In the previous episode of this series around d/a converters, we looked at external dacs. In this article, we’re going to look at bits and sampling frequencies. What is the difference? What is required? And where are we headed in this “spec race”?
Practically all the music we listen to is stored digitally. And most of the vinyl we listen to is also initially recorded “digitally”. That doesn’t mean it sounds the same, of course, because all of the mastering for vinyl is different because the medium demands it. But now we digress.
When recording music, the signal path is usually partly analog (after all: a microphone transmits an analog signal) and partly digital. This is because it eventually passes through an Analog to Digital Converter (ADC) so engineers can mix and master. In fact, that practically never happens in the analog domain anymore. It’s too complex, time-consuming and therefore far too expensive. By the way, purely analog recordings can be found. Both on vinyl and on tape. And we have experienced that they can be beautiful.
Now recordings can be made at various qualities. From CD quality – 16 bit / 44.1 KHz- to high-res – 24 bit / 192 KHz – or even higher: 32 bit / 384 KHz for example. Not to mention: DSD. From 1-bit 2.8 MHz to 1-bit 11.2 MHz.
Usually a studio records in at least 24 bit / 44,1 KHz or for example 24 bit / 96 KHz. This gives them freedom to mix and master properly. Then from this master, new versions are made again for the various streaming services and other media. Think CDs, tape or vinyl.
On the playback side, you don’t really have to worry about the various versions. Most, modern d/a converters can handle all bit sizes and sampling rates. With the exception of 32-bit files and perhaps certain DSD files. However, those are very rare.
Sampling rate
We’ve already bombarded you a bit with jargon: masters, sampling rates, bit sizes. Just what are these? Let’s start with samplingrates.
You can think of sampling rate as the amount of pictures a video camera can shoot. The more “pictures” the smoother the video. A movie with 10 frames per second looks like a stop-motion video. But a video with 60 frames per second looks smooth and razor sharp: even in complex scenes; there is a no more motion blur.
In audio, a higher sampling rate mainly gives more detail and a bit more air. It also seems to run a little smoother. However, your author finds that the gain above 24 bit / 96 KHz is really minimal. It is incredibly difficult to observe differences. If they are even noticeable.
Nyquist
Now this is only one part of the story, because something else is at play: Nyquist. A certain Mr. Harry Nyquist figured out that the sampling frequency had to be at least twice the desired of the required bandwidth. This in connection with so-called “aliasing“.
This is why Philips chose a sampling frequency of 44,100 KHz. This gives a bandwidth of 22050 Hz. Philips chose to use a for CD bandwidth of 20 Hz to 20,000 Hz. By putting the Nyquist frequency – 22050 – outside our audible range, less sharp and less complex, digital filters can be used. Smart.
Bit depth
Then bit depth. This affects the dynamic range. The more bits, the more dynamic freedom the engineer has. 16 bits gives 96dB of dynamic range. 24 bits already gives 144dB. 32 bits would come out to 192dB dynamic range. This is of course absolutely idiotic, since with humans the pain threshold is around 120dB. So why 144dB or even more idiotic: 192dB dynamic range?
That has everything to do with the recording headroom and post-processing. If someone is recording a large orchestra, it is nice that the recording does not clip when the big drum is beaten vigorously, for example. Or when a singer is in full emotion, just too close to the microphone, and he or she sings a bit too loud.
It is also nice when sound engineers mix and master. That includes compression, eq’s and other forms of manipulation. The more leeway an engineer has, the less that gets “broken” during post-production.
These elements also come into play in photo and video editing. Recordings are also often made in a so-called “raw” format. These (huge!) files can be manipulated very heavily without losing any information. This is not the article (or platform) to go deep into video editing, but there are certainly similarities here. For example, both disciplines require recordings in very high resolution because of the processing stage.
Now we also see that more and more devices can handle 32bit floating point. Recordings in 32bit floating point can no longer clip, which of course is ideal for sound engineers. Our guess is that 32bit floating point will also find its way to the playback side. Perhaps this means – finally – the end of “digital sound”.
Dithering
Simply put, Dithering is a process by which you add noise to the digital signal. This noise – which comes in countless shapes and sizes – ensures that rounding errors are avoided. And partly because of this noise.
Dithering is mostly used when we switch between sampling rates during mixing and mastering. We interviewed Brendon Heinst of Trptk about dithering. His explanation can be found below.
What is needed?
We wrote a piece on hearing last year. In it we “calculated” how much resolution is needed to meet the dynamic range of our ear. That turned out to be around 24 bit / 96 KHz. It also happens to be a point where we find it very difficult – if not impossible – to hear differences between even higher resolutions.
Our opinion is that 16 bit / 44.1 KHz is just fine, but that it does make sense to release recordings in higher quality. However above 24 bit / 96 KHz the gain is very, very small. Perhaps some dacs are more comfortable working with higher resolution material, but that is mainly in the dac technology: not the recordings.
Next Episode
In the next episode, we will discuss jitter. What is jitter? What forms are there? And when does it affect you?