While software plug-in emulations of compressors and gates may include options for external side-chain access, or for filtering the side-chain signal, it seems that many users are ignoring these features because they're not entirely sure what they are for. Essentially, they do the same job as their hardware counterparts and this short feature aims to explain exactly what a side-chain is and how external access and filtering may be used.
Both gates and compressors normally work by monitoring the amplitude or intensity of their own inputs, so that they can detect whether the level is above or below the threshold value chosen by the user. While some classic hardware compressors use an alternative system, where the threshold is fixed and the user instead sets the amount of compression by adjusting the input drive level, the outcome is much the same: if the signal rises above the threshold, the compressor or gate responds.
The part of the circuitry that monitors the input level is known as the side-chain, and it controls that part of the circuitry that adjusts the gain of the main signal path. In previous articles I've likened this to an engineer who listens to the sound level coming over his monitors and then makes adjustments using a mixer fader when the sound gets too loud.
In most situations the side-chain monitors the input level, although there are some designs, particularly limiters, that monitor the output level instead. Although it is significant to the technical operation of the device, this circuit topology detail doesn't greatly affect the way the user sets up the compressor.
One very simple way of modifying the behaviour of a dynamics processor is to apply high- or low-cut EQ to the signal before it reaches the side-chain. This doesn't affect the tonality of the main signal path but it does affect the way in which the side-chain responds to what it hears. In the case of a gate, side-chain filtering is often built-in and is used to make the gate less sensitive to the extremes of the audio spectrum so as to prevent, or at any rate reduce the likelihood of, false triggering from sources such as microphone spill. The best way to explain this process is to consider an example.
Let's assume that we have a microphone on a tom-tom that is part of a drum kit. We have set up the gate to get rid of the low-level ringing that often occurs as a result of sympathetic resonance (usually when you hit the kick drum). Because this resonance is at a fairly low level, we can set the gate threshold so that the gate closes shortly after the tom has been hit. So far so good. But if you have a cymbal near the tom and hit it hard, the amount of spill into the tom mic might be enough to cause the gate to open. By putting a high-cut filter before the side-chain, we can EQ some of that cymbal sound out of what the side-chain 'hears' so that it is less likely to cause the gate to open in error. Yet, when the gate opens, you can hear the tom-tom without the EQ.
Similarly, if the bass drum was being picked up in the snare mic and causing the snare gate to open unnecessarily, you could put a low-cut filter before the side-chain input to reduce the risk. Drawmer gates have both high- and low-cut filters with controls to vary their frequencies, so that the gate essentially triggers from the frequency range bounded by the two filters. Material falling below the cutoff frequency of the low-cut filter or above the cutoff frequency of the high-cut filter is largely ignored. This filter arrangement has since been adopted by a number of other hardware and plug-in gate manufacturers.
To set up the filters most effectively, you need some way to hear what the side-chain is hearing and this is usually provided by a 'side-chain listen' or 'key listen' switch. When this switch is active, you hear the filtered version of the signal, enabling you to adjust the filters to pass as much as possible of the wanted sound and as little as possible of the spill that is causing you problems. Once the filters are set, you can turn off the key listen and adjust the gate threshold and release time to get the best results. It is important to realise that the side-chain signal doesn't have to sound 'nice', as nobody is going to hear it. What you are trying to do is emphasise the amplitude difference between the strongest part of the wanted signal and the unwanted ones, and that usually demands very severe and harsh EQ settings.
In the case of drums, don't worry if the closing of the gate sounds a little unnatural on its own when soloed or if the occasional drum or cymbal breaks through, because these artifacts tend to get completely hidden when you add the drum overhead mics (which should not be gated). What you are trying to achieve is an improvement in separation and control. Although the individual elements may not sound perfect in isolation, the end result should sound much better than if you chose not to use a gate at all.
Side-chaining In Popular Sequencers
Users of Sonar and Cubase SX face more challenges when it comes to side-chains and neither application currently offers the same direct routing as those mentioned above. While there have been some innovative plug-ins that use two components, to set up a side-chain source (FXpansion's Vocoder being one of the earliest examples), these are few and far between.
Several Cubase users have expressed their frustration at this on Steinberg's online forums and there have also been several discussions on the SOS forums, focusing on workarounds to allow side-chaining within Cubase SX. The technique is not always immediately obvious, but in the Q&A section of SOS September 2005, Sam Inglis explains a workaround using quadro or surround channels. Martin Walker also explains in the April 2003 gating article mentioned above how you can use Steinberg's MIDI gate plug-in as an alternative solution. Promisingly, Steinberg appear to be responding to the groundswell of user feedback: while there is no mention of side-chains in the new features list for Cubase 4, their online FAQ states that the new VST3 standard supports side-chaining and, while the functionality does not feature in the first release, it will be implemented 'within the Cubase 4 product cycle'. Watch this space! Matt Houghton
Rather than feed the gate's side-chain from its own input, you may find that you have an external side-chain input that allows you to control the gate from a separate signal source (usually another audio track in the case of a plug-in). One way to visualise this is to imagine an engineer adjusting a fader according to the level of a track he's hearing over the monitors but the fader is in fact controlling a different track altogether (see diagrams, right).
One of the oldest examples of how this might be useful is to place the gate in a track that has an over-busy bass guitar part, then trigger the gate's side-chain from the kick-drum track. This forces the gate to open only when a kick drum is present after which it closes at the rate you set using the release control. As a result, the only bass guitar you hear is that which plays at the same time as the kick drum. The rest gets gated out. How natural this sounds varies, depending on how the bass part was played originally, but this trick can tighten up an otherwise messy bass part. You can also feed in rhythmic signals to produce controlled 'chopping' effects: for example, breaking up a synth pad in order to create rhythmic bursts.
Yet another trick possible using external keying is ducking, although the gate must have an attenuation control that allows you to input negative as well as positive values for this to work. Ducking is traditionally done using compressors, but a number of modern hardware and plug-in gates have the necessary features to allow them to function as very efficient duckers. Normally a gate has an attenuation control that determines by how much the signal level is reduced when the gate is closed. If the attenuation can be set to a negative value, as it can in Logic's gate, then the signal level actually increases when the gate is closed. What's more, you can dial in by exactly how many dBs you want the level to increase when the side-chain input signal falls below the gate's threshold.
A good example of how this feature might be useful is placing the gate in a busy guitar track and then triggering the side-chain from the lead vocal. You can then adjust the gate so that when the vocal is present, the gate is triggered and the signal level passes through at unity gain. When the vocal is absent, the gate closes and the guitar part is boosted by however many decibels of negative attenuation you have set on the gate controls.
The reason for doing this is so that you can have the guitar part a couple of dBs louder in the absence of vocals. Then, when the vocals are present, the guitar level drops back a little to allow the vocals to be heard more clearly. The technique is exactly the same as the one used by radio DJs, where a ducker is used to drop the level of the record whenever the DJ speaks. This is a useful way of keeping the vocals audible in a busy mix without having to automate the instrument levels manually.
We can access the side-chain in exactly the same way with many compressors and, as you're probably aware, compressors can also be used to perform ducking in conjunction with a side-chain input from another track. However, the end result of filtering the side-chain in a compressor is quite different to what happens in a gate, even though the object of the exercise is still to make the device more or less sensitive to specific frequencies.
For example, to use a compressor for de-essing, we simply insert a peaking (band-pass) filter into the side-chain input path and boost the signal in the region where sibilance occurs (typically 4-8kHz). If we were to boost that range by 10dB, then the compressor would be 10dB more sensitive to those frequencies and apply gain reduction whenever those frequencies were loud enough to push the signal level over the threshold. Of course, a simple de-esser like this reduces the gain of the entire audio signal whenever there is enough level in the sibilance range to trigger the threshold, which means that over-use can lead to the vocals taking on a rather unnatural lisping quality. However, it still provides a good example of this principle.
By setting the peaking filter to around 40Hz, the compressor can be made most sensitive to very low frequencies, which may be useful when trying to reduce the subjective impact of popping on a vocal track. Although this won't eliminate pops, it will reduce them in level.
While most software plug-in compressors don't allow you to insert your own filters into the side-chain input, there are often dedicated de-esser plug-ins available that combine the necessary compressor and side-chain filters. With a suitably equipped hardware compressor, you can patch any type of equaliser into the side-chain input using conventional patch cables. Note that some hardware compressors include a side-chain insert point, which makes it even easier to patch in your equalisers.
One workaround for creating a de-esser, using a compressor plug-in with an external side-chain input, might be to copy the audio being processed to another audio track, and set up the compressor to access that as its side-chain source. Then you use a filter plug-in on your newly created side-chain control track to locate the problem area, such as sibilance, and boost those frequencies as much as possible, taking care not to allow the levels to clip. You can use your bounce function to make this EQ change permanent if your DAW takes its side-chain feed before any insert plug-ins added to the control track. Now the EQ'd track will control the compressor so that it is most sensitive to the region you have boosted.
Compressors are also widely used for ducking, but in my view they are not as predictable in this role as gates. Why? The amount of gain reduction applied by a compressor depends on the ratio you set, and on the amount by which the input signal rises above the threshold. If you control the ducking process using a vocal track as the side-chain input, the most gain reduction will be applied when the vocal is loudest, when what you really want is either more gain reduction to drop the backing parts when the vocal is quieter or a system that always drops the level by the same amount once the vocal track rises above the threshold. With a gate you can be very precise as to how much gain reduction is applied but using a compressor it will always vary to some extent. However, the higher you set the ratio, the less the amount of gain reduction will be affected by the absolute level of the side-chain signal, as long as it is high enough to cross the threshold. For this reason, a compressor with a ratio control that goes right up to hard limiting is the best choice for ducking.
A potential problem when using either a gate or compressor for ducking is that the 'duck' can't happen until fractionally after the controlling signal has started. This can be fixed in a DAW by advancing the control track in time so that the duck happens fractionally early, and setting the release and hold times so that the 'un-ducking' rate is appropriate. You might also consider using plug-ins that have an inbuilt look-ahead facility, although most will probably not look ahead far enough for this purpose. Of course, level-automation is used to do this manually and more precisely in many cases, so duckers are not quite so indispensable as perhaps they once were.
Most of the time, compressors and gates are set up to process their own input signals, so all you have to do is patch them into an insert point. I hope this brief introduction to side-chains will inspire you to explore some of the more esoteric applications of gates and compressors, whether hardware or software.