What Is Sample Rate and Bit Depth? Do They Matter?
With HiFi services like Tidal, Qobuz and even Apple Music offering 24-bit depth audio with sample rates as high as 192 kHz, the average listener might be wondering: what does any of this mean, and does it really matter? Incessant audio-nerd bickering aside, let’s break down what these concepts are and what the numbers are telling us. If you’re new to all of this, I’ll be trying my best to put all of this into easily understandable language and examples.
What is it? Not to be snarky, but it is literally the rate at which audio is sampled. A 44.1 kHz sample rate means that for every one second of music, there are 44,100 individual points at which frequency information was mapped. These points are then played in rapid succession over the course of one second.
Let’s use an relatable example to clear that up: sample rate is to audio what frame rate is to video. For example, most TV shows have 24 to 30 frames (individual still pictures) composing 1 second of video. For sports broadcasts, 60 frames per second is common. Because these frames are played rapidly, our brains interpret the phenomenon as continuous movement rather than individual instances.
Let’s run with the image example for a little longer to illustrate the next point on sample rate. How many frames per second (FPS) does the human eye see? There seem to be two answers to this question: A) 60 FPS is our upper limit, or, B) some people might be able to process more than 60 FPS. So for the purposes of simplicity, let’s just say the human eye can’t discern 60 FPS from 300 FPS like it can discern 60 FPS from 10 FPS, and most of the scientific community would likely agree.
If 60 frames per second is the upper threshold of our vision, how many samples per second serve as the upper threshold for our ears? The answer to that would be 40,000 samples per second (40 kHz), or, to keep things in terms of existing file formats, 44.1 kHz. Remember this, I’ll come back to it in a moment – but first, let’s briefly go over Nyquist Theorem.
Here we have one of the core commandments of audio: sound waves can be accurately and realistically represented so long as samples can be taken at twice the speed of the highest frequency soundwave.
Again, let’s simplify that concept, or at least spell out its real-world application: the upper threshold of human hearing is 20 kHz – that is, waves that repeat 20,000 times per second. Thus:
20 kHz (Shortest/highest pitched wavelength within the range of human hearing)
2 (samples per wavelength needed for accurate sound reproduction)
40,000 samples per second (40 kHz) to accurately reproduce the whole range of human hearing.
This has, for some reason, become a controversial point to make in some audio communities. Higher sample rates such as 96 kHz and 192 kHz are used for recording music, due to an acoustic effect known as aliasing that happens as a result of the phenomenon in ultra-high frequencies, which ends up effecting audio in an audible range. That is a can of worms that I won’t be opening here – my point is that when it comes to simply sitting down to listen to music, a standard 44.1 kHz sample rate can, in theory, accurately deliver every single frequency your brain and ears can possibly process.
Sample rate is responsible for accurate frequency reproduction. Bit depth, on the other hand, is responsible for accurate volume reproduction. In slightly more technical terms, we’ll refer to this volume accuracy as Dynamic Range. Dynamic range is the amount of space between the loudest and quietest sound in a recording. For example, 16-bit audio has 96 dB of dynamic range, or, we can say: the absolute loudest sound in a file will proportionally be 96 dB louder than the quietest sound in the same file. All volume dynamics can be imagined on a scale from 0 to 96dB in 16-bit recording.
24-bit audio unsurprisingly has a greater dynamic range: volume is scaled from 0 to 144 dB.
Let’s go back to a visual parallel to better understand this. If sample rate is like frames per second, what is bit depth? Perhaps we can think of it as resolution: 720p, 1080p, 4k, etc. In videos and photos, 1080p denotes that 1080 pixels compose the vertical dimension of a photo. 1080p resolution will look sharper and more realistic than a 720p resolution, as more pixels are being used to represent an object, thus resulting in a finer depiction of color gradients and things of that nature.
Now instead of color gradients, think of volume gradients to imagine bit depth. Audio with a 16-bit depth contains 65,536 steps, while audio with a 24-bit depth has 16,777,215 steps (don’t worry about what exactly steps are, just think of them as a sub-unit of a bit). This gives 24-bit audio a much finer precision when it comes to accurately representing delicate volume ratios (dynamic ranges) between, for example, the sound of a pick hitting a guitar string and the ensuing C note that rings out, or the dynamic range between that same C note fundamental and its higher harmonics.
I noted earlier that human psychoacoustic limitations prevent us from truly hearing a difference between a 44.1 kHz sample rate and a 96 kHz sample rate. There may not be something quite as concise as Nyquist Theorem when it comes to just how perceptible differences in bit depth resolution are, but it’s a quandary that is certainly worthy of consideration.
So can you enjoy high-quality music in a standard 16-bit depth and 44.1 kHz sample rate? The answer is decidedly yes. The answer to whether or not, or how much, 24-bit audio improves quality is a little more up in the air than whether or not higher sample rates improve audio quality for listening purposes (most likely they do not). Don’t fall for manufactured pressures to make sure your favorite tracks have a 192 kHz sample rate and 32-bit depth. Whether you think you can hear the difference or think it all sounds exactly the same, trust your ears and get back to enjoying your music.
Curious about incorporating these concepts into your listening setup? Check out the extensive selection of DAC’s and DAP’s available from Audio46.