A bitcrusher is a form of distortion, and its purpose is to lower the quality of a digital audio signal. It creates some regnizable lo-fi effect reminiscent of old gaming consoles like gameboys or the super Nintendo. But it can also be used more subtly to add a bit of grit or a bit of warmth to your sound. Just like any distortion effect really.
Any audio signal, when it's recorded into a digital format, will be cut into small chunks. Each chunk will hold a value, and when you put all these values together, you get the waveform of the sound you recorded.
It's exactly like a photo you see on your computer screen, it's made of pixels and each pixel holds only one colour.
Let's take a moment to see how that works a bit more in detail. It will make everything easier to understand the bit crusher.
When the sound is cut into slices, the number of slice per second is called the sample rate. The default sample rate in the majority of music softwares is 44100Hz, so that's 44100 slices per second. That's a big number.
That's because to record a note properly, we need a sample rate at least twice as fast as the frequency of the note we want to record.
That's called the Nyquist limit.
So we have at least the two peaks of the waveform, and we can know its frequency.
On average, the highest frequency a human ear can hear is around 20 000Hz, so we need at least a sample rate of 40 000Hz to record it. With a sample rate of 44 100Hz, we have a little room to be safe.
I open a parenthesis
As a little trivia, in video, the standard sample rate for sound is 48 000 Hz. That's because in cinema we have 24 frames per second, so having a sample rate that is a multiple of 24 made it easier to synchronize the sound with the image.
end of the parenthesis
Now for each slice, it will record a value and the precision of that value is called the bit depth. Like in any computer, this value will be noted as a series of 0 and 1. Don't worry, you never see those number, it's just the software doing its thing.
But the thing is, the software will have a limited amount of slots to write the value of each slice. And these slots are called bits.
It it writes the value on 2 bits, it will have only 4 different options to write it.
00, 01, 10 and 11.
So the more bits y ou have to write these values, the more precise it will be.
Generally a wave file have a bit depth of 16 or 24 bits, which have a definition of respectively 65 000 and 16 millions different values possible for 1 slice.
Now, the purpose of a bit crusher is to lower the quality of an audio signal, and this is exactly how it will achieve that: it will reduce the sample rate, and the bit depth. These are the 2 main buttons you'll see on almost all bitcrusher.
So what does it mean for the sound?
When you lower the bit depth, it will introduce some stepping on the shape of the waveform.
And changing the waveform means changing the harmonic content. So this will add harmonics above the fundamental. The more you lower the bit depth, the more the waveform will look like a square wave. That's what will make you sound sound more like old video games.
For example the Sega megadrive or genesis had sounds in 8 bits and others up to 12 bits, and gameboys had them in 4 bits.
But don't let this fool you, you can make very powerful sounds at a low bit depth. The Roland tr-909, which is an iconic drum machine, had a bit depth of only 6 bits.
Lowering the bit depth also reduces the dynamic range, because you'd have less possible values between the minimum and maximum amplitude. So you can lose subtle differences in volume.
And it can also bring up the volume of the noise compare to the volume of the signal. It can also introduce some quantization noise, which sounds like white noise with a low pass filter.
If you want to have a cleaner sound, it's sometimes better to boost the volume of the sound that goes into the bitcrusher, so the synamic would be higher.
Because of this noise that can be brougth up, I prefer to have this kind of effect in synthetised sounds instead of recordings
And when you reduce the sample rate, which is also called down sampling, you will introduce a lot of artifacts to the sound.
As said before, to record a frequency properly, we need a sample rate of at least twice that frequency. So when we reduce the sample rate, some higher frequencies in the sound will become simply too high for the sample rate.
And because we don't get enough samples to describe them, they will be misinterpreted as lower frequencies. And this what we call aliasing. So it will add inharmonics to the sound in completely random places, even below the fondamental.
This create a particular sounding distorsion that is very characteristic of digital. By the way, a lot of digital distortions suffer from the same aliasing problem. Because a distortion adds high harmonics, some of these are too high for the sample rate, and so they bleed back as lower frequencies.
If you want to avoid the aliasing sound, you can use a low pass filter just before your bit crusher. Filtering out the higher frequencies will prevent them from causing problems. So you can get a much smoother distortion this way. You can even make it sound like an old tape recording. So a filter is a really good companion to a bitcrusher.
This trick is cool to create more subtle effect that can help a drum track or a vocal track pop out a bit more for example.
You can even try to put an LFO on a low pass filter or to a band pass filter. It will add and remove the random harmonics from the aliasing, at will create a distinctive oyoy sound.
If the original sound is a static wave form, lowering the bit depth will create a waveform that is static, but with steppings. Whereas reducing the sample rate will create some steppings in the waveform, but with motion.
And because the speed of this motion depends on the frequency of the note and the sample rate, it can be a good idea to resample it if you want the same motion for every note.
(correction: some bitcrushers will indeed introduce some stepping in the waveform of the signal, because that's how they operate.
But simply reducing the sample rate usually doesn't add those stepping. It would simply make less points per second to describe the waveform, and the software playing it back would recreate the waveform by drawing a curve passing by all those points.)