In the previous article in this series, we initiated a conversation about limiting in order to get a rough idea of what limiting is, and what it’s doing. Now we will take a deep dive into the limiter’s gut and get our hands dirty!
If you are in a hurry, skip down to the TL;DR section of this article to get a summary, as well as a video explanation of the subject.
A limiter is a very particular kind of device, bridging the gap between more conventional compression processing and saturation. There is a common definition stating that a limiter is a compressor with a ratio above 10:1. This means that if your input signal exceeds the threshold by 10 dB, there will only be 1 dB above the threshold at the output.
But this definition is not satisfactory for the kind of limiter we use at the end of the mastering chain, more commonly known as a True Peak Limiter.
True peak limiting means that at the very moment where a sample has a value superior to the threshold of a limiter, it will be caught by it. You can think of it as a compressor with an attack and an RMS window equal to zero.
Now, True Peak Limiters also have an infinite ratio, which is also a very important point; an audio sample cannot exceed the threshold. Think of it as a kind of warranty implied by a mastering limiter. If you have set the threshold to -1 dBTP, the audio signal will never go beyond this value. This is why a limiter could also be seen as a saturation processor, because it hard clips the input signal. Theoretically, you could replace a limiter by a clipper to get the same kind of warranty, but the sonic results will most probably be problematic and quite undesirable.
Let’s look at what happens to very simple signals when we send them through a limiter. We will first look at the spectrum analysis of a sine wave at a frequency of 440 Hz, with the FLUX:: Elixir limiter on and off. The oscillator generates the tone at a -6 dBTP level, and Elixir has its threshold set at -9 dBTP. So, we should see a perfect -3 dB of gain reduction when Elixir is on.
Maybe you are wondering if there is any difference between the two previous pictures. And yes, there is a small 3 dB difference between the two peaks, which means that we did not add any saturation in the process.
Now, to make it a bit closer to sound we will have to handle with limiters, let’s modulate the amplitude of the sine wave. For this we have simply added a tremolo with a frequency of 4 Hz. It goes from unity gain to -inf dB.
Now, we can see that there are some additional frequencies here. These are added by the fact that the limiter is engaged and disengaged by the amplitude modulation, and its envelope adds some harmonic distortion. What should be kept in mind is that the harder the peak is, the more saturation will be added.
If you use them right, you could get a light version (none True Peak) of Elixir using FLUX:: Alchemist, Solera or Pure Compressor plugins. engage the infinite ratio option, set the delay to the same value as the attack, then play with the release and hold time to get the desired result.
Smoothing the clipping
To prevent and reduce any distortion added by a limiter, we use the envelope very much like a compressor.
Didn’t we say that a limiter has an attack of 0 ms? Well, not exactly. What we want to be sure of is that no sample can exceed the threshold. Using an attack of 0 ms is a solution but it also generates additional saturation that we want to avoid. So what could we do about it ? This is where lookahead comes in handy.
A lookahead, as the name implies, allows the algorithm to look ahead, before the signal. So, if we know in advance when the signal will pass the threshold, we could then manage to open the envelope before that happens.
Remember, because it is still, unfortunately, impossible to go back in time, lookahead will add latency to the signal.
Another way to understand it is to look at a block diagram of a limiter.
There is a detection circuit that will tell when the signal passes the threshold. In a traditional compressor this moment will trigger the envelope applied by the processing block. So, in this regard, the processing in a limiter is always kind of late. Now, the lookahead is a simple delay at the input of the processing stage.
This attack time allows for a softer clipping of the signal. It is not often seen as a parameter on the user parameter, but almost all modern True Peak limiters have this hidden under the hood. In Elixir, the attack time also depends on the input signal, to achieve a more musical result.
The release time is more straightforward to understand than the attack time, as it is the same thing as in a compressor. The release time is the time for a limiter to completely stop processing the signal once the signal goes back below the threshold. It has a strong impact on the quality of a limiter.
- Set it fast to get a snappy result, with a more saturated character
- Set it slow to get a softer result, with a more compressed or pumping character
As for the attack time, the release in Elixir is dependent on the input signal.
Is True-Peak really True-Peak ?
A True-Peak limiter will always guarantee that you never exceed its threshold. At least, as long as you never do any kind of sample rate conversion after the processing!
Remember, in a digital audio workstation (DAW), we work with a digital representation of sound. To do so, we have sampled the audio signal at a certain sample rate (44.1 kHz, 48 kHz, 96 kHz, 192 kHz, etc.). So it is possible that the original signal had, between two samples, a value of higher value. After a resampling, this value may appear and generate a value above the threshold of the limiter. This phenomenon is known as intersample peak.
To prevent this effect, many limiters use oversampling. It is often hidden under the name intersample peak detection. Using oversampling will increase the resolution of the limiter and prevent intersample peaks from passing through. In Elixir, the oversampling only happens in the detection algorithm, while always being in sync with the processing algorithm.
A limiter is a dynamics processor. It bridges the gap between compression and saturation. A mastering grade limiter is characterized by an infinite ratio and a true-peak detection. This guarantees that a signal will never exceed the threshold of the limiter.
Because a limiter has a very strong behavior in regard to the input signal, it can generate distortion. To prevent it as much as possible, plug-in constructors use a complex envelope strategy involving looking ahead with attack time, and often give the user a way to control the release time.
Alas, oversampling is often used in limiters to prevent intersample peaks from passing through the limiter.
Next article in this series
Part 3 – Advanced processing and Dolby Atmos mastering