How Sample Rate Works
When analog sound is digitized, an analog-to-digital converter (ADC) captures the amplitude of the waveform at regular intervals. Each captured measurement is called a sample, and the number of samples taken per second is the sample rate (also called sampling rate). A higher sample rate captures more detail from the original signal, enabling more accurate reproduction during playback.
The Nyquist-Shannon Sampling Theorem
The theoretical foundation for choosing a sample rate is the Nyquist-Shannon sampling theorem. It states that to accurately reconstruct a signal, the sample rate must be at least twice the highest frequency present in that signal. Because human hearing extends to roughly 20 kHz, a sample rate of at least 40,000 Hz is needed to reproduce the full audible spectrum. This is why the CD standard uses 44,100 Hz.
Common Sample Rates and Their Uses
8,000 Hz : Traditional telephony and narrowband voice.
16,000 Hz : Wideband speech recognition and voice AI applications.
44,100 Hz : CD-quality music playback.
48,000 Hz : Professional video, broadcast, and streaming platforms.
96,000 Hz and above : High-resolution audio production and mastering.
Sample Rate vs. Bit Depth
Sample rate and bit depth are complementary dimensions of digital audio quality. While sample rate determines how often the signal is measured (temporal resolution), bit depth determines how precisely each measurement is recorded (amplitude resolution). Together, they define the overall fidelity of a digital audio file.
Sample Rate in Speech and Voice Technology
In speech processing, choosing the right sample rate balances accuracy against computational cost. Most automatic speech recognition (ASR) engines operate on audio sampled at 16,000 Hz, which captures the frequency range essential for intelligible speech while keeping file sizes and processing demands manageable. Voice AI systems that receive audio at a different rate typically resample it before inference.
Sample Rate Formula
The relationship between sample rate, duration, and total samples is straightforward: Total Samples = Sample Rate x Duration (in seconds). For example, one second of audio at 48,000 Hz produces 48,000 individual samples.