Understanding ADCs: How Sound Becomes Digital

Every digital audio file, from streaming music to podcast recordings, starts as an analog signal — the continuous vibrations of air pressure — and is then converted into digital numbers that a computer can store and manipulate.
This process is handled by an Analog-to-Digital Converter (ADC).

Let’s break down how ADCs work, how sound is sampled, and why parameters like sample rate and bit depth matter.

Step 1: The Analog Signal

Sound in air is a continuous wave:

• Pressure varies smoothly over time
• Frequency and amplitude encode the pitch and loudness of sound
• Example: a 440 Hz sine wave is a pure tone at “A4”

Mathematically:

p(t) = A · sin(2πft + φ)  

Where:

• p(t) = instantaneous pressure
• A = amplitude
• f = frequency
• φ = phase

Step 2: Sampling — Measuring the Wave

Sampling is taking discrete measurements of the continuous waveform at regular intervals.

• Sample rate (fs) = how often we measure per second
• Nyquist theorem: fs must be at least 2× the highest frequency we want to capture

Example:

• Human hearing: 20 Hz – 20 kHz
• Minimum fs = 40 kHz
• Standard CD audio uses 44.1 kHz

Sampling creates a sequence of values:

p[n] = p(t) evaluated at t = n / fs, n = 0, 1, 2, ...  

This gives a time series of discrete points, each representing the waveform at that instant.

Step 3: Quantization — Rounding Values

Real-world pressure can have infinitely fine values.
Digital systems have finite resolution, so each sampled value is rounded to the nearest allowed level.

• Number of bits = resolution
• 1 bit → 2 levels (very poor)
• 16 bit → 65,536 levels (CD quality)
• 24 bit → 16,777,216 levels (high-resolution audio)

Quantization introduces a small error:

q[n] = round(p[n] / Δ) · Δ  

Where Δ = smallest step between levels.
This is why higher bit depth = lower quantization noise.

Step 4: ADC Pipeline — How the Converter Works

1. Anti-aliasing filter: removes frequencies above Nyquist limit to prevent aliasing
2. Sample-and-hold: holds the instantaneous voltage long enough for measurement
3. Quantizer: rounds the held voltage to nearest digital level
4. Encoder: outputs the number as a binary code

Simplified diagram:

• Analog → Filter → Sample & Hold → Quantizer → Digital Output

Step 5: Understanding Bit Depth and Dynamic Range

Bit depth determines how finely amplitudes are represented.

• Dynamic range (dB) ≈ 6.02 × bits

Examples:

• 16-bit → 6.02 × 16 ≈ 96 dB
• 24-bit → 6.02 × 24 ≈ 144 dB

Dynamic range = difference between the quietest and loudest representable signals.

Step 6: Nyquist Frequency and Anti-Aliasing

Aliasing occurs when high frequencies appear as lower “ghost” frequencies.

• To prevent it, a low-pass filter removes frequencies above fs / 2
• This highest allowed frequency is the Nyquist frequency

Example:

• fs = 44.1 kHz → f_max = 22.05 kHz
• Any signal above 22.05 kHz must be filtered out

Step 7: Oversampling and Noise Shaping

Modern ADCs often sample at higher rates and then downsample.

• Oversampling spreads quantization noise over a larger frequency range
• Noise shaping moves noise out of audible band
• Result: lower perceived noise without increasing bit depth

This is how 24-bit quality can be approximated using high-performance ADCs.

Step 8: From ADC to Storage

After sampling and quantization:

• Each sample is a number stored in memory
• Sequence of numbers → digital waveform
• Can be compressed (lossless or lossy) for storage or streaming

Example: 16-bit, 44.1 kHz stereo audio:

44,100 samples/sec × 2 channels × 16 bits = 1,411,200 bits/sec ≈ 176 KB/sec  

Step 9: Summary of Key ADC Concepts

Step 10: Why ADCs Are So Important

ADCs determine:

• Fidelity: how accurately the waveform is captured
• Noise floor: low bit depth increases hiss
• Frequency accuracy: sampling rate limits highest reproducible frequency
• Perceived quality: affects music, speech, and measurements

Without a good ADC, no matter how good your speakers, microphones, or headphones are, the captured sound will be limited and imperfect.

Digital audio is essentially frozen snapshots of continuous vibrations.
Understanding sampling, quantization, and ADC behavior is key to producing, measuring, and appreciating sound in the digital world.