Types
| Type | Cause | Character |
|---|---|---|
| Harmonic (HD) | Non-linear transfer | Integer multiples of input tone |
| Intermodulation (IMD) | Non-linearity mixing two tones | Sum / difference frequencies appear |
| Clipping (hard) | Output hits rail | Flat peaks, rich odd harmonics |
| Clipping (soft) | Gradual saturation | Rounded peaks, warm overtones |
| Crossover | Class-B amp zero-cross dead zone | Kink at zero — audible hiss on quiet parts |
| Slew-rate limit | Amp can't follow signal | Triangle-ish at high frequencies |
| Quantization | ADC / low bit depth | Granular noise |
| Group delay / phase | Non-flat phase response | Transient smearing |
Audibility thresholds (ballpark)
| Condition | THD threshold |
|---|---|
| Expert listener / reference gear | ~0.1% |
| Trained listener / good gear | ~0.3% |
| General listener / music program | ~1% |
| Heard as obvious distortion | > 3% |
| Tube amp "warm" territory | 1–5% |
| Overdriven guitar | 10% – 50%+ |
Measurement tips
- THD+N includes noise — always report which (THD vs THD+N) and at what level.
- Test at multiple frequencies — distortion rises with output level and frequency.
- Harmonic content matters: even harmonics (2nd, 4th) sound warm; odd (3rd, 5th) sound harsh.
- Masking: intermediate 2f_1 ± f_2 IMD products are less masked than harmonics — often more audible.
Converting percent THD to dB
Specifications quote distortion either in percent or in decibels below the fundamental. The conversion is dB = 20 × log10(THD ratio). A THD of 1% is 20 × log10(0.01) = −40 dB; 0.1% is −60 dB; 0.01% is −80 dB. A 2% figure is 20 × log10(0.02) = −34 dB. Each tenfold reduction in distortion is another 20 dB, so a spec that moves from 0.1% to 0.001% is a 40 dB improvement, far below the audibility thresholds in the table above.
Intermodulation example: feed 19 kHz and 20 kHz tones into a non-linear stage. Second-order products appear at 1 kHz (the difference) and 39 kHz (the sum); third-order products appear at 2 × 19 − 20 = 18 kHz and 2 × 20 − 19 = 21 kHz. The 1 kHz product falls in the audible band even though both inputs were ultrasonic, which is why two-tone tests expose problems that single-tone THD misses.
Where distortion comes from in practice
- Running out of headroom is the most common cause. Distortion rises sharply as output approaches the supply rail, so a clean amplifier spec at 1 W says little about 100 W.
- Feedback reduces distortion in proportion to loop gain, which falls with frequency. THD therefore usually rises toward 20 kHz.
- Output stages are a classic source: Class B crossover distortion is worst at low levels, while Class A stays smooth but wastes power as heat.
- Digital paths add quantization distortion at low levels and clipping above full scale. Dither converts the quantization distortion into steady, benign noise.
- Transducers dominate. Loudspeaker distortion is typically far higher than any amplifier, often 0.5 to several percent at low frequencies.