When it comes to mixing drums, few things have studio newcomers tearing their hair out more often than compression. Is a compressor supposed to even out the levels in a performance? Or lengthen sustain tails? Or make individual hits more punchy? Or pull up low-level details? Well, compression could do any of these things, or none of them, depending on how you set the compressor up — hence all the confusion!
In this article I’d like to clarify matters by identifying a series of distinct drum-mixing tasks for which engineers commonly use compressors. I’ll explain in each case what compression settings you might choose for best results, whether the drums you’re working with are one-shot samples, loops, virtual instruments, or live recordings. And I’ll focus specifically on tackling the kinds of undesirable processing side-effects that blight many project-studio mixes. As we’ll see, this can require more sophisticated compression methods, but I’ll also suggest various alternative remedies that can bail you out when compression proves unable to deliver the goods.
Another frequent source of head-scratching amongst less experienced engineers is that so many real-world compressors have different control sets. For this reason, I’ll restrict myself to the most common configuration, found in compressors bundled with pretty much every mainstream DAW, which is based around five main controls: Threshold, Ratio, Attack, Release, and Make-up Gain. However, I won’t go over the very basics of what these controls do (check out my ‘Compression Made Easy’ feature in SOS September 2009 if you need a refresher: http://sosm.ag/compression-made-easy). Instead, I’ll concentrate on how best to set them up in practice for each specific drum-processing task.
Let’s start with the most common function of compression, whether you’re dealing with drums or any other instrument — to automatically even out unwanted level variations. In drum terms, that means turning down louder hits to more closely match softer ones, so the processed part can be more stably and reliably balanced against other instruments. Here’s how you can set up a compressor for this purpose, assuming for the moment the simplest scenario, in which I’m dealing with a single ‘instrument’, either an acoustic drum recorded live, or some kind of programmed electronic sound.
First, dial in attack and release settings of around 1 and 100 ms respectively, and choose a ratio of, say, 3:1. Then reduce the threshold until you see activity on the gain-reduction meter for all but the softest hits. You’ll almost certainly want to finesse the ratio based on what you hear. Some hits still too loud? Increase the ratio. Dynamics flattened too much? Then lower it.
While this should yield a sound that holds its place in the mix more solidly, the job’s not complete — you must also make the best of the attack and release times. For pure balancing applications, go for an attack time as fast as you can get away with, so that it catches the fast-moving transient at the onset of each hit. However, many compressors can respond swiftly enough to put kinks into the start of a drum’s audio waveform, altering the drum’s attack timbre, especially when you’re processing the slower-moving waveforms of low-pitched sounds. So make a point of listening to the attack portion of the processed drum signal while you massage the attack time, and actively choose a timbre you favour.
It also makes sense for the release time to be as long as possible, so the compressor’s gain-reduction reset occurs mostly after the sustain tail of each drum hit has finished. Otherwise, you’ll change the drum’s decay characteristics. If you set the release too long, however, the gain-reduction won’t reset fully between hits, thus weakening the compressor’s level-correction effect. So, in practice, you’ll almost always need to adjust the release to suit the tempo and complexity of your drum part. For example, you’ll likely want a much shorter release time for a fast 16th-note tambourine than for a rock ballad’s ponderous back-beat snare. Bear in mind, too, that both attack and release times may impact upon the amount of gain reduction, so don’t be scared to revisit the threshold and ratio controls as necessary.
One of the most powerful ways to stress-test any processing decision is to toggle the processor’s bypass switch and directly compare what you’ve done with the untouched signal. What complicates this for compression, though, is that compressors sometimes make the processed output hotter, and most of us have an instinctive bias in favour of louder sounds. It’s therefore very important that you use your processor’s make-up gain control to match the subjective loudness of the compressed signal with that of the bypassed signal before critically evaluating your efforts. If you don’t, your ears may vote for the loudest signal, and not necessarily the best!
But, given that the very nature of compressors is that they dynamically alter the loudness of the processed signal, there’s also a choice to be made about which elements of the processed signal you should loudness-match with the bypass signal. For instance, in the simple level-balancing setup I described above, it would be a mistake to try to match the loudest hits in the compressed signal against the loudest hits in the bypassed signal, because this would mean most of the hits in the compressed signal would end up sounding louder. By loudness-matching the softest hits of the two signals, on the other hand, the compressed sound will end up being quieter than the unprocessed sound most of the time, so there’s no way the loudness bias can fool you into accepting processing unless it really does sound better. Of course, there’s always a risk here that the loudness bias may dissuade you from a pursuing a truly workable compression setting from time to time. But I’d say that’s the lesser evil, since project-studio mixes on the whole suffer much more frequently from over-processing than under-processing!
Incidentally, many plug-in compressors provide an ‘auto make-up gain’ facility, which aims to automatically loudness-match the processed and unprocessed signals. While the underlying goal of such functions is a noble one, and can work well with less percussive sources, I personally find them all but useless for typical drum-processing applications, regardless of their sophistication. The reason for this is that their fundamental goal is to match overall loudness, whereas I almost always want to match the loudness of some specific element of the signal: the softest hits in the above level-balancing task, for instance, or the levels of the transients or sustain tails in other circumstances, as we’ll see in a moment.
Trouble Letting Go
As useful as that straightforward level-balancing setup can be in many situations, there are plenty of real-world recordings that’ll fox it. For example, if you want to even up a pattern in which there’s a flurry of drum hits very close to one another in time, you may have to set such a short release time (so the gain reduction resets between hits) that it causes an unnatural gain ‘bump’ in the sustain character of isolated hits. In fact, if the release time gets short enough, you may even start getting audible distortion, as the compressor’s gain-change element starts affecting the shape of the signal waveform, not just its level envelope.
Faced with this choice between insufficient level control and unpalatable artifacts, too many small-studio engineers wave the white flag and resign themselves to the unwanted side-effects. A much more positive attitude is to see the impasse as a message from your compressor that you’re asking it to do something beyond its capabilities — in other words, that you should look for a tool that’s better suited to the job at hand. In this case, for instance, chopping up the part into beat-sized chunks (practically a single-click process in most DAWs these days) provides one solution, allowing you to correct the levels of the faster hits with ‘stepped’ per-beat gain offsets, so that the compressor only has to deal with the more widely spaced hits. Alternatively, SoundRadix Drum Leveler and Melda MDrumLeveler are just a couple of plug-ins that provide a very similar scheme in a more ‘set-and-forget’ format.
Another insoluble problem is mostly restricted to live recordings, and specifically those where the performer hasn’t just varied how hard they’ve hit the drum, but also the manner in which they’ve hit it, such that the timbre varies considerably from hit to hit. No amount of compression will make a spindly-sounding mis-hit reside happily alongside fuller-sounding on-the-money hits, so there’s a definite risk of over-processing if you try to address this with compression alone. Using your DAW’s audio-editing facilities to paste good hits over bad ones can take the pressure off the compressor, allowing it to concentrate on the pure level-balancing job it’s more comfortable with.
Rebalancing Within Submixes
Balancing a single-instrument recording with compression can occasionally be heavy lifting, but the tough really get going when audio contains more than one instrument — in other words, when you’re tackling things like prefab drum loops, sampled breaks, live-kit overheads, drum room mics, and rhythm-section submixes. And a whole set of challenges arises when you wish to compress one element in a multi-instrument drum track more than the others. Let’s look at a real-world example...
Imagine you have a drum loop with kick, snare and shaker, and you want to use compression to turn down the kick. Perhaps it’s a heavy hip-hop loop you want to fit into more of a pop production. Assuming the kick is louder than anything else, in theory all you need do is adjust your compressor’s threshold so that it only flexes its gain-reduction circuit in response to the kick-drum peaks. But what if that doesn’t make enough of a difference? Even with a 100:1 ratio, a typical compressor is only designed to prevent the signal exceeding its threshold level, but if you pull the threshold down beyond a certain point, you’ll start triggering compression on the snare, undesirably altering that instrument’s balance relative to the shaker.
This is where it becomes helpful to EQ the signal feeding the compressor’s level-detection circuit (or ‘side-chain’). For our drum loop, you could cut away all the high end from the side-chain signal, such that the compressor no longer ‘hears’ such a high snare-drum level, and continues to trigger its gain reduction only on kick-drum hits, even at a lower threshold setting. By the same token, if you wanted the compression to react mostly to the shaker, you might try high-pass filtering the side-chain signal; or if the snare were too loud, you could use a band-pass filter (or a more complicated EQ curve) to emphasis its mid-frequencies in the compressor’s level detector. The beauty of this system is that the side-chain EQ doesn’t directly affect the audio — all it does is change how the compressor reacts.
Side-chain EQ is so useful that it’s included on many general-purpose compressors, while others let you send audio directly to the compressor’s side-chain from elsewhere in your DAW, so you can use a separate EQ plug-in to carve away at the detector signal instead. Even if your compressor has neither facility, you can usually achieve the same ends by applying equal but opposite EQ curves before and after the compressor. So, returning to our drum-loop example, you might use a high-frequency shelf to cut 12dB above 100Hz pre-compression, and restore the missing high end with a complementary 12dB shelving boost post-compressor. In the analogue domain this kind of stunt would cost you added noise, plus the cut and boost EQ curves might not match very accurately, but the low-CPU digital EQs bundled with most DAWs can handle the task without any appreciable sonic degradation. (If yours doesn’t, try Blue Cat’s Triple EQ or Melda’s MEqualizer, both of which are cross-platform freeware.)
There are limits to what even side-chain EQ can handle, though. High-pass filtering the side-chain might stop the compressor reacting to the kick in our loop, but the snare may still out-gun the shaker in terms of level, even just at high frequencies, in which case there’s simply not enough level differentiation to allow the compressor to get an independent grip on the shaker. So what then? Well, seeing as we’ve introduced the idea of feeding an external signal into a compressor’s side-chain, why not program a MIDI drum loop specifically to trigger the compressor’s gain reduction with 100 percent reliability? It doesn’t matter what kind of sound you use for that, as long as it has a well-defined attack, because it won’t be heard in the mix.
Of course, for recorded live drumming, constructing a MIDI trigger part could be tedious, even if your DAW’s grid notionally lines up with your drummer’s playing, because in practice the performer’s natural timing will continually wander somewhat from the grid lines, and you’ll have to tweak the MIDI trigger part bar by bar. Fortunately, with live drums it’s common to use multi-mic recording techniques, so it’s usually much easier to do any internal rebalancing by simply adjusting the faders and EQ settings of individual mic channels than by processing the mixed drum-kit signal in such a heavy-duty manner.