Understanding Time Constants in FabFilter Pro-L 2

Understanding Time Constants in FabFilter Pro-L 2

FabFilter’s Pro-L 2 is one of the most popular limiters in use today, and for good reason. It sounds great, it’s faster to use than some other limiters, and it has a wide range of advanced features if you want to get tweaky.

That said, certain aspects of it are widely misunderstood, and not clearly explained. This leads to results that often sound smaller, squishier, less dynamic, and less alive than they could be.

The attack knob is the most confusing part of the plugin. It’s tempting to assume function akin to the attack time on a traditional compressor, where it controls the slope of gain reduction, and making it longer allows more transient to pass through.

A bit of critical thinking suggests that can’t be the case, though: If more transient *did* pass through, how would the limiter still limit at the ceiling? How would it effectively limit at all?

(There’s a longer discussion to have, about how “more transient” is often automatically assumed to mean “better sound,” and why that’s frequently false in practice. But that’s for another time.)

Instead, the attack time is a crossover point; a digital fork in the road. 

Pro-L 2 contains two limiting stages: One for short transients, another for longer sustained peaks. This is clearly explained in the documentation (First sentence on page 13), and it makes sense in practice. Limiting applied to short peaks, where only a few samples are above the output ceiling, can release very quickly, maximizing level while being mostly transparent. Longer peaks often benefit from longer release times, which minimize distortion.

What’s less clearly explained: The transient stage of the Pro-L 2 releases near-instantly, *regardless* of the release setting. Sustained peaks, on the other hand, *are* governed by the release setting. 

FabFilter’s documentation hints at this: “The Attack and Release knobs control how quickly and heavily the release stage sets in. Shorter attack times will allow the release stage to set in sooner; longer release times will cause it to have more effect.”

But that’s vague and confusing. None of my colleagues or students, not one, have correctly understood the mechanics of the plugin just from reading the description.

The attack, in fact, governs Pro-L 2’s determination of *which* peaks are considered transients and *which* are considered sustained. Any peak shorter than the attack time will be treated as a transient, and released near-instantly. Any peak longer than the attack time will be considered sustained, and released according to the release time. Put another way: The transient stage lasts for the length of your attack time, at which point the sustain stage takes over.

In practice, it’s not all-or-nothing. Peaks with durations approximately equal to the attack time have their limiting split between the two stages, and the more a peak’s length gets further away from the attack setting in either direction, the more it will be primarily handled by the relevant limiter stage. And the different styles are legitimately different, significantly influencing the overall response. (The “Safe” style, for example, only uses the slower release stage.)

This is easy to confirm if you look at Pro-L 2’s gain reduction display— see the examples below.

1. A short impulse, just a single excursion of a waveform, with the Pro-L 2 attack maxed to 10000ms and a 0ms release. 

As you can see, limiting is released near-instantly. This is because the peak is far shorter than the 10000ms attack setting, so it’s handled entirely during the transient stage.

2. The same short impulse, attack still maxed, release now also maxed: 

Here, even a drastic change in release time has no impact on the shape of the gain reduction. This is because the signal’s peak is still very fast, far faster than the attack setting. So all of the gain reduction is still handled by the transient stage of the limiter, and the release time has no effect.

3. The same short impulse signal, but with the attack now at 0ms:

Now, *because of the attack setting*, the release time is active. Because the attack is at 0ms, even a super-short impulse is longer than the attack time. This means Pro-L 2 moves immediately past its transient stage, and the entirety of the limiting is handled by the release stage.

4.  Again, same short impulse, but now with a slightly slower (but still fast) attack.

You can see the slightest little hint of the release shape at work- the majority of the limiting is handled by the transient stage, and the last fraction of a dB of limiting is released slowly.

5.  An 808 bass note, with 0ms attack and a moderate release.

As expected, the release is moderate, a bit slower than would be expected from the amplitude of the 808. Some gain reduction is still applied even after the signal is below threshold, and the release shape has a slightly nonlinear curve.

6.  The same 808 bass note, with attack maxed to 10000ms

Here the release action is immediate, as expected. It appears gradual only because the 808’s decrease in amplitude is gradual. The release of the gain reduction precisely mirrors the amplitude of the 808, and the limiting is released as soon as the signal’s below threshold.

Based on all of the above, we can make the following inferences, and apply them to practical work:

1. Using mostly/entirely the transient stage will result in a louder master. Because the sooner Pro-L 2 gets out of the way, the sooner the limiter’s not reducing level anymore.

2. To maximize loudness while retaining punch and dynamics, use short lookahead times, long attacks, fast releases. This is most often the best starting point for commercial music. Choosing to mostly skip the release stage means limiting mostly only the peaks, and minimizing the amount of limiting applied to parts of the signal that aren’t actually peaking.

3. Setting Pro-L 2 this way makes it close to, but not quite, a clipper.

4. That seems like it should be bad, but it isn’t. Many top mastering engineers clip their A/Ds for loudness. Others purposely use similarly aggressive limiting settings. It’s often the most transparent way to push loudness.

5. With those settings, the primary concern is accidentally creating too much crunch. If that happens, just gradually increase the Lookahead, decrease the Attack, and increase the Release.

6. But, don’t play it safe by default. Finding the right amount of acceptable distortion often increases the impact of the music, maintains more perceived transient content, and minimizes the amount of limiting applied at a given amount of gain. In short: all else equal, using Pro-L 2 means a louder result, and a less squashed sound.