My Cart
Your Shopping Cart is Empty
Types of Distortion in Audio Systems and How to Reduce Them
|
0 comments
posted by Vincent Zhang
How to identify harmonic distortion, intermodulation, clipping and other signal errors in audio amplifiers and their interfaces.
Distortion is a departure from the expected signal after allowing for the intended gain and a fixed delay. Linear distortion changes the relative amplitudes or timing of existing frequency components. Nonlinear distortion can generate harmonics and intermodulation products. Direct pickup of supply ripple, mains hum or RF adds interference. These effects can coexist, and a single 1 kHz THD measurement samples only one operating condition.[1][2]
1. What Distortion Means in a Signal Chain
IEC 60268-3 describes amplifier measurements for total harmonic distortion, individual harmonic orders, modulation distortion, difference-frequency distortion and dynamic intermodulation distortion. Separate gain and phase measurements characterize the amplifier’s linear response.[1][19]
- THD and THD+N measure different quantities. THD combines the RMS levels of the measured harmonics and is commonly reported relative to the fundamental. A THD+N measurement removes the fundamental and combines the remaining in-band harmonics, noise, hum and other spurious signals into one RMS residual. A single result gives no breakdown by source. Report the test level and analysis bandwidth, and inspect the residual spectrum or measure output noise with the test tone muted when troubleshooting hum.[2]
- Audibility depends on the stimulus and playback conditions. In Lee and Geddes’s 2003 experiment, 34 listeners rated 21 stimuli, and neither THD nor IMD showed a statistically significant relationship with the ratings in that experiment. The frequencies and levels of the distortion products and masking by the programme affect how readily listeners hear them.[8][18]
2. A Taxonomy of Common Distortion Types
The table below maps each mechanism to its usual cause, a measurement that reveals it and the first design action worth taking. It is a diagnostic aid, not a ranking: one nonlinearity can produce several of these symptoms at once.[3][16]
| Type | Physical cause | Measurement that exposes it | First countermeasure |
|---|---|---|---|
| Harmonic / waveform | Curved transfer characteristic of an active device, core or transformer | THD, THD+N and individual harmonic orders vs level and frequency[1] | Linear device, clean bias point, symmetrical topology |
| Intermodulation (IMD) | Two or more tones mixing in the same nonlinearity | SMPTE RP120 (60 Hz + 7 kHz, 4:1) or a high-frequency two-tone test[3][4] | Avoid stage overload and improve open-loop linearity |
| Clipping / overload | Required output exceeds available voltage or current | THD vs output power; clipping and overload-recovery behaviour[7] | Power and gain headroom, correct load, clean recovery |
| Transient / slew | Slew-rate limiting, restricted small-signal bandwidth or inadequate phase margin | Square-wave / step response, full-power bandwidth, slew-rate calculation[5] | Speed margin, linear input stage, compensation, input band-limiting |
| Linear response / interface | Non-flat gain, frequency-dependent group delay or frequency-dependent loading | Frequency and phase sweeps[1][11] | Check bandwidth, loading and source impedance |
| Component nonlinearity | Capacitance that varies with signal voltage | Harmonic spectrum versus level and frequency[10] | Choose a suitable dielectric and check the AC voltage across the part |
| Additive interference | Supply ripple, ground coupling or RF pickup | Output noise and hum spectrum[2][24] | Improve return routing, shielding and supply filtering |
Figure 1. A sine wave in linear operation and a larger sine wave limited by symmetric output swing. Ideal symmetric hard clipping of a sine wave produces odd harmonics. Unequal positive and negative limits can also produce even harmonics.
2.1 Harmonic and Waveform Distortion
A nonlinear element such as an output transistor, vacuum tube, transformer core or loudspeaker motor can generate output components at integer multiples of a single input tone. The harmonic spectrum depends on the shape and symmetry of the nonlinearity. Audibility depends on each component’s level and frequency, as well as masking by the programme. Some listeners describe amplifiers with predominantly low-order harmonics as warm or full.[8][18]
In a balanced push-pull output stage, the two halves operate in opposite phase. Matching their transfer characteristics reduces even-order products, and equal quiescent currents flowing through opposite halves of a centre-tapped primary cancel their steady magnetizing flux. A conventional single-ended output transformer that carries the output valve’s DC plate current typically uses an air-gapped core to reduce the risk of saturation.[12][26]
2.2 Intermodulation Distortion (IMD)
When two tones pass through a nonlinearity, they can produce intermodulation components at |n·f1 ± m·f2|, where n and m are positive integers. The mixing mechanism defines the product. Its frequency can coincide with a harmonic for some choices of input tones. Single-tone harmonic tests and multitone IMD tests probe different effects of the same nonlinear circuit.[3][17]
SMPTE RP 120 testing commonly uses 60 Hz and 7 kHz with a 4:1 low-to-high amplitude ratio. CCIF/ITU-R and IEC difference-frequency distortion (DFD) methods use two equal-amplitude high-frequency tones. Specify the analysis method because the reference levels used by CCIF and DFD can differ. IEC 60268-3:2018 specifies an 80 Hz separation for its DFD procedure unless there is a reason to choose another spacing.[4][19]
With a 19 kHz and 20 kHz test pair, linear summation produces a beat envelope at 1 kHz while the spectrum contains just the two input tones. Nonlinear mixing in the circuit can produce an actual 1 kHz output tone, the second-order difference product, and third-order products at 18 kHz and 21 kHz.[3][17]
Design note. Check each stage’s signal swing and operating point, and improve its linearity at the intended level. Compensation changes stability and the loop gain available to suppress distortion. Current-feedback op-amps can offer high slew rates that vary less with input step size. Measure IMD in the complete circuit at its intended gain and load.[6][25]
2.3 Clipping and Overload Distortion
Voltage clipping starts when the requested output exceeds the amplifier’s available voltage swing at the test load (Figure 1). Current limiting is a separate overload mechanism and can distort the output before the voltage swing limit is reached. The resulting waveform and recovery depend on the load and protection circuit. Check for delayed recovery after either form of overload.[7][21]
Use the gain structure, speaker sensitivity and listening distance to estimate the required output. When starting from an average listening level, include the programme’s peak-to-average ratio. An additional 3–6 dB above the estimated programme peak is an example of a design margin. Verify available voltage swing and peak current with the intended load, then use a distortion-versus-output sweep and an overload-recovery test to establish the operating limit.[7][21][22]
2.4 Transient, Slew and Bandwidth-Related Distortion
Slew-rate limiting is a large-signal nonlinearity. Limited small-signal bandwidth can round a step, and inadequate phase margin can cause overshoot or ringing. For a sine-wave output of peak-to-peak voltage VPP and frequency f, the required peak slew rate is SR = π · VPP · f. Severe slew-rate limiting can make the output approach a triangle with reduced amplitude. Use small-signal steps to check settling and large-signal tests to check slew rate.[5][25]
Transient intermodulation distortion (TIM) describes unwanted mixing under rapidly changing, large-signal excitation. The IEC dynamic intermodulation distortion (DIM) procedure uses a band-limited square wave combined with a sine wave to probe this behaviour. A linear input stage, sufficient large-signal speed and appropriate input filtering help avoid transient overload. In transformer-coupled stages, primary inductance works with the source and reflected load impedances to shape the low-frequency response. Leakage inductance and winding capacitance influence the high-frequency response. Interleaving can reduce leakage inductance, with a trade-off in winding capacitance.[1][13][19][27]
2.5 Linear, Phase and Interface-Related Errors
Non-flat gain and frequency-dependent group delay can change a waveform without adding new frequency components in a linear time-invariant circuit. A capacitor’s frequency-dependent reactance contributes to the circuit’s linear response. A capacitance that varies with the instantaneous voltage can generate nonlinear distortion, particularly when significant AC voltage appears across a high-permittivity ceramic or certain electrolytic capacitors. Check the actual signal swing and bias at the capacitor.[10]
A line-level connection from a preamplifier to a power amplifier uses impedance bridging. An input-to-source impedance ratio of at least 10:1 is a common starting point for limiting signal loss. For example, a preamplifier output impedance below 1 kΩ and an amplifier input impedance above 10 kΩ meet that ratio. At the speaker terminals, a 4 Ω load rating describes the nominal load the amplifier is designed to drive. Check the speaker’s impedance minimum and phase as well as its nominal rating.[11][14]
3. Six Design Levers for Lower Distortion
Device selection, topology, power supply, layout, loading and feedback interact. Choose the changes that address the dominant error at the intended signal level and load.
Figure 2. Common symptoms and design changes to investigate for four groups of causes.
3.1 Component Linearity and Operating Point
- Vacuum tubes: choose a device with a wide, smooth linear region and set the bias point away from cutoff and saturation. In push-pull stages, matched transfer characteristics improve even-order cancellation, and balanced idle currents minimize residual DC flux in the output transformer.[12]
- Transistors and ICs: select op-amps with adequate slew rate and gain-bandwidth product for the required output swing; excess speed margin reduces the input differential voltage the first stage must handle and can lower distortion.[5][6]
- Passives: low-temperature-coefficient metal-film resistors keep divider ratios stable as parts warm up. Film capacitors and low-voltage-coefficient C0G/NP0 ceramics are options where significant AC voltage appears across the part. Check capacitance, voltage and current ratings for the application.[10]
- Magnetics: high-permeability, low-loss core material, an air gap for single-ended service, and interleaved or sectional windings to control leakage inductance and winding capacitance.[12][13]
3.2 Circuit Topology
Choose a topology that supplies the required gain and load drive while keeping each stage within its linear operating range. A buffer can isolate a voltage-gain stage from a low-impedance load. A well-balanced push-pull stage suppresses even-order distortion. A single-ended Class A stage avoids the device hand-off that can cause crossover distortion in a Class AB output stage. In Class AB, setting the idle bias correctly reduces the crossover notch. Compensation sets the stability margin and influences how much loop gain remains available for distortion correction.[12][16][23][25]
3.3 Power Supply Stability
Supply ripple and shared return currents can add hum and noise to the output. Under load, supply sag can also reduce available swing or modulate the signal. Set the allowable ripple using the circuit’s supply rejection at the ripple frequency and its output-noise budget. Local regulation, filtering and separate supply or return paths can reduce coupling. For switch-mode supplies, check the output spectrum for switching residue and any products that fall in the audio band.[24]
3.4 Signal Path and PCB Layout
Keep sensitive traces short and route high-current supply and output returns away from low-level signal returns. Use shielded interconnects where needed and keep the feedback sense and reference nodes clear of load-current voltage drops. Choose the feedback takeoff point with the output network and the loop’s stability in mind. If the loop senses after an output filter, verify the phase margin across the intended range of loads.[23]
3.5 Load Control and Headroom
Loudspeaker impedance varies with frequency, driver resonance, the crossover network and voice-coil temperature. The conventional amplifier damping-factor specification is DF = Znom / |Zout(f)|, using the speaker’s nominal impedance and the magnitude of the amplifier output impedance at a stated frequency and measurement point. To estimate effective damping at the speaker terminals, include the cable’s round-trip series impedance. For an 8 Ω nominal speaker, a damping factor of 20 corresponds to 0.4 Ω of source impedance. Assess the resulting response with the actual impedance curve. A speaker-side Zobel can compensate a driver’s rising impedance over part of the band when the crossover or output filter needs a flatter load.[14][15][28]
3.6 Negative Feedback
Negative feedback reduces small errors generated within a stable loop. In a voltage-series configuration, voltage sensing and series input mixing raise input impedance and lower output impedance. In a current-shunt configuration, current sensing and shunt input mixing lower input impedance and raise output impedance, which helps regulate output current. The available peak current still depends on the output stage, supply and protection limits.[20][21]
For a small output disturbance within the loop, the residual amplitude is approximately 1/|1 + T(f)| times the uncorrected amplitude, where T(f) is the loop gain at the disturbance frequency. The correction generally weakens as loop gain rolls off. Verify stability with the intended load. Voltage swing and slew limits still apply in overload.[9][16][23][25]
4. How to Verify: Measurements That Matter
- THD, THD+N and individual harmonics. Measure versus output level and frequency, and report the load, test level, analysis bandwidth and weighting if used. The Rohde & Schwarz application note suggests starting with the second through fifth harmonics. Include higher orders when they are significant and within the analysis bandwidth.[1][2][19]
- IMD. Use a defined SMPTE and high-frequency two-tone method. Record the tone frequencies, amplitudes, load and reporting convention. Add a DIM test when large, fast transitions are relevant to the application.[1][4][19]
- Square-wave and step response for overshoot, ringing and settling; full-power bandwidth for slew-related limits.[5]
- Frequency and phase sweeps. Use the specified resistive load for repeatable amplifier measurements. Repeat with the intended speaker or a representative reactive load to check load-dependent behaviour.[19][23]
- Clipping onset and overload recovery, including whether the output sticks to the rail after a transient.[7]
- Residual and noise spectra. Inspect the waveform and spectrum after removing the test tone to identify crossover artifacts and spurious tones. Repeat with the tone muted to check the idle noise and hum.[2][16][24]
State the analysis bandwidth when comparing results. With a 10 kHz fundamental and a nominal 20 kHz analysis bandwidth, harmonics above the second are outside the band, and the second is at the band edge. Use a wider analysis bandwidth for out-of-band harmonic measurements and report that bandwidth with the result.[1][2]
5. A Practical Priority Checklist
- Set gain and headroom from the expected programme peaks. Check each stage for overload.
- Confirm amplifier load rating, speaker impedance curve and cable resistance; add impedance compensation if needed.
- Clean up the supply and grounding: ripple, isolation and return paths before exotic components.
- Set bias points and device matching; choose symmetrical topology where it cancels the dominant products.
- Check AC voltage across signal capacitors and choose dielectrics and ratings appropriate to each position.
- Apply feedback with verified stability margin, then re-measure the complete system.
6. Frequently Asked Questions
- Is a lower THD figure always better sounding?
- A THD result gives the combined level of the measured harmonics under stated test conditions. In Lee and Geddes’s 2003 experiment, THD and IMD did not correlate significantly with the listening ratings. To assess a particular system, examine the spectrum and the levels and loads at which the distortion occurs.[2][8]
- Why are tube amplifiers often described as "warm"?
- Some single-ended designs produce a prominent second harmonic, and listeners may describe the result as warm. The spectrum also depends on the tube, bias, load, transformer and feedback. Push-pull symmetry can reduce even-order products.[12][16]
- Does negative feedback remove distortion?
- Feedback can reduce small distortion components generated within a stable loop. The correction at each component’s frequency depends on the loop gain there. The circuit still needs sufficient voltage swing, output current and slew rate for the signal and load.[9][23][25]
- How much amplifier headroom do I need?
- Start with the expected programme peak. If your starting point is an average level, account for the programme’s peak-to-average ratio. An extra 3–6 dB above the expected peak is one possible margin. Verify it against the amplifier’s measured output limit into the intended load.[7][21][22]
- Do "audio-grade" capacitors reduce distortion?
- Look at the dielectric, capacitance, bias and AC voltage across the part. X7R and other high-permittivity ceramics can generate appreciable distortion when the signal voltage across them is large. Film or C0G/NP0 capacitors can reduce that mechanism where their value and ratings suit the circuit.[10]
- What should I check first when an otherwise good-measuring system sounds harsh?
- Clipping or near-clipping at the listening level, then grounding and supply noise, then load interaction. These are the mechanisms a single mid-band THD number most often misses.[2][16]
Find More
References
- Rohde & Schwarz, "Tests of Audio Amplifiers in Accordance with Standard IEC 60268-3" (application note, PDF). https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/00aps_undefined/RCS0702-0032.pdf
- Audio Precision, "More about THD+N and THD". https://www.audioprecision.com/news/more-about-thdn-and-thd
- Rod Elliott (Elliott Sound Products), "Intermodulation Distortion". https://sound-au.com/articles/intermodulation2.htm
- Audio Precision, "More about IMD". https://www.audioprecision.com/news/more-about-imd
- Texas Instruments, "Ramping Up on Slew Rate" (SLOA332A, PDF). https://www.ti.com/lit/an/sloa332a/sloa332a.pdf
- Texas Instruments, "Input Voltage Effects on Slew Rate for Voltage and Current Feedback Amplifiers" (SBOA610, PDF). https://www.ti.com/lit/pdf/sboa610
- Rod Elliott, "Amplifier Sound — What Are The Influences?". https://sound-au.com/amp-sound.htm
- Lidia W. Lee and Earl R. Geddes, “Auditory Perception of Nonlinear Distortion” (AES Convention Paper 5891, 2003). https://aes.org/publications/elibrary/elibrary-page/?id=12465
- James M. Fiore, “Limitations On The Use of Negative Feedback.” https://eng.libretexts.org/Bookshelves/Electrical_Engineering/Electronics/Operational_Amplifiers_and_Linear_Integrated_Circuits_%28Fiore%29/03%3A_Negative_Feedback/3.05%3A_Limitations_On_The_Use_of_Negative_Feedback
- Texas Instruments, “Selecting Capacitors to Minimize Distortion in Audio Applications” (SLYT796A). https://www.ti.com/lit/an/slyt796a/slyt796a.pdf
- Rane, “Technical Support” (audio load and source impedance guidance). https://www.ranecommercial.com/legacy/tech.html
- The National Valve Museum, “Valve Technology.” https://www.r-type.org/articles/art-010q.htm
- Aiken Amplification, “Output Transformers Explained.” https://www.aikenamps.com/index.php/output-transformers-explained
- John Siau, “Damping Factor Isn’t Much of a Factor” (Benchmark Media Systems). https://benchmarkmedia.com/blogs/application_notes/audio-myth-damping-factor-isnt-much-of-a-factor
- Philip Newell and Keith Holland, “Loudspeakers: Effects of Amplifiers and Cables — Part 4” (EDN). https://www.edn.com/loudspeakers-effects-of-amplifiers-and-cables-part-4/
- Rod Elliott, “Distortion and Feedback.” https://sound-au.com/articles/distortion%2Bfb.htm
- Georgia Institute of Technology (W. M. Leach course notes), "Intermodulation Distortion" (PDF). https://leachlegacy.ece.gatech.edu/ece4435/sp10/imdist.pdf
- Earl R. Geddes and Lidia W. Lee, “Auditory Perception of Nonlinear Distortion – Theory” (AES Convention Paper 5890). https://aes.org/publications/elibrary/elibrary-page/?id=12464
- International Electrotechnical Commission, IEC 60268-3:2018, Sound System Equipment – Part 3: Amplifiers, §§15.12.5–15.12.9. https://webstore.iec.ch/en/publication/32788
- Collaborative Compendium on Analog Integrated Circuits, “Introduction to Feedback” (feedback topologies). https://bmurmann.github.io/COCOA/contents/partII/partII.2.html
- Texas Instruments, “How to Avoid Overcurrent Shutdown in TAS58xx Devices” (SLAAEU8). https://www.ti.com/lit/an/slaaeu8/slaaeu8.pdf?ts=1781162997734
- Texas Instruments, TPA0162 data sheet (audio crest factor and peak power). https://www.ti.com/lit/gpn/tpa0162
- Texas Instruments, “TPA324x and TPA325x Post-Filter Feedback” (SLAA788A). https://www.ti.com/lit/an/slaa788a/slaa788a.pdf
- Texas Instruments, “Guidelines for Measuring Audio Power Amplifier Performance” (SLOA068A). https://www.ti.com/lit/an/sloa068a/sloa068a.pdf
- Texas Instruments, “THD+N and Slew Rate” (SDAA309). https://www.ti.com/lit/an/sdaa309/sdaa309.pdf
- Hammond Manufacturing, 125ASE single-ended output transformer data sheet (gapped core and DC current rating). https://www.hammfg.com/files/parts/pdf/125ASE.pdf
- Eero Leinonen, Matti Otala and John Curl, “A Method for Measuring Transient Intermodulation Distortion” (JAES, 1977). https://aes.org/publications/elibrary-page/?id=3379
- Analog Devices, MAX9744 data sheet (speaker impedance compensation). https://www.analog.com/media/en/technical-documentation/data-sheets/max9744.pdf
Related Posts
0 comments