Sample rate explained: what 44.1, 96 and 192 kHz change

Samples per second and the Nyquist limit

A digital recording stores a signal as a run of measurements of its level, taken at a fixed rate. At 44.1 kHz there are 44,100 of them every second in each channel. How finely each one is stored is a separate property, the bit depth.

The sample rate decides which frequencies the recording can hold. The sampling theorem says a signal can be rebuilt exactly from its samples as long as it contains nothing at or above half the sample rate, a limit called the Nyquist frequency. For a CD at 44.1 kHz that limit is 22.05 kHz. Doubling the rate doubles the limit and changes nothing below it.

"Exactly" is meant literally. Drawings of digital audio as a staircase show each sample held until the next one. The reconstruction filter after conversion turns the samples back into the one smooth waveform, confined to the band below the Nyquist frequency, that passes through all of them. With that filter in place there are no steps for a higher rate to make finer. What a higher rate adds is room above the old limit.

Aliasing, and the filters at both ends

Anything above the Nyquist frequency that reaches the converter doesn't vanish. It folds back into the audio band as a false tone called an alias: a 30 kHz tone sampled at 44.1 kHz comes out at 14.1 kHz, where it can be heard and can't be taken out again. So a recording chain filters away everything above the limit before sampling. Playback needs a filter too, after conversion, to remove the mirror images of the music that conversion leaves above the limit.

At 44.1 kHz both filters face the same squeeze. They have to pass 20 kHz and block 22.05 kHz, a gap of about 2 kHz, which takes a steep filter, and steep filters ring around sharp transients. Modern converters do the steep part digitally, at a much higher internal rate, where it can be made precise. A non-oversampling DAC leaves the digital filter out altogether, which the R-2R explainer covers.

Why 44.1 kHz, and where 48 kHz comes from

The CD's rate came from the machines that first recorded digital audio. In the late 1970s digital masters went onto video recorders through PCM adaptors, which stored three samples per audio channel on each video line. PAL video carries 294 usable lines per field at 50 fields a second, and NTSC 245 at 60, so both come to 14,700 lines a second, and three samples on each makes 44,100. The rate also clears 20 kHz with the 2 kHz transition band the filters needed, and the CD standard, IEC 60908, fixed it for consumer audio.

48 kHz became the rate for film, video and broadcast sound, and most higher rates are multiples of one or the other:

DSD works differently, as a one-bit stream at 2.8224 MHz or a multiple of it, and doesn't fit this table.

What 96 and 192 kHz change

A higher rate gives the filters room. At 96 kHz they have from 20 kHz to the 48 kHz limit to roll off, 28 kHz to work in, so they can be gentle. In the studio that room helps processing as well: effects that add harmonics, such as saturation, can create frequencies above the old limit, and at 44.1 kHz those would fold back as aliases.

It adds little you can hear directly, since human hearing tops out around 20 kHz in young ears and falls with age. Whether a 96 or 192 kHz release sounds different from a well-made 44.1 kHz one is argued over, and blind tests haven't settled it. A hi-res release can also come from a different master, with its own EQ and level, which makes a fair comparison between two versions hard.

The extra range has costs. An amplifier or tweeter that isn't linear at ultrasonic frequencies can turn content up there into intermodulation products that land inside the audible band. And at the same bit depth, a 192 kHz file takes about 4.4 times the space of a 44.1 kHz one.

Sample rates on a DAC, and what reaches it

A DAC's spec sheet lists the highest rate each input accepts. The Henry Audio DA 256 stops at 24-bit/192 kHz on every input, with no upsampling ahead of its chip, and its review points out that no major streaming service goes past 24/192. The Denafrips Enyo 15th takes PCM up to 1536 kHz over USB and I²S, a figure its review ties to upsampling in software such as HQPlayer with the DAC in NOS mode.

What arrives also depends on the computer. Many systems mix all sound at one output rate, so a 44.1 kHz track played through an output set to 48 kHz is resampled on the way. A good resampler does that cleanly, but the samples reaching the DAC are no longer the ones in the file. Set the output to the rate you play most, or use a player with exclusive or bit-perfect output, which can switch the rate track by track.

A few devices show you the result. The DA 256 signals the incoming rate by the color of its LED when you press its Prog button, and the Douk Audio U2 Pro, a USB digital-to-digital converter, has a front indicator that confirms the resolution being passed through to the DAC.

Spotting an upsampled file

Upsampling can't create content the recording never had. A CD master converted to 96 or 192 kHz still has nothing above 22.05 kHz, and its spectrum drops off a cliff at or just below that frequency, where the master's anti-aliasing filter cut in. A 48 kHz master shows the same cliff near 24 kHz, and a file made from an MP3 or AAC cuts off lower still, where the codec's own filter stopped. The hi-res audio checker draws the spectrum of any file you drop into it, and the analysis stays on your device.

Genuine hi-res content usually fades into the noise gradually above 20 kHz. A quiet top band proves little on its own, since many recordings carry very little energy up there, so look for the sharp edge at or just below 22.05 or 24 kHz. And when you choose a converter from the DAC guide, check the rate each input accepts against the files you own.