lesson

Updated 6 days ago ยท 1 view
Have you ever watched a car commercial where the car speeds forward, but its rims look like they are spinning backward? That same visual illusion happens in digital audio โ and when it does, your crisp high frequencies turn into harsh, fake low-pitched tones.
Pulse-Code Modulation and Sampling
Real acoustic sound is a continuous wave of air pressure changes. To capture it digitally, audio interfaces use Pulse-Code Modulation (PCM), taking thousands of instantaneous measurements of the wave's amplitude every second.
The sampling rate (fsโ), measured in Hertz (Hz), is how many times per second the analog-to-digital converter measures that wave. For example, standard CD audio takes 44,100 snapshots per second (44.1ย kHz).
๐Interactive/animated diagram showing an analog sine wave being sampled into PCM digital points. A smooth continuous blue sine wave moves across the screen. Vertical dashed lines drop down at regular time intervals T = 1/fs, placing bright circular sample points along the wave. Below it, an indicator shows the sample clock ticking at 44.1 kHz. Clean, light background (#ffffff), dark blue (#1e2945) wave, cyan (#0284c7) sample dots, labeled axes (Time vs Voltage).
How many snapshots do you actually need each second to capture a specific musical pitch without losing its identity?
The Nyquist-Shannon Theorem
In 1928, telecommunications engineer Harry Nyquist determined telegraph telegraph transmission speeds, which Claude Shannon proved mathematically for information theory in 1949: to accurately capture any waveform, your sampling rate must be at least twice the highest frequency present.
Because a single cycle of a sound wave has one positive peak and one negative trough, you need at least two sample points per cycle โ one high and one low โ to register that an oscillation took place at all.
The highest frequency a digital system can record without error is called the Nyquist frequency (fNโ), defined as exactly half the sampling rate: fNโ=2fsโโ
๐Comparison diagram showing two sampling conditions. Top: A 10 kHz wave sampled at 40 kHz (4 samples per cycle), showing accurate waveform reconstruction with green checkmark. Bottom: A 10 kHz wave sampled at 12 kHz (only 1.2 samples per cycle), showing that the sample points trace an entirely different, phantom low-frequency wave with a red warning badge. Clean light styling, labeled 'Nyquist Condition Met: fs >= 2f' vs 'Under-sampled: Creates Alias'.
What happens when an incoming sound frequency exceeds this Nyquist boundary? Let's trace where the extra energy goes.