Audio Distortion Threshold Test: Audibility Analysis

This comprehensive discussion covers the design philosophy and technical innovations behind high-fidelity audio amplifiers. GrimmAudio follows a bottom-up approach, questioning conventional designs and examining existing solutions. The key distortion mechanisms addressed are: (1) Phase modulation distortion, where unstable crossover points cause audible phase changes as the ear is extremely sensitive to phase variations; (2) Thermal distortion, where transistors heat and cool rapidly (milliseconds) within the audio frequency range, causing inconsistent behavior since all transistor parameters depend on temperature; (3) Harmonic distortion, minimized through second-order feedback systems with constant loop gain up to 20kHz achieving -130dB distortion. Solutions include cascoding circuits to maintain constant voltage across transistors, keeping power dissipation constant, and dividing power across approximately 60 pairs of output transistors in Class AB design. The feedback controversy is resolved by applying either no feedback or high feedback, as the middle ground (20-30 dB) is the worst choice due to the feedback mixer's inherent nonlinearity.

High-fidelity audio amplifiers require prioritizing distortion minimization over bandwidth expansion, as the inverse relationship between bandwidth and distortion means that to achieve wide frequency response covering the entire audible spectrum (20 Hz to 20 kHz), system distortion must be maintained at nearly zero levels; this requires careful design of the driver stage, proper neutralization of tubes to eliminate parasitic oscillations, and implementation of feedback loops that provide approximately 14 dB of feedback while maintaining signal integrity, with the driver stage being the most critical section where most problems and residual distortions originate.

Negative feedback reduces harmonic distortion because both fundamental and harmonic signals are reduced by the same factor. Since harmonics are typically much smaller than the fundamental, their relative contribution decreases significantly. The Total Harmonic Distortion (THD) measures the ratio of harmonic content to fundamental signal. Lower THD indicates better linearity. This makes negative feedback essential for high-fidelity audio amplifiers.

This comprehensive section covers the core principles of negative feedback in audio amplifiers. The Williamson amplifier pioneered voltage feedback by sensing transformer output voltage, while current feedback senses voltage across a series resistor. Voltage feedback loses signal when output is shorted and varies with speaker impedance, while current feedback increases feedback when shorted, potentially damaging the amplifier. Phase response varies with frequency, with only 180° providing true negative feedback; other frequencies cause oscillation. Feedback only works for correlated signals, not random noise, which actually worsens performance. The twin coupled amplifier addresses these limitations through push-pull feedback from both output tube cathodes, reducing susceptibility to impedance variations and eliminating distortion from tube switching.

Phono stage equalization can be implemented as a feedback loop or as a filter toward ground. The feedback method can result in significant headroom loss (up to 40 dB) in high frequencies. Moving coil cartridges require amplification from 1 to 10,000, presenting technical challenges. Inductive coils control amplifier gain and current. Open-loop circuit design (no feedback) represents an alternative approach where circuits work independently and respond very quickly. Negative feedback was first attempted in tube amplifiers in the 1950s but became practical with semiconductor development. High-quality circuits may avoid electrolytic capacitors entirely. Fully balanced audio signals can be transmitted using RCA connectors by keeping the cold lead separate from ground.

Creating a scientifically valid blind test requires addressing multiple challenges: (1) Sufficient runs (typically 20+) to achieve statistical significance; (2) Proper level matching using software controls rather than potentiometers; (3) Elimination of all physical tells including relay clicks and environmental sounds; (4) Remote verifiability so others can inspect files and procedures. The video demonstrates using specialized relay-switching equipment that fully disconnects signals during switching to prevent any mechanical cues. Testing oversampling filters digitally using identical source files ensures only the filter differences remain as variables.

The double-blind ABX test is a scientific method used to compare two audio situations objectively. In this test, the listener first hears condition A, then condition B, and finally hears condition X without knowing whether X is A or B. The operator randomly selects which condition X represents. If the listener cannot reliably identify the difference, it indicates that the difference between A and B is imperceptible to human hearing. This method eliminates bias from brand recognition and expectation.

The ABX testing method is the scientific standard for comparing audio files. In this test, a listener hears three files: A, B, and X (where X is either A or B). The listener must identify whether X matches A or B. To be statistically valid, the test must be repeated multiple times. This methodology eliminates bias by preventing the listener from knowing which file is which. The test requires that files differ only in the parameter being tested (such as bit resolution) and that volume levels are identical between files.

A double-blind audio test is a scientific methodology where neither the listener nor the test operator knows which audio component is active, eliminating external variables that could influence perception. In Mike Borghese's test at Fiera HiFi Milano, participants who heard two nearly identical crossover filters (differing by only 0.5 dB in frequency response) scored between 40-60% correct identifications, which falls within the expected range for random guessing. This demonstrates that while measurable technical differences exist between audio components, they may not always be perceptible to human ears, challenging the common assertion that 'only the ear can hear' audio differences.

Blind audio testing is a scientific methodology used to determine whether listeners can genuinely perceive differences between audio equipment when visual information is removed. The test involves placing pedals in mystery boxes that cannot be seen by participants, who then play through both options without knowing which is which. This approach eliminates bias from brand perception, price expectations, and visual cues, allowing participants to judge purely on sound quality. The methodology is commonly used in audio engineering and music technology to evaluate whether perceived differences between equipment are real or influenced by psychological factors.

Audio clipping, traditionally considered a dangerous distortion artifact, can be strategically employed as a professional mixing technique to enhance mix loudness and punchiness; unlike limiters that compress transients and reduce dynamic range, clippers transparently cut off audio peaks at a controlled threshold, preserving the punchy character of transients while creating additional headroom that allows mixes to be pushed louder during mastering without audible distortion.

The harmonic excitation technique, pioneered by Aphex in the 1970s with their Aural Exciter technology, uses specialized distortion modules (tape, uart, and voice) combined with linear phase filters to generate new harmonics from existing frequencies, allowing producers to dramatically enhance the power, depth, and professional quality of their music by selectively boosting specific frequency ranges like sub-bass, mid-bass, and brightness without introducing phase problems.

The Oxford Inflator is a saturation/harmonic exciter that adds high-end energy to tracks. Apply it before reference plugins and always A/B test with gain matching to ensure peak volume remains the same. This plugin makes tracks sound louder and fatter without changing the actual peak level. It's an optional but recommended step in the coloring phase of mastering.

Mid-side processing separates center (mono) from sides (stereo), providing more control than traditional stereo processing. Always audition in mono to ensure compatibility with mono playback—the mono image represents what listeners hear most clearly. Apply excitation/saturation only to the side component to add character and width without affecting the clean mono center. Use subtle master reverb as a blending factor to fill empty gaps and make the track perceive as bigger and warmer. Different distortion types serve different purposes: warm vintage-style, retro-style, tube triode, and dual triode each create distinct sonic characters. FabFilter Saturn allows band splitting for granular saturation control. Distortion plugins can adversely affect quiet parts by adding excessive noise to already quiet sections. Bass should always remain mono to avoid phase issues and muddiness. High frequencies can be pushed wider for added dimensionality. These advanced techniques require careful application—overdoing any effect creates audible artifacts. The key principle remains subtlety: these tools enhance the natural character of your mix rather than overwhelming it.

Clipping occurs when audio exceeds zero decibels in digital formats, flattening waveforms and adding harmonics; hard clipping creates harsh, square-wave-like distortion while soft clipping and saturation preserve waveform roundness and produce more pleasant, analog-like warmth. The impact of clipping varies by sound type—harmonic-rich sounds like saw waves may tolerate clipping with minimal audible change, while percussive sounds like pianos produce immediately undesirable results. While final master export should avoid clipping, strategic use of soft clipping during production can achieve louder mixes with lower peaks while preserving punch, offering an alternative to compression that maintains dynamic impact.