Digital distortion types are typically the result of bit depth and sampling rate limitations - when these are reduced, it results in aliasing, clipping, waveshaping, and more types of familiar distortion. To avoid these, we can increase the sampling rate via oversampling and bit-depth by recording at least 16 bits.
The distortion types we’re covering here are often related, so we’ll also look into those relationships in this video. Also, we’ll get into some technical, kind of dry topics, but it’s worth watching if you want to better understand this stuff when you're producing or mixing.
Let’s start with aliasing distortion - aliasing is basically harmonics that form in the high-frequency range due to our sampling rate being too low to support those high frequencies. So let’s say we’re using a 44.1kHz sampling rate, and we have a harmonic that hits 25kHz, this would bounce back down.
Since the highest supported frequency with this sampling rate is 22,500Hz, or the sampling rate divided by 2, the 25000Hz frequency can’t be supported.
As a result, it “folds back” in an equal and opposite manner. The sound of these harmonics is often unpleasant, and they can cause phase cancellation to other high frequencies.
Although the effect will be subtle, let’s listen to a track with the high frequencies distorted, to cause harmonics to form above the highest supported frequency. Then, we’ll turn on oversampling which will lessen the effect of aliasing.
Sample rate distortion is very similar to aliasing - when introducing this type of distortion we’re reducing our sampling rate on purpose, reducing the highest supported frequencies and subsequently causing harmonics in the high-frequency range. Although it’s not a common distortion type, it can be used creatively.
Let’s take a listen to it.
Reducing the bit depth causes multiple forms of distortion, including clipping, quantization noise or error, and waveshaping when the rate is reduced significantly. The bit depth is responsible for approximating the signal’s amplitude - the higher the bit depth, the more accurate the approximation, or recreation of the waveform.
When we reduce the bit depth, we lower the amplitude ceiling, making clipping distortion more likely.
Next, we create a bigger difference between the original waveform and the digital or quantized waveform. This results in quantization error, which more or less sounds like noise.
Lastly, if we reduce the bit depth enough, the waveform can completely change due to clipping, and general unsupported amplitudes that the waveform had originally.
Let’s listen to it by using this bit-depth reduction plugin, and notice how significant the changes are when lowered to 8-bits and below.
We touched on this last chapter, but let’s look it again - in a digital system we’re usually working with 64-bit processing. This means our signal can go well above 0dB before clipping occurs - however, by the time it reaches the output, it’s converted back to 24-bit processing.
This is why meters in DAWs usually don’t show red when overs happen on a track or bus, but it does show as red at the output.
As for the clipping itself, it’s basically the transient being chopped off at the top, however much it goes over 0dB. As a result, we get white noise whenever this hard clipping occurs.
Let’s take a listen to hard clipping being introduced, and then listen to the delta or the null of this clipping, to hear exactly what is being changed about the signal.
Pre-ringing is a newer term, but it refers to the effect that linear phase processing has on transients - in short, when we introduce linear phase processing the signal is delayed by a set amount. Our DAW then tries to compensate for the offset timing.
In doing so, the original signal, and the delay compensated signal clash very slightly, causing mild but sometimes noticeable phase cancellation to our transients.
Although subtle, let’s listen to one track processed with zero latency, and one processed with an aggressive linear-phase setting.
You may be familiar with the term lossy file - this is basically an MP3 or similar file type in which the information has been deleted to reduce the file size. As you’d imagine, this process causes distortion - which we can emulate with this Codec plugin.
The main cause of distortion in these files is bit-rate reduction. The bit rate is a multiple of our bit-depth and sampling rate, for example, 16 bits x 44,100 samples x 2 to make the signal stereo, resulting in 1,411,200 bits per second or 1,411kbps.
MP3s range from about 128 - 320kbps. So in short, to create a lossy file type we reduce the sampling rate and bit depth, causing noise, potential clipping, a reduction to high frequencies, and potential aliasing if filters aren’t used.
Let’s listen to this codec plugin, and solely focus on the effects of reducing the kbps.
If a sample or a group of samples drop out or are deleted during the conversion process from lossless to lossy file types, we’ll notice an audible drop in the signal, which sounds somewhat like a click. We can emulate this with this codec plugin.
Let’s take a listen to what these “dropouts” sound like.
Encoding error or distortion is another term you may have heard - in short, it’s very similar to lossy distortion but it typically refers to the clipping aspect. When online streaming services convert WAV files into lossy file types, they reduce the bit-depth, causing less accurate amplitude approximations.
Subsequently, transients that wouldn’t have peaked are now hitting a lower ceiling, resulting in hard clipping distortion. To illustrate this, I’ll use this plugin by apple that lets us monitor a WAV file being encoded.
Inter-sample peaking occurs when the digital waveform is converted back into an analog waveform by a D to A convertor, which occurs in any amplifier that converts digital audio into analog for speakers, earbuds, etc. During this conversion, peaks at 0dB, or close to it, can be incorrectly read.
This results in those peaks being amplified above 0dB, resulting in clipping distortion. Fortunately, the better D to A converter you have, the less of a problem these are.
Let’s try to emulate it though with a hard clipper. Notice that the effect is subtle, and honestly, doesn’t have the worst sound.
You may have noticed that some limiters have true-peak detection and limiting functions - these were designed to reduce inter-sample peaking. Ironically, they introduce a separate form of distortion, similar to the linear-phase distortion we covered in chapter 5, in which transients are affected via phase cancellation.
With that in mind, if you see a limiter with true-peak limiting, not detection but limiting, it’s best to disable it for the sake of preserving transients.
Let’s listen to a mix limited with true-peak limiting, and one without, and let me know if you can hear a difference.