmonic distortion and intermodulation distortion.
Both forms of distortion derive from nonlinear electronic components — devices with outputs that are not strict multiples of their input — that necessarily form the core active (amplifying) circuits in audio‑signal processing blocks. These exist throughout the analogue signal path, from a microphone transducer’s or instrument’s output to the analogue‑to‑digital converter. On the reproduction and monitoring side, similar causes of distortion exist within the digital‑to‑analogue converter, mixers and other signal‑processing blocks, and monitor amplifiers.
Harmonic Distortion
As the name suggests, sources of harmonic distortion add harmonics — frequency multiples — of the input signal to their outputs. The presence of added harmonics can have several harmful effects, which degrade audio quality in audio capture, post‑production, and reproduction signal chains when present in more than minute amounts. Harmonic distortion, for example, degrades sonic transparency due to exaggerated spectral density and, at sufficient amplitudes, reduces the perceived realism of acoustic instruments and naturally occurring sounds. At even greater distortion amplitudes, musicians put this latter trait to creative effect with many instrument pedals, such as overdrive and fuzz stompboxes, creating sounds almost unrecognisable as deriving from the raw instrument output, but here the deviation from the pure instrument tone is purposeful. In production and reproduction gear, the goal nearly always is to preserve the original signal as accurately as possible irrespective of its origin.
For the purpose of reading spec sheets, we’re interested in very low levels of distortion, which are given in (fractional) percent or (negative) dB. Equipment manufacturers most often state harmonic distortion as an aggregate measure of harmonics within a frequency band, which they refer to as THD (total harmonic distortion) or as THD+N (total harmonic distortion plus noise), depending on measurement method.
The measurement begins with an ultra‑pure (very low distortion) sine‑wave generator that provides a stimulus to the device under test. The analyser monitors the device-under-test (DUT)’s output, and filters out the stimulus frequency — the fundamental, in this exercise — through a narrow notch filter. What remains within the measurement band are the harmonic residues plus noise, which the analyser measures and presents as a fraction of the fundamental’s amplitude. Analysers can apply narrow‑band filters centred on the harmonics as well, eliminating the noise component of the measurement.
As we’ve seen with other parameters reported on spec sheets, THD and THD+N figures are only meaningful if the test conditions are stated, and these need to agree between spec sheets if device‑to‑device assessments are to be accurate. These include, at minimum, the test frequency and the measurement bandwidth. Most commonly, 1kHz is used for the former, which allows space for a good number of harmonics to show up within a typical measurement bandwidth of 20Hz to either 20kHz or 22kHz. But, as the BBC would put it, “other test frequencies are available”, so be sure to check when making comparisons between competing products.
Note that the modest upper frequency limit of the measurement bandwidth means that you don’t have to increase the test frequency much before you limit the in‑band harmonics to just a few. For example, at a 1kHz test frequency, 20 harmonics fit within a 20kHz measurement bandwidth (with some attenuation possible in the last one). But at a 5kHz test frequency, three harmonics barely fit in a 20kHz measurement bandwidth, and at 7kHz you’re down to one.
While judicious amounts of harmonic distortion can serve musical interests, the same is not true of intermodulation distortion (IMD), which is virtually always discordant.
Intermodulation Distortion
While judicious amounts of harmonic distortion can serve musical interests, the same is not true of intermodulation distortion (IMD), which is virtually always discordant. Like harmonic distortion, intermodulation distortion results from nonlinearities in signal‑processing blocks. In the case of IMD, however, the distortion signal components are the sums and differences of frequencies in the source audio.
Intermodulation distortion is tested by injecting multiple sine waves at harmonically unrelated frequencies. These sine waves, visible here, would later be filtered out so that the amplitude of the distortion artifacts can be measured.
Measuring a device’s IMD requires two test frequencies that are not harmonically related. A test system generates two or more ultra‑pure sine waves as the input signal to the DUT. The tester’s signal analysis section uses narrow notch filters to remove the test frequencies from the DUT’s output signal, and measures the amplitude of the residual sum and difference frequencies. Test standards for IMD call out the test frequencies and their relative amplitudes so, when comparing two competing product’s spec sheets, it’s important to check that IMD measurements comply with the same standard.
Like most THD and THD+N measurements, IMD specs reflect a spot test of complex behaviour. They do not test the DUT’s distortion performance across the entire audio range, which can vary particularly at the upper end of the audio spectrum. So, while comparisons of competing products’ spec sheets do provide valid and valuable performance comparisons as long as their test methods and operating conditions match, they cannot tell the entire story of products’ distortion performance, and they cannot predict exactly what your ears will experience under real‑world audio production or reproduction applications. In other words, spec sheet comparisons serve as a crucial first‑step evaluation of competing products, but they cannot entirely replace critical listening or critical thinking about what level of performance is necessary to satisfy your goals.
This series is produced in association with Audio Precision, Inc.