lesson

Updated 6 days ago Β· 1 view
When a singer's voice moves air molecules into a microphone, that acoustic vibration becomes an electrical current.
How you store that signalβas an infinitely continuous physical voltage or as a series of sliced numbersβchanges everything about its frequency response, noise floor, and overload behavior.
Continuous Voltage vs. Discrete Data
An analogue audio signal is a continuous physical voltage that directly mirrors (is analogous to) the fluctuating pressure waves of sound.
A digital audio signal breaks that smooth curve into discrete numeric measurements through sampling (measuring time) and quantization (measuring amplitude).
πInteractive comparison diagram of Analogue vs. Digital Audio. Left panel: Smooth continuous sine wave labeled 'Analogue Voltage (Continuous Time & Amplitude)'. Right panel: The same wave overlaid with a grid showing vertical sampling intervals (time, fs) and horizontal quantization levels (amplitude, bit depth), with discrete red dots plotted on the grid intersections. A toggle button allows toggling between 'Raw Discrete Staircase' and 'Reconstructed Waveform'. Minimal style with clean card container (#ffffff), slate border (#e2e8f0), dark navy text (#1e2945), vivid blue wave (#2563eb), and red sample points (#ef4444). Responsive down to 350px width.
How many times per second do we need to take those digital snapshots to capture the full range of human hearing?
Bandwidth and the Nyquist Theorem
Bandwidth is the range of audio frequencies a system can accurately capture and reproduce, typically from 20Β Hz to 20Β kHz for human ears.
In 1928, telephone engineer Harry Nyquist determined that to accurately capture any frequency f, your sample rate (fsβ) must be at least double that frequency: fsββ₯2fmaxβ.
Because human hearing extends to 20Β kHz, standard CD audio uses a sample rate of 44.1Β kHz, giving a Nyquist frequency (the absolute maximum capture limit) of 22.05Β kHz.
πEducational visual showing the Nyquist Theorem and Aliasing. Top section: A 10 kHz wave sampled at 44.1 kHz has plenty of sample points (>4 per cycle) and reconstructs accurately. Bottom section: A 25 kHz input wave sampled at 44.1 kHz (below 2x) produces false, mirror 'ghost' frequencies (aliasing) folding back down to 19.1 kHz (44.1 - 25). Clear visual callouts showing 'Low-pass Anti-Aliasing Filter' removing frequencies above 22.05 kHz. Colors: #1e2945 text, #f8fafc background, #2563eb correct signal, #ef4444 aliased distortion.
If sample rate controls the frequency bandwidth, what governs the difference between a whisper and a deafening crash?