We Can't Go On... (Part 1)
...Beating like this. We start a major new series on how to make your drum machine - no matter what its spec - sound more interesting, both sonically and rhythmically. Your guides are Chris Meyer and Matt Isaacson.
Your drum machine is a sophisticated piece of technology, yet your percussion parts still sound lifeless and robotic. True for you? Don't panic - we're starting a major series on creative drum programming.
A COMMON ELEMENT in the vast majority of music we hear today is The Beat - drumming, percussion, rhythm, whatever you want to call it. The advances in technology these last few years have allowed more of us to produce more parts of the music we record, and to take over the role of the traditional drummer or percussionist to create The Beat.
Great? Not to some people's ears. What has made all this possible - the drum machine - is, next to the tape recorder, arguably the most used and certainly the most cursed piece of equipment used in the creation of music today. It's been accused variously of putting ordinary drummers out of work, turning popular music into little more than an elaborate metronome, and making modern dance a mockery.
But one thing is certain: nobody is going to give up the convenience and new possibilities inherent in one person being able to create all or most of the parts of a piece of music. So the metronome goes on.
The aim of this series is to make all of us (we hope) better pseudo-drummers and percussionists by pointing out common traps and pitfalls, and giving good hard advice on how to make both our percussion programming and our percussion sounds more lively and more authentic - or failing that, at least a lot more interesting.
DRUM MACHINES JUST don't sound like drummers. It's nothing mystical or intangible - there are honest factual reasons why this is so, whether or not you've been able to put your finger exactly on them (and don't feel bad if you haven't - many, many others haven't, either). Following the idea that to Know Thy Enemy is the best way to conquer (and that's what this series is all about), it's a worthwhile exercise to spell it out.
Analogue drum machines (of which the Roland TR808 is the last example still in common recording-studio use) don't sound right simply because their electronically generated sounds fall so far short of simulating the characteristics of real percussion instruments, even your deaf old Aunt Mary could tell the difference. That hasn't stopped these machines finding a niche for themselves in modern music, of course, and even the limitations of their programming facilities can be got round by sampling those analogue voices into a digital drum machine.
But on digital drum machines, the short length of sampled drums is often just as big a problem. The type of memory that drum machines use to store sounds is expensive. True, the price of memory has been coming down, but so has the average price of drum machines - and amount of memory (and, therefore, length of sounds) is one of the first things to get cut.
The first noticeable effect of this is that sounds in drum machines end up far shorter than their original parents - pale imitations of the real thing. Toms go "blat" instead of "thooommm", cymbals go "tish" instead of "tiiishhhhhh" (usually cutting off abruptly instead of decaying smoothly into silence), and so on. Open hi-hats are rarely long enough for the closed hi-hat to actually cut them off. It's like a person abruptly cutting off a word at the end of a sentence, instead of winding down his or her inflection - subtle, and not something we think about every day (if every year), but a psychoacoustic "hint" that our ears and brain notice.
Ambience is another lost property — the sound of the room around the drum and of the drum itself shuddering to a stop. Most of this shows up after the bulk of the sound itself has died away, and ends up getting trimmed off. This makes many sampled drums sound unnaturally close, which tends to work against the volume level of the drum or percussion device itself — if something is soft, you expect to hear as much reverb as the source sound itself. If a sound is extremely high in level, this usually implies that it has been hit harder, which also tends to excite more reverb in the room in which it was hit (and more shuddering in the drum shell).
In some cases, ambience is purposely left out of the sound. Since drum machines are in one sense trying to be all things to all musicians, most of them come with standard issue, middle-of-the-road sounds. Aside from memory constraints, the sounds are recorded in such a way as to strip them down to their generic essentials, using head damping, isolation, and close miking in acoustically dead rooms - in a word, dry. In multitrack recording, drums are often recorded this way with the assumption that ambience will be reinserted during mixdown - ambience-free tracks leave the widest number of options open. Perhaps unfortunately, the same assumption has been applied to most drum machines.
"The memory that drum machines use is expensive. True, the price of memory is coming down, but so is the price of drum machines - and memory is one of the first things to get cut."
A third issue related to lack of memory is the data format used to store sampled drum sounds. Memory that is eight bits wide costs two-thirds as much as memory 12 bits wide (the current accepted standard for sampling instruments). Linear eight-bit data has very little dynamic range - only 48dB, or less than the difference between silence and normal conversation. Turn it up, and more noise creeps in (plus a digital audio phenomenon known as quantisation noise, where the quieter the sound gets, the more distortion and noise creep in).
Most digital drum machines use an encoding scheme known as eight-bit COMDAC, where 12 bits of dynamic range - 72dB - are squashed into eight bits of data. This means there are some holes in the range of volume a drum or percussion sound can have as it fades away, which often results in some high-frequency noise. So sounds without a lot of high frequencies such as toms, congas, and bass drums need to be heavily equalised (filtered) to cut out this noise. The sound ends up cleaner, but dull. Listen closely to a drum-machine tom next to the real thing - it lacks the "whack" of the initial hit, and much of the sustaining character of a real tom.
Next is an effect that's caused both by lack of memory, and by few people realising that it is that important - multiple copies of the same sound. You can hit a cymbal in many different places, and get many different sounds. Next time you're next to a drum-kit, listen to the way the cymbal wavers on the stand when you hit it again, even if you are trying to hit it in the same place. Some drum machines use the same tom transposed to different pitches to fake having multiple toms. True, if toms were perfect, different sizes of the same manufacturer should sound alike — but they're not, and they don't. And any ambience that does manage to get left with the original sound gets transposed along with the rest of the sample - so the authenticity of such sounds suffers. The same goes for retuned snares — making a real drum head tighter doesn't change the size of the shell or the snares that vibrate against the head themselves, it's just that retuned samples make it sound that way. And a hi-hat has a virtually infinite range of sounds, depending on how hard it is clamped shut. If at least one of the authors hears one more string of semiquaver closed hi-hats without any dynamics being wilfully created, he will murder.
Ah, yes - dynamics. Many, many drum machines do not take any form of dynamics into account. No one short of a robot can do anything precisely the same way over and over again. Even if a drum machine lets you adjust the levels of the sounds against each other, you often can't change the levels of any one sound throughout the piece or on any given take, though some machines, such as the Korg DDD1 and Kawai R100/R50, feature force-sensitive pads.
But volume is only a part of dynamics - drums and percussion instruments are complex beasts. When you hit something harder, not only does it tend to get louder, but the timbre of the sound changes (more stick crack, more snare, whatever), and even the pitch itself changes subtly from the skin head getting stretched further (increased tension equals increased pitch). Where you hit the drum can also have as much effect on pitch, and even more on timbre. Once again, you may not know this, but your ears do.
The "robot feel" of drum machines does not apply just to the sounds themselves - timing comes into play, too. Whether they can help it or not, very few people can play at a perfectly constant tempo (on any instrument). And there's a lot to be said for the subtle placement of hits just ahead of or behind the beat - some drummers have tracked this down to just a few milliseconds, which is finer than the vast majority of auto-correct functions available (or even what a drum machine or sequencer claims to be "free time").
Drum machines also play back sounds differently than the sounds themselves "play". When you hit any drum or percussion instrument that has not finished sounding from the last time you hit it, the energy and the state it was in affect how the new strike excites the drum (and, therefore, how it sounds). The more "loose" the physical device, the greater this change (ie. crash cymbals sound different every hit; items such as woodblocks or xylophones nearly always sound alike on multiple strikes). No electronic percussion instrument yet developed is able to take this properly into account. Most drum machines just restart the sound from scratch, which just isn't the same.
Another thing that just isn't the same is the way we play the sounds on a drum machine. To be fair, Roger Linn did us all a great service when he allowed us to hit pads in real time on the front of a drum machine as a way of programming drum patterns — it's certainly more interactive than having to chart out patterns on paper and in our head, and then enter them by hand into a computer. However, hitting small hard plastic squares with our fingertips isn't quite the same as a drummer or percussionist swinging sticks at, or whacking hands across, the real thing. Well-trained is the person who can do a tom roll on one pad with two fingertips. Often, this means entering the more difficult, or faster, parts in step time.
"There's a lot to be said for placing bits ahead of or behind the beat — some drummers have tracked this down to a few milliseconds, which is finer than the auto-correct functions on machines."
Neither method encourages playing drum or percussion parts the way a drummer or percussionist would. Result? We get boom/chock/boom/chock bass and snare patterns, as opposed to all of the little half-hits and flams more typical of the Real McCoy.
Then there is the important, and quite common, crux - most of us aren't really drummers. Therefore, we do not program drum patterns the way a real drummer or percussionist would. Now, some of you may comment that this is a cop-out, and that it is equally valid to do things a drummer might not think of or even can't do. No disagreement there. However, many are still incorrectly imitating real drummers, and like it or not, there is a whole tradition of how we're used to hearing real drummers sound that sends off alarm bells in our heads when somebody does something blatantly wrong.
SO, WE'VE DISCUSSED some of the things which can make a rhythm track sound synthetic or unexpressive — things that give it away as sounding "not real". We've also seen that many of these things are inherent to drum machines, particularly older ones or those in a modest price range.
Can we get around these problems? Well yes, but there's some bad news in advance - a lot of the solutions may mean getting instruments or other devices you don't already have. If you have not yet made an investment in electronic drum and percussion devices, then this series is just in time. If you are looking to upgrade, perhaps you can pick up some advice on better (and occasionally, non-obvious) ways to spend your money. And in any event, drum machines themselves are getting better, and are starting to address some of the problems listed above.
But what if you already have one of the poor, broken-down drum machines we've been maligning from an ivory tower in the above paragraphs, and don't see many extra pound coins floating your way in the near future? Well, the rest of this instalment is dedicated to you.
If you find yourself in the position of being constrained to work with these devices, you may at this point be wondering whether you are doomed to suffer with rhythms which always fall somehow short of what you wanted. The answer? Depends. Honestly, there isn't much hope of making a £250 drum box on its own sound like Billy Cobham behind a large Tama drum-kit - with such a goal, it should be obvious by now that satisfaction shall not soon be yours.
On the other hand, many recording applications are not so critical. Much of the ability of a reasonably good drum machine to simulate recorded drums can arguably be credited to the way in which real drums and drummers have been used in the past in pop recordings - as providers of metronomic rhythms and muffled sounds meant to stay well in the background, especially when vocals are the main attraction. In these musical situations, a producer may opt for a mediocre drum box program in favour of a more expensive studio drummer, with relatively little effect on the final recording or the ability of the listener to tell if it is a drum machine (or a drummer who sounds like a drum machine). And anyone who has been around real drums knows that they tend to sound a bit different from what you hear on recordings. This is no accident - a lot of work is done to get them that way.
The good news is that this points to one approach to treating our problem on the drum box itself - namely, avoid programming techniques that cause rhythms to jump out of the mix and say, "I'm a machine". This calls for some discipline, and careful listening to both the sound made by your box and to any recordings you may be using as musical or stylistic references. The goal here is to avoid two common problems: sounds that are obviously not created using real drums; and rhythmic arrangements that no human drummer would or could be expected to play.
"Retriggering a sound on drum machines cuts off the previous hit and starts it all over again. And if you play fast enough, you get to a point where only the very beginning of the sound is heard."
Here are a few specific cases, starting with fast licks and rolls. Now, the snare sound in your beat box probably sounds remarkably like a well-recorded medium-loud single hit on a perfectly set up, isolated studio drum. Only when rapidly repeated does it become annoyingly obvious that the sound is exactly the same every time you hit it. As mentioned before, retriggering a sound on most drum machines cuts off the previous hit and starts it all over again. If you play fast enough, you get down to a point at which only the very beginning of the sound is heard - now you've really got a machine-like sound.
In the case of real snare drums, the heads continue to ring from each hit even while other hits are piled on, and fill in the spaces between the hits with a natural, random sound that's definitely not identical for every hit - not even for Mr Cobham. In what is known as a "crush" roll, the goal is to play so rapidly and smoothly that the stick strokes literally blend into the sustained ringing of the drum heads against the snares - this is what's responsible for the "rolling" sound. This sound has little in common with the sound our machines make repeating the sample's attack over and over again, making the "crush" technique nearly impossible to program. Do everyone a favour - forget it.
Careful listening to the three-, four-, and five-stroke rolls played by drummers (more slowly, usually as demi-semiquavers) will reveal the volume dynamics which make them effective - an accent on the last stroke, with sometimes a mini-crescendo building up to it, such that the roll in effect becomes a single event that leads your ear along into the rhythm, rather than an obtrusive burst of machine-gun fire that breaks it up. This sort of dynamics takes some practice to program - you will probably have to slow things down a bit, and may need separate takes to record the changing levels of individual drums (although, as we've said, velocity-sensitive pads are starting to find their way onto lower-priced drum boxes).
In spite of the difficulty it takes to create, you should resist the temptation towards heavy re-use of one such event to save programming time - these rhythmic devices are most effective when used sparingly, as occasional embellishments to the rhythm, and with some element of surprise. If exactly the same roll comes in at exactly the same time every fourth bar, we might as well go back to square one...
Much of what's just been said applies to solos and fills, too. Hold the speed down somewhat and play "around the set", rather than homing in on one drum. This lets the sounds ring out longer and minimises the artificial character of abbreviated drum sounds. Work dynamics in as much as possible - not just to showcase the dynamic capabilities of your machine, but with an ear towards the motion-suggesting power that makes percussion dynamics musically useful. A good fill does not stand out from the rest of the rhythm so much as it gives it an extra kick in the arse. And remember - the more off-the-wall it is, the less it needs to be repeated before it starts shouting "I'm a machine" again. Solos should take the listener somewhere (and back?) instead of grinding wheels in one place.
Carelessly used, ride cymbals can appear rather blatant when, after a moment, the cymbal suddenly disappears. Ride cymbals are so named because a drummer can "ride" one with one hand, laying down a steady stream of light quavers and semiquavers to form the top layer of a rhythm. They're large, heavy, and ring for a long time, "hanging" in the air between strikes - so long that a complete hit and decay would easily exhaust the sound capacity of almost any drum machine. In order to capture that "hang", drum-machine ride cymbals use a second or so of unadulterated cymbal (adequate to span the gaps between slow quavers), after which the sound is digitally faded down (or "enveloped") rather abruptly. This compromise solution was obviously intended for the "standard" style of ride cymbal playing, and sounds artificial if the strikes are left too far apart or allowed to die in an open space.
Try to lead the ride rhythms into a crash cymbal hit or other bright instrumental accent, to distract the ear away from the sudden cutoff of the last strike. Some subtle dynamics - even as simple as accenting the even beats or all odd 8th notes - can help transform sequencer cymbal monotony into something with a bit of organic flow.
ASIDE FROM RHYTHM programming techniques, there is the issue of the sounds themselves. As we mentioned earlier, drum-machine sounds tend to be pretty devoid of ambience - either on purpose, or because of memory constraints. Back when drum machines were first introduced, and expensive enough to be found only in professional studios with large mixers and stacks of outboard effects, this wasn't a problem. Now, however, drum machines are widely used in situations where none of these things is present, and we get to listen to the sounds straight out of the box.
"Apply reverb only to certain sounds, rather than to the drum box as a whole. Send just the snare through to accentuate the second and fourth beats, or leave the hi-hats out to keep the upper frequencies open for other instruments."
The good news is two-fold. One, some manufacturers are beginning to recognise this problem and are providing their boxes with sounds that are more useable on their own; and two, just a little bit of help goes a long way here. As with drums-in-a-box, ambience-in-a-box has taken spectacular nosedives in price during the past couple of years, with decent-quality stereo digital reverbs now available for as little as £250. Reverb, in particular, will help restore what is missing from dry drum machine sounds, putting them into a larger acoustic space in which the ambient image is also part of the rhythm, and diverting the ear away from specific shortcomings - such as filling in holes in the rhythm. The sudden cutoff at the end of a ride cymbal sound can be effectively extended or masked by a medium-to-long decay reverb, for instance. Reverb of almost any decay time will provide a space for drums to swell into - in particular, fast licks will sound much more full when reverb is there to pick up where the drum machine leaves off.
Here, as elsewhere, the "less is more" axiom applies. Long reverberation times work best with slow or sparse rhythms, or if kept low in the mix. For busier or more uptempo tracks, gated reverb or small-room settings may prove more effective - they will give the desired sonic enhancement without overwhelming the mix, even when mixed in at a fairly high level.
Another useful technique is that of applying reverb only to certain sounds, rather than to the drum box as a whole - for example, kick the decay time back up and send just the snare through, to accentuate the second and fourth beats, or leave the hi-hats out of the reverb to keep the upper frequency spectrum open for voices and other instruments.
With drum machines that have individual sound outputs (and always assuming your mixer has enough inputs to cope), this is straightforward. Stereo-out drum boxes can also be used to achieve this setup with no external submixing - use the pan control to send the sounds you want reverberated to one output, and the rest of the sounds to the other output. If you have access to a second reverb unit (beg, borrow, or whatever), try using different types of reverb on different sounds at the same time (a common studio practice).
Digital delay lines (DDLs) have, until recently, been less expensive than digital reverbs and therefore more widely used. While they might seem to be an obvious tool for creating ambience, they must be used carefully - in contrast to the diffuse echoes created by a reverb unit, the discrete echoes generated by a delay line are none too subtle, and can wreak rhythmic havoc if not kept under control. In fact, at longer delay times corresponding to quavers and semiquavers, delay lines can be considered rhythmic tools as opposed to ambience (we'll discuss that much later in the series). Keep echo times shorter, and use little or no echo feedback - you will have less trouble with the digital delay taking over. Ambience effects are more often connected with echo times found in rooms or halls - one-tenth of a second or less. Alternatively, you can use slightly more feedback and longer delays with some equalisation to eradicate treble and midrange frequencies (in either the echo send or return) to take the sharp edges off those repeats.
If your delay line has settings in the chorus/doubler range (20-40mSecs), there is a simple and rather effective route to stereo small-room simulation at your disposal. Pan the sounds to be processed fully to one side of the output mix (you may even want to do this to the entire mix). Run this set of sounds through the delay line, and pan its output fully to the other side of the mix with the delay line's Mix control set to echo only. Use a delay time of about 30mSecs, with no feedback and (optionally) a small amount of modulation to the delay time. Finally, balance the level of the echo output to match that of the program mix in the other channel. The echo generated by the delay line falls too closely on the heels of the source sound(s) to be heard as a distinct echo. Instead, your ears do the work of combining it with the original sound to create a pseudo-stereo image in the form of the original sound. This is admittedly a bit cheap, but it works - and is not all that blatant, either.
When nothing else is available, you may be able to use the acoustic properties of one of the rooms in your dwelling (or wherever it is you're recording). Ideally, you would use the largest room, but if this is the living room, it may be too well damped by carpeting.
For best results, the speaker(s) and mics should be as far away as possible from each other, and should not point directly at one another - try the walls, ceiling, or particularly the floor (if it's not carpeted). The mics should be free-standing, as opposed to placed in the corners or right against the walls - but feel free to experiment. Be careful about placing a speaker or mic too far in front of its partner - phase cancellations may result. Naturally, you'll also have to remove, tape, or nail down anything that rattles, and keep people away while recording. This is actually a low-rent version of the "live chamber" echo widely used in recording studios, especially before the advent of plate and digital reverbs.
Next time, we'll confront the possibility that keyboardists have the best tools for making good drum and percussion sounds, and discuss how to exploit them to our own ends. Until then...
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!