Every audio file we listen to, from a voice message to a high-resolution track, starts from an operation that is simple to describe and surprisingly deep: sampling. Understanding how it works lets you choose the right recording settings, avoid conversion mistakes and stop chasing bigger numbers for no reason.
From a continuous signal to samples
Sound in the air is a continuous variation in pressure. A microphone turns it into an electrical voltage that changes over time just as continuously. A computer, however, cannot store an infinite number of values: it has to measure that voltage at regular intervals and record each result as a number.
The sample rate tells you how many measurements are taken per second. At 48 kHz the analog-to-digital converter records 48,000 values every second, roughly one every 20.8 microseconds. Each value is called a sample.
The key point is that, if certain conditions are met, those discrete samples allow the original signal to be reconstructed exactly, not as the staircase approximation often shown in popular diagrams.
The Nyquist-Shannon theorem
The sampling theorem states that a signal containing no frequencies above a certain limit can be reconstructed perfectly if it is sampled at a rate more than twice that limit.
As a formula: if the highest frequency present is f_max, you need a sample rate f_s > 2 × f_max. Half the sample rate is called the Nyquist frequency: at 44.1 kHz it is 22.05 kHz, at 48 kHz it is 24 kHz.
Since human hearing generally reaches about 20 kHz, and that limit drops with age, a sample rate of 44.1 or 48 kHz covers the entire audible band with some margin.
The steps don't exist
Many programs display a zoomed-in waveform as a series of steps. That is just a graphical representation. The digital-to-analog converter, thanks to its reconstruction filter, outputs a smooth continuous curve that passes through every sample. If the signal respected the Nyquist limit, that curve matches the original.
Why 44.1 kHz and 48 kHz specifically
The two most common rates have different origins:
- 44.1 kHz is the audio CD standard. The value is tied to early equipment that recorded digital audio onto video media, and it has remained the reference for distributed music ever since.
- 48 kHz is the standard for video, television, film and much professional production. If you record a podcast that will also end up on a video platform, or audio for a film, 48 kHz is the most consistent choice.
In my work I recommend choosing the rate based on the final destination and keeping it throughout the project, so you avoid unnecessary conversions.
What about 96 or 192 kHz?
Higher rates do not make sound "more detailed" within the audible band: they only push the Nyquist frequency higher. They do have some legitimate uses, though:
- Extreme processing: heavily slowing down a sound (sound design) or applying large pitch shifts can benefit from content recorded above 20 kHz.
- Gentler filters: with a wider margin, the anti-aliasing filter can be less steep.
- Some non-linear plug-ins (saturation, distortion) generate high harmonics; working at higher rates, or using the plug-in's internal oversampling, reduces aliasing.
The costs are real: files two or four times larger, more CPU load and fewer tracks you can handle in real time. For spoken voice, podcasts and most music recorded at home, 48 kHz is more than enough.
Quick comparison
| Rate | Nyquist | Typical use | Space per minute (stereo, 24-bit) |
|---|---|---|---|
| 44.1 kHz | 22.05 kHz | Music, CD | about 15 MB |
| 48 kHz | 24 kHz | Video, podcasts, broadcast | about 16.5 MB |
| 96 kHz | 48 kHz | Sound design, archiving | about 33 MB |
The math is simple: samples per second × bytes per sample × channels × 60. At 48 kHz, 24-bit (3 bytes) and 2 channels you get 288,000 bytes per second, about 17.3 million bytes per minute, which is roughly 16.5 MB.
Common mistakes to avoid
- Mixed rates in the same project: importing a 44.1 kHz file into a 48 kHz session without conversion makes it play higher in pitch and faster. Modern DAWs usually convert automatically, but it is worth checking.
- Interface and system out of sync: if the operating system uses 44.1 kHz and the DAW 48 kHz, you may get crackles or the driver may refuse to open. Check your audio control panel settings.
- Repeated conversions: every sample rate conversion is a form of processing. A good conversion is nearly transparent, but chaining many of them makes no sense.
- Believing higher is always better: a 192 kHz file made from a 44.1 kHz recording contains no extra information.
In short
- Sample rate is the number of measurements of the signal taken per second.
- The Nyquist theorem guarantees exact reconstruction if you sample at more than twice the highest frequency.
- 44.1 kHz comes from the CD, 48 kHz from video: choose based on the final destination.
- Higher rates only help with specific processing and cost storage and CPU.
- Keep a single rate per project and avoid repeated conversions.
