Are your tracks out of balance? Do other limiters leave artifacts or coloration in the processed sound? Is it possible to retain dynamic range in today’s music styles? If you seek the answers to these and other questions, read on. McDSP’s most recent release, the ML4000 plug-in, just might be the answer to these and other questions.
The ML4000 is actually two plug-ins: a high-resolution brick wall look-ahead limiter called the ML1, and a multi-band dynamics processor called the ML4. Both are designed for music, mastering, post, and live sound. Unique controls and sophisticated algorithm modes allow the ML4000 to handle a variety of material. Like any McDSP plug-in, the ML4000 sounds great, but a little knowledge goes a long way in differentiating yourself from the rest of the audio engineers out there. This Tips and Tricks article focuses on the ML1 mastering limiter basic operation followed by specific applications for mastering.
The ML1 brick wall look-ahead mastering limiter is designed to provide the ultimate in transparent limiting. The basic operation of the ML1 is fairly simple – select the desired maximum output ceiling, and then the level, or input threshold, at which signal levels start to be limited. The difference between the output ceiling and the input threshold is the maximum amount of signal increase.
The ML1 uses a very small amount of look ahead (just over 1 msec, or about 50 samples at 44.1 kHz) to minimize the distortion that would otherwise occur from instantaneous limiting. This low latency allows the ML1 to be used on other instrument busses (drums anyone?) without exceeding the limit of the automatic delay compensation of Pro Tools. The ML1 can even be used for live sound – main speaker arrays can be protected and volume maximums can be easily maintained.
To further optimize signal integrity and sound quality, the ML1 uses three stages of limiting for optimum signal peak processing. Each audio sample gets the ‘triple pass’ to insure the final output is of the highest quality and does not exceed user selected maximums.
The other controls in the ML1 – Knee, Mode, and Release – provide unique and powerful adjustments to the limiter sound itself, or what is referred to as the ‘limiter effect’. These controls allow the ML1 to realize a wide variety of responses making it suitable for any application. Let’s look at these controls at little more…
The Knee mode softens the transition between unlimited and limited signals. While the Knee control will reduce overall loudness as the Knee percentage is increased, the limiter effect becomes even more transparent. Note the output ceiling is not exceeded, even with extreme Knee control settings.
An analogy for the Knee control setting range would be the state of loudness in music production over the last few decades:
- Knee = 0% – contemporary 2000’s loud as possible sound
- Knee = 25% – mid 1990’s sound
- Knee = 50% – mid 1980’s sound
- Knee = 75% – mid 1970’s sound
- Knee = 100% – mid 1960’s sound
Now perhaps the record companies would like to stay in the contemporary range? Well, I will leave that to you folks. Perhaps another way to state the affect of the Knee control is to say that 25% to 100% is probably what the audio engineer wants, and 0 to 10% is probably what the record company prefers.
The Mode control determines overall limiter algorithm behavior. The five options in the Mode control are:
- Clean – the most transparent limiter mode. Signal levels are adjusted with the least amount of measurable distortion. Great for general mastering, and other uses where the effect of the limiter is ideally inaudible.
- Soft – Slightly louder than the Clean mode, however also very transparent.
- Smart – Intelligent limiting minimizing signal distortion while increasing signal levels more than the Soft mode. This is the best mode for the widest variety of material – the limiter uses input signal dynamics to produce the optimal response.
- Dynamic – Louder than the Smart mode, and adding a hint of compression pumping under certain conditions.
- Loud – As loud as possible while keeping signal distortion minimal.
- Crush – Louder than Loud, with some signal distortion. Think drum buss limiter for a rock band.
The Release control determines how quickly the limiter recovers from a peak event, stops applying the limiter gain reduction, and returns to a state where the algorithm is only boosting the signal by the difference between the output ceiling and input threshold. What is not always obvious is the use of the Release control to augment overall loudness and the perceived limiting effect. Short release times (less than 20 msec) increase overall loudness. Longer release times (400 msec and greater) decrease overall loudness, but also minimize the audible limiter effect.
The combination of the Mode, Knee, and Release controls provide the uses various ways to balance the overall program loudness, limiter transparency, and even audible signal distortion. The ML1 mastering limiter has something for every application.
ML1 Mastering Applications
Speaking of applications, using the information in the previous section, here are some examples of how to best use the ML1 configuration as a mastering limiter.
Track Level Balancing
Because the ML1 mastering limiter is capable of a good deal of signal level increase with minimum amounts of distortion, it does excel at making tracks louder. But before we get into the hot water of making everything louder, let’s just look at how all the tracks in a given project (good grief, should we still call these things ‘records’?) can be brought into ‘loudness alignment’ – making the approximate levels of each track near enough to each other for a consistent listening experience.
Start by choosing a good output ceiling that will be applied to all the tracks, such as –1.0 dB. Next, audition the loudness portion of each track, and see where it falls within this limit. Some tracks may already be very near (or above) the selected output ceiling level, and others will be below. Note the each tracks approximate peak levels. Now for each track, set the input threshold of the ML1 to the noted level, bounce (or process in sub real-time using the Audiosuite version of the ML1 configuration). So for example, if the ‘quietest’ song had peaks at –9.0 dB, then move the orange triangle along the input meters, or the Threshold control to indicate –9.0 dB. If the loudest song was already at –1.0 dB, set the Threshold control to –1.0 dB for that track. Once such a processing pass is performed on each track, the signal peaks will consistently be at –1.0 dB, and the relative loudness of each track a little closer to the others than they were before.
At this point we have only considered the track peaks. However it is common to have track peaks that are roughly equal, while the overall loudness is different. The above process can be applied to estimated track average levels, as indicated by the signal ballistics in the ML1 input meters.
But suppose there is a great deal of dynamic range to the tracks in the project (record!). Or perhaps you would like to retain some dynamics in the mix (surely then you must have made some records). Here’s an alternative that nicely highlights the unique Knee control in the ML1 mastering limiter.
Choose the ‘good’ output ceiling, such as –1.0 dB again. Audition the loud portions of each track. Set the input threshold to –10.0 dB for all the tracks so the difference, in this case 9.0 dB, is the amount of signal increase it would take to bring the ‘quietest’ track to roughly 0.0 dB (you can consider track peaks or rough average levels). Now we’ll manipulate the Knee control to bring level balancing to all the tracks. For tracks that are already near the –1.0 dB limit, choose a knee of 75%. For those tracks that have the lowest levels, use a Knee value of 0 to 25%.
The tracks with the lowest overall level will be increased the most, while the tracks that were already the loudest will actually be reduced a little in overall loudness. Using the Knee control, the dynamic range of the ‘quiet’ tracks is preserved – the limiter is only boosting those tracks, but not applying much limiting as the input threshold is at or just above the signal peaks in those tracks. The fact that these tracks are using a hard-knee is ok – very little of the signal is getting limited. Meanwhile, the ‘loud’ tracks are hitting the limiter much more, and passing through a soft-knee that just might return some signal dynamics to the track, while bringing it into balance with the levels of the ‘quiet’ tracks by reducing overall loudness.
For one more variation, try the same trick with the release control. Choose an output ceiling of –1.0 dB, and an input threshold of –9.0 dB. For the ‘loud’ tracks, choose a longer release time (500 msec or greater), while the ‘quiet’ tracks are processed with a much faster release (20 msec or less). Now the limiter recovery from the ‘loud’ track signal peaks takes longer, so these tracks overall loudness is reduced. The ‘quiet’ tracks’ signal peaks recover in much less time, and so overall loudness in increased.
With all these scenarios, I suggest sticking with the same Mode (Clean, Soft, Dynamic, etc.) for all the tracks. Also, if the goal is transparency, I do not recommend the Loud or Crush modes.
Volume Increase without the ‘Limiter Effect’
One of the things folks do not like about limiters is that they sound like limiters. The ‘limiter effect’ can distract even a casual listener, and/or alter the original performance in an unintended and undesired way. The ML1 mastering limiter is designed to minimize this effect, however when you find yourself in such a situation, increase the Knee control value for a soft-knee approach to limiting. Knee control values of 25% or more can reduce the ‘limiter effect’ significantly. Also try the Clean, Soft, and Smart modes to get the most transparent performance out of the ML1 limiter.
Make It Really Loud
There are still times where some tracks need to be as loud as possible. Try a Knee control between 0% and 20%, and go for the Loud or Crush modes. While these modes maximize signal levels, they do not minimize signal distortion as much as the other modes. But then that may be what you want. Do not overdo the input threshold control, especially in the Crush mode. The input threshold should be at or just above the input signal peaks. If input levels are already near the maximum 0.0 dB, try to stay at an input threshold higher than –12.0 dB (i.e. in input threshold no more than 12.0 dB below the input peaks).
Soft to Loud Transitions
One of the most difficult things in making a good limiter is creating one that does not ‘plunge’, or scoop out short portions of the audio as the music transitions from a quiet passage to a loud passage and back. And even when the limiter algorithm is capable of achieving the desired ‘non-plunge’ result, there is still some required user setup.
Observe signal peaks and rough average levels during the quite passage. Set the input threshold to be roughly equal to this level. Choose a ‘good’ output ceiling, such as –1.0 dB in the example above. Now during the quiet part, the limiter will boost the signal by the difference between the input threshold and the output ceiling. During the loud passage, the limiter will transition to its transparent ability to preserve material that is already ‘loud enough’.
Now there are some folks that would point out ‘this is exactly the problem in today’s music – there is no quiet or loud part!’ They would be correct. But once again the ML1 comes to the rescue with its unique Knee control. Take the Knee control, which I have assumed has been set to 0% so far in this example, and move it to 25%. Audition the quiet to loud transition again. Try the same thing with 50%, 75%, and 100% Knee control settings. As the Knee control is increased, a more gentle effect is heard, and by the 100% setting, the limiter will likely sound more like a compressor. Go for a Mode setting of Smart or Dynamic in these cases as well. Sure results may vary, but clearly the opportunity to preserve some dynamics while achieving consistent peak levels is within reach using the ML1 mastering limiter.