The debate's been going on for as long as drum machines have been around: can they really sound human? Nicholas Rowland takes jazz drumming as his test case and tries to find out...
In this month's nail‑biting instalment of The Rhythm Method, we're going to investigate various techniques for making programmed patterns sound more 'human', as well as looking at some short cuts to generating rhythm tracks with an apparently improvised feel. While the focus of our attention will be jazz patterns, the subtext is all about injecting the milk of human kindness into beat boxes in general. So even if you think that jazz is something musicians only do when they get too old to play music that people actually want to listen to, stay tuned.
On the face of it, jazz drum programming appears to be a contradiction in terms. Jazz music is supposed to be all about the spontaneous expression of heart and soul, while drum machines and sequencers are soulless machines, the very opposites of spontaneity, creativity and having a good laugh down the boozer after the gig. That was certainly true 10 years ago, when drum machines simply didn't have the technical facilities to compete with humans on a jazz tip. First, the sounds themselves were often not realistic enough to be appropriate for jazz (though, to be fair, this was more an attitude of mind than a valid technical issue). Second, and more importantly, early drum machines just didn't offer the necessary control over dynamics and quantisation which are necessary if you want to emulate the subtle nuances of a live drummer in full flow.
...thanks to cut and paste, you can quickly generate drum tracks which have an apparently improvised feel.
These days, there are no excuses. Armed with the most basic GM module/workstation and the humblest of computer sequencers, you can produce jazz patterns that not only sound convincing, but swing with the best of them. The only real limit to your creativity is your time. Sequencers and drum machines only put out what you put in. If you want to create a rhythm track based around the idea that each bar is different from the next, then you'll have to be prepared to program every single variation yourself. From my own experience, I know it can take many hours to recreate the kind of spontaneous‑sounding jazz track that any drummer worth their salt could lay down in a single take. Be prepared.
One question which is perhaps worth spending a few lines considering is what exactly differentiates a rhythm played by a human from one created by a machine. Setting aside the issue of sounds and ambience for the moment, can most people actually tell the difference between a recording featuring a real drummer and one driven by a beat box? It was probably easier to distinguish in the early days, when a combination of lazy programming and a lack of onboard memory meant that drum machines gave themselves away by undue repetition. The lack of control over dynamics also meant that drum machines really did sound like metronomes — not so much because of the regularity of timing, but because of the total consistency of the sounds. What makes music 'human', on the other hand, is the minor inconsistencies in the playing, in terms of timing, dynamics and the variations inherent in acoustic instruments. There's also this ephemeral notion of 'interpretation' — which can, perhaps, be defined as an ability to creatively bend the rules to enhance the emotional pleasure of the music. Or to put it another way, if it ain't got that swing, it don't mean a thing.
As I mentioned last month, dynamics (the relative MIDI velocity levels of the different instruments) are crucial to creating a sense of movement within any style of drum pattern. Creating convincing jazz patterns requires even more attention to detail in this matter. Obviously, the easiest way to achieve a human feel is simply to program your rhythms in real time, using a velocity‑sensitive MIDI keyboard, drum pads or drum machine buttons. I'd recommend this as your standard approach with cymbal parts, which often provide the fluidity of movement within a rhythm. (In jazz, it's the ride cymbal which is the dominant time‑keeping instrument, as opposed to the hi‑hats). Most sequencers offer a mixture of pattern‑based and linear recording, so it's easy enough to build up a basic track from a series of step‑time created patterns, then go back and record a new 'live' cymbal line over the entire track. Try also setting the quantise function to a very fine resolution, or turning it off altogether. You can normally go back and correct any really wayward beats after the event, using the over‑quantise function.
What originally really used to get up people's noses about drum machines was the fact that they kept 'inhumanly' strict tempo — a charge which is still levelled at sequenced music per se. There are two issues here. One is about variations in tempo across the whole track — in other words, the fact that people naturally speed up and slow down during different bits of a song. There's no reason why sequenced music shouldn't also speed up and slow down, and thanks to the wonder of sequencer tempo maps it's very easy to build this kind of variation into a song. In fact, whatever the style of music, one trick is to nudge the tempo up by a couple of beats when you hit the chorus or playout, and take it down a few notches in the bridge from the introduction to the first verse, or the bridge from the middle eight to the next verse, and so on.
The second issue concerns the minuscule variations in timing that occur within a pattern. Here we're touching on a human foible known in drumming circles as playing behind or in front of the beat. The fact is that the majority of human drummers (and, for that matter, most other musicians) rarely hit the notes right on the button. Some will have a natural inclination to play slightly early, others will play slightly late; some can go back and forth as the music demands. Playing behind the beat will drag the song back and make the track sound slightly slower than it actually is. You notice this in a lot of slow blues and funk numbers, where often the whole band hits everything slightly late. Playing ahead of the beat gives the song real urgency, making it sound faster even though the tempo hasn't actually changed. Again, this is easily replicated on most sequencers (and some drum machines), which allow you to shift patterns or entire drum tracks by a specified number of MIDI ticks. It's worth experimenting with this function, particularly on the snare when you've got a regular beat on the two and the four. But don't overdo it, or your drummer will just sound out of time.
Some sequencers and drum machines take this a stage further, with intelligent quantise functions which alter the MIDI velocity of certain beats, while also shifting the timing of certain beats by tiny amounts. But whereas early applications of this function imposed the changes randomly, it's now based on more careful analysis of the rhythmic pulse of particular styles of music. Personally, I think these functions work best when they're applied sparingly — for example, to a fill or particular drum phrase rather than across the entire track. (See the examples box for further discussion of this.) Otherwise the drumming just sounds wrong rather than 'human'.
As most people are aware, an acoustic drum doesn't just get louder when it's struck harder, it also changes timbre, rising in pitch and exhibiting more pitch‑bend. Cymbals will also change timbre according to where they are struck on their surface, and also how rapidly they are played. Some drum machines and sound modules simulate this through multi‑sampled voices which will change according to MIDI velocity. If a sampler is the source of your drum voices, you can also easily set up velocity‑sensitive cross‑fades between different pitches of the same sound or, indeed, different sounds.
A similar effect can be achieved with more humble equipment. For example, the standard GM kit offers a choice of two ride cymbals, plus a more 'clangy'‑sounding ride 'bell'. As a matter of course, I would use at least two of these sounds within a jazz ride pattern, if not all three. It really does make a difference. Similarly, when programming two bass drum notes in quick succession, try using a softer, more rounded one for the first beat and a sharper, heavier sound for the second.
So far, so good. But while technology may be on our side in terms of making individual patterns sound more human, I appreciate that not everyone has the time and patience to laboriously trawl through a drum track beat by beat, instrument by instrument, tinkering about with individual velocities, timing values and so on. However, thanks to the power of cut and paste, you can quickly generate drum tracks which have an apparently improvised feel.
Dynamics are crucial to creating a sense of movement within any drum pattern.
The process starts with the creation of a 1‑ or 2‑bar 'master' pattern. With a jazz track it might be the archetypal jazz cymbal rhythm, underpinned by a basic bass and snare figure. This is then copied to several pattern locations — an easy enough job whether you're using a stand‑alone drum machine or a computer‑based sequencer. You then call up one of these copies and start deleting, adding or moving a couple of cymbal beats here, a couple of snares or basses there. Maybe just delete every fifth cymbal note — whatever. The trick is not to think too hard about what you're doing, and for this reason I often work in step time, because then it's hard to second‑guess the end result. What you should end up with is a family of 1‑bar patterns, all based around the master rhythm, yet each one slightly different. When chaining these together to form the song, simply assemble them in a random order. Hence the first verse might consist of patterns 1/2/3/4, but the second would be 2/4/2/3, and so on. Again, don't try to second‑guess the result. When played end to end, the finished rhythm track might sound a bit iffy, but once you've got the rest of the instruments in place the result should sound more coherent.
With a sequencer, applying this technique is even easier. For example, in the edit page of a program like Cubase you can easily sub‑divide your master and variation patterns into smaller sections — half‑bars, or even quarter‑bars, for instance, and then use these smaller building blocks to build up the complete drum track.
Once the other parts are in place, it's worth going back to the drum edit page and tweaking the patterns to better fit the structure of the track. For example, there might be places where the insertion of a crash cymbal would provide an accent or mark the division of a bar.
Et voila! What you now have is a rhythm track with a large element of unpredictability about it — almost as good as a having a machine with its own mind!
Your starter for 10 is a cluster of four jazz patterns based around the archetypal jazz cymbal pattern. In rhythmic terms there's nothing to blow your socks off. What I'm more interested in here is illustrating some of the techniques explained in the main article. Before we get into that, though, some basic house rules: whereas rock and pop rhythms are based on even‑length notes, jazz rhythms are divided into uneven sections based on combinations of quavers, dotted quavers and semi‑quavers.The archetypal jazz hi‑hat/cymbal rhythm (you'll know it when you hear it, honest) is based on a quaver/dotted quaver/semi‑quaver combination. However, this is played with a triplet feel, so in programming terms it's easiest to program in triplets. So the patterns this month were all created using a quantise value of 8th‑note triplets (that is, 12 steps to the bar.) You'll see that a few examples actually use 16th‑note triplets, but for the sake of making the grids easy to understand, I've kept the grids quantised at 12 steps to the bar.
In Part 1 we had three dynamic levels, but this month I've used four. And instead of giving them absolute MIDI velocity values, I've suggested a range of dynamic levels to play with. Ideally, you would tweak each beat/instrument, so that no consecutive values are the same. Finally, while this month's examples are all based on jazz patterns, the general points covered in the main body of the text are equally applicable to any style of music where a human feel is required.
JAZZ PATTERN A
This is the archetypal (some might say cliché) jazz rhythm, played using the ride cymbal. Note, though, that even with a simple pattern such as this, we're using the ride and ride bell sounds to add a little extra interest to the rhythm.
JAZZ PATTERN B
With more blobs on the grid you can see that this is a more intricate version of pattern A. Again we're mixing the ride bell and ride cymbals to give more interest at the top end. I've also doubled up the tambourine with the pedal hi‑hat to add emphasis to the second and fourth beats in each bar.
JAZZ PATTERN C
The snare fill towards the end of the second bar uses a mix of two snare sounds to simulate the changing timbre of an acoustic drum. The 'double diamond' (brush snare, fourth beat, second bar) is a flam, when two notes are played close together.
JAZZ PATTERN D
This example mixes sounds again — bringing in the metallic sounds of a timbale over the top of the brush snare during a short roll. The psychedelic wedges are supposed to indicate the fact that one gets louder and the other gets quieter — but I guess you'd probably worked that out for yourself. Timbale sounds are often quite strident, so you might have to play around with the relative volumes of the snare and timbale to get the most pleasing effect.
This is a simple jazz fill pattern which has been part‑doctored using a Cubase groove quantise, although only the bit contained in the white box. The results are shown more clearly in the enlarged section. The result is a lazy, lumpy snare fill which sounds very human indeed! I would probably apply this quantise to any other instruments which were playing at this time, so that the whole 'band' sounds locked in together, even though they're all 'out of time'. Rather than applying these quantise treatments wholesale across a song, I tend to use different treatments in different parts of the song. There are times when you want things to sound tight and absolutely locked into the groove, other times when you want to 'push' a chorus, or 'pull back' a verse. This may sound corny, but think of the different character of each part of a song, then look at how you might enhance them through playing about with the micro‑timing of beats.
You can also try a reverse approach. A lot of people will quantise the rhythm track to death while being happy to input the rest of the instruments in real time. What you can consider is copying instrumental tracks to the drum tracks, then using the note‑ons of the melodic parts as the basis for quantising the drums.
If you watch a jazz drummer playing, you'll notice a lot of tapping away at the snare drum in between the main beats or accents. These are often known as ghost notes and, while you often can't hear them when a band is in full flight, they add a characteristic background wash to a recording or live performance. The purists out there might consider even taking all the attack off the ghosted sound and then feeding it through a short reverb back into the main drum mix. This also goes some way towards simulating the way that the snare will vibrate whenever other drums in an acoustic kit are struck.
JAZZ IMPRO 1, 2 & 3
You've probably already got the idea from the main body of the text, but here's a couple of patterns which have been chopped up and reassembled.
The toms in Jazz Impro 1, especially in Section 1, give the pattern a strong theme, so when putting the track together I've used this as the first bar in each four‑bar section.