As a desktop musician, you've undoubtedly amassed a collection of audio files packed with loops of all types. The trouble is, unless you created them yourself, everyone else has the same loops. If you've used a particular loop in a piece of music, then heard it pop up somewhere else, you know the problem. Of course, not overusing canned audio files of any kind is the real solution, but you can also do quite a bit to spice up the loops and other audio files in your collection. In this article, I'll take you step-by-step through a number of tried-and-true techniques that go beyond just applying your favorite DSP effect.
Let's start with a brief look at the history of tape music and the various techniques employed by its pioneers. Why? Because applying tape-music techniques to the digital world can yield a number of interesting and not overly explored variations for your loop library. In fact, most of what I'll cover can be done with a sequencer and sample editor (the modern equivalents of tape) and without any DSP plug-ins.
As the word loop suggests, the use of loops in electronic music dates back to the earliest days of tape music. The first electronic music studio dedicated primarily to the use of tape techniques was opened in Paris in 1948 by Pierre Schaeffer. He and Pierre Henri, who joined a year later, pioneered many of the tape techniques that are the forerunners of sample-editing and digital-audio-sequencing techniques used today. Primarily, they recorded everyday sounds on tape and then cut and spliced them to form collages — a style they called musique concrète.
Guitarist Les Paul, another early experimenter with tape techniques, is credited with inventing multitrack recording, tape echo, and sound-on-sound in the 1950s. Minimalist composers made extensive use of tape techniques beginning in the 1960s. In particular, Steve Reich's two speech-loop tape pieces It's Gonna Rain and Come Out introduced slowly shifting phase relationships between two tape loops of different lengths. John Lennon's Revolution 9 (1968) is almost entirely a collage of clips and loops of sounds from the EMI sound archive. From the 1950s to the late 1980s, tape techniques played a major roll in virtually all styles of music.
In the world of tape there are only a few basic strategies. You can record sound and play it back. You can cut the tape up and splice the pieces together in a different order. You can play the tape backwards by reversing the reels and flipping them over. You can vary the speed (in the early days, speed could vary between two settings; now speed can vary continuously). You can dub the playback of one or more tape players onto another tape recorder. With multitrack recorders you can also dub some tracks to others on the same machine. And finally, you can play loops on one or more tape players. That opens the door to flanging as well as the shifting phase techniques just mentioned. All of those tape techniques can be replicated in digital-audio-sequencing and sample-editing software.
A CLEVER TURN OF PHASE
You most likely have an effects box or plug-in that offers flanging — a process that involves splitting a signal into two copies, slightly delaying one of the copies, and varying the delay time. Flanging was originally a real-time tape process (developed at the Abbey Road studio) in which an audio signal is recorded and played back simultaneously on two tape recorders. The term flanging refers to applying pressure to the flange of the feed reel on one of the recorders, thereby slowing it down and causing a slight delay in one of the copies. Varying the pressure varies the tape speed, producing the familiar whooshing sound associated with flanging.
Mixing two copies of the same audio file while slightly delaying one of them causes some frequencies to be attenuated while others are enhanced. That is the result of shifting the phase of the individual sine wave components of one signal with respect to the other. For example, if the delay causes a half wavelength shift at a particular frequency, that frequency will be cancelled entirely, whereas if it causes a full wavelength shift, that frequency will be doubled in level. Other amounts of shift cause varying degrees of attenuation or enhancement. If you don't vary the delay time, you'll get a coloration of the signal without the more noticeable motion associated with flanging.
Slightly delaying a loop and mixing it with a copy is easy in a multitrack audio program. Just put the loops on different audio tracks and nudge one of the loops slightly forward in time. It's a process that works well with atmospheres and ambient loops. Shifting the copy by a slightly different amount for each repetition of the loop adds slight timbral variations that make the loop more interesting without being obvious. Moving adjacent repetitions of the loop by different amounts produces a small gap or overlap, depending on which copy is moved more. That may be masked by the track playing the other copy of the loop, but if there's an audible problem, it's easily repaired by slightly adjusting the loop lengths or applying a short crossfade.
Fig. 1 shows an example created in Emagic Logic. The two tracks holding the individual loops are shown at the top; Logic's List editor (below) shows the amount of shift for each loop on the second track. The shifts are shown in ticks (often called pulses), which in Logic are 1/960 of a quarter note. At 120 bpm, which is the tempo of the example, a tick is roughly 0.5 ms. The shifts in Fig. 1 range from 2 to 20 ticks or 1 to 10 ms, which is pretty much the useful range for this process. To calculate the shift per tick in your sequencer, divide the tempo in bpm into 60,000, then divide the number of ticks per quarter note into that. You can hear the example in the MP3 file Oyster.mp3.
As you'll discover, some shifts sound better than others, and the only way to know is to try them out. For that, it's convenient to set the working tempo so that the loop fills an exact number of measures, because that makes it easy to cycle a single repetition while nudging one of the copies. If you plan to repeat the loop a large number of times in your song, you don't really need to set up an individual offset for each repetition. You can save time by doing a small number of repetitions, then rendering the result to create a new, longer loop. For the best results, use a repetition number that doesn't divide evenly into the size of your song sections. For example, if your song is divided into 16-bar sections, render an odd number of repetitions so the new, longer loop doesn't cycle exactly with the song sections.
The classic Reich-style phase
In 1965 and 1966, Steve Reich composed two minimalist pieces consisting entirely of repetitions of a speech loop. Copies of the same loop were played on two tape recorders, but one copy was spliced slightly shorter than the other. As a result, the loops started in synchronization and slowly drifted out of phase. The key word here is slowly. The minimalist nature of the process requires that the evolution be barely perceptible, with the result that at various times you suddenly become aware that you are hearing an entirely new cadence or rhythm.
Creating that kind of phase process in a digital audio sequencer is again extremely simple. Place single copies of the clip to be looped on two tracks, and use a high zoom level to slightly shorten one of the clips by a few milliseconds. (You don't necessarily have to restrict yourself to the same clip for both tracks.) Then loop both clips for enough repetitions that the clips eventually come back into alignment.
You can calculate the number of repetitions required to come full circle by dividing the time difference into the total time for the unaltered clip. For example, if the original clip is three seconds long and you shorten the copy by 10 ms, you will need 300 repetitions (3,000/10). At three seconds per repetition, that will take 900 seconds or 15 minutes.
You might not have use for a 15-minute, minimalist, shifting-phase loop in any of your songs, but there's more here than meets the ear. And the process is not just applicable to spoken word loops — it produces very interesting results with background parts such as strings, pads, and background vocals. The idea is not to use the whole thing, but to cull it for interesting new loops. The audio example StringPhase.mp3 is the first minute of a 12-minute piece made using a string loop from the PowerFX Downtown Orchestra collection. The example StringClips.mp3 is made up of nine different loops cut from the full 12-minute piece. (Each loop repeats four times.)
It is easiest to find and extract loops if you start by setting the song tempo so that the original loop occupies a single bar. Once you've looped the clips, render the two tracks to a new audio file, mute the original tracks, and place the rendered file on a new track. Either listen to the whole thing, marking measures of particular interest, or simply audition single measures at random until you find something you like. When you find a repetition you like, listen to several repetitions on either side of it — there will be very slight differences, and you may find one you prefer. When you've zeroed in on a specific repetition, slice the audio file at the measure boundaries, and you have a new loop.
The big-gulp approach
Slow evolution is not always the best way to go, and in fact, for percussion loops, such as drums or rhythm guitar, it works best to shorten one of the loops by a quantized value such as a 16th note. With that approach, you get a different rhythm pattern for each repetition of the loop. You might want to use the whole cycle or extract interesting individual loops as in the previous example. If you use a pitched part such as rhythm guitar, be sure to avoid chord changes, because each repetition shifts the chord change earlier in the shortened loop.
Fig. 2 shows two four-bar rhythm-guitar parts recorded using Steinberg's Virtual Guitarist and Cubase SX. In the example, the lower part is shortened by a 16th note. Therefore, the phase shifts by a 16th note every four bars, and a complete cycle takes 256 bars (8.5 minutes at 120 bpm). As in the prior example, you can cull interesting four-bar loops from a rendering of the complete cycle. But in this case, you can also slice out a long loop from anywhere in the rendered file. Because the evolution is by 16th-note jumps rather than by a few milliseconds, longer selections don't slip out of phase. You can hear a 32-bar version in the audio example BoogiePhase.mp3.
A loop typically consists of a single audio file, but not necessarily. Software sequencers make it easy to group parts (audio or MIDI) from several tracks and then loop the group. That gives you much more latitude to introduce variations using DSP effects and automation on the individual tracks of the loop over longer periods of time than you have with a single repetition.
Fig. 3 and Fig. 4 show a four-bar composite loop created in Propellerhead's Reason. Each of the pads of Reason's ten-pad sample player Redrum is used to play a loop or effects clip. All but one of the pads is routed to its own mixer channel, with the remaining pad, a rap-vocal loop, routed to the modulation input of a vocoder. The vocoder's carrier is supplied by a send bus from the mixer. Automation is used to create the carrier mix as well as to control the vocoder level in the mix. Delay, chorus, and reverb send effects are also used, with the delay return having its own mixer channel.
Fig. 3 shows the four-bar, 10-track loop in Reason's piano-roll style editor. The individual bars indicate gating of the Redrum pads, which were set up to play as long as the gate was held. The four-bar composite loop was made into a group in the track view of the sequencer (not shown), and the group was then looped eight times to create the piece BellTree.mp3.
Fig. 4 shows the Reason rack with the sequencer's mixer-automation view at the bottom. Automation was used for channel levels, for muting the kick drum and hi-hat channels, and for the return of the phaser effect. Although the loop repeats every 4 bars, the automation spans the entire 32 bars.
Against the grain
Granular techniques are not usually described in terms of looping and slicing, but viewing them that way can help demystify the topic. Granular processing is essentially just sequencing grains (slices) of an audio file. One thing that sets granular processing apart from the techniques I've described so far, as well as from run-of-the-mill beat-slicing, is that the grains are extracted in real time. In other words, you don't first do the slicing and then later do the rearranging.
Another difference is that the grains are not necessarily played at their original speed. In essence, you have control of individual grain pitch and playback speed. The resequencing of the grains is often controlled by some repeating process such as an LFO or free-form contour generator. It's the repetitive nature of the grain sequencing that relates granular techniques to loop processing.
Granular synthesizers typically offer real-time control of four main parameters: grain location (within the audio file), grain size, grain density (the rate at which grains are triggered), and grain pitch (the speed at which the individual grain is played). There are often additional controls for smoothing (applying an attack-decay envelope to each grain), randomization of the four main parameters, and reverse playback (applied to individual grains).
Individual grain playback is typically not looped, but you can achieve the same thing by setting the grain density to match the grain size and leaving the grain location fixed. Ordinarily, grain sizes are kept very small (a few milliseconds or even a few samples), but that isn't a requirement. If you set the grain size to the entire audio file, for example, you can loop the audio file as a single grain. That's somewhat contrary to the spirit of granular processing, but using large grains — say, an eighth- or 16th-note long — can be very interesting.
When using tiny grains, the analogy to looping is affected by grain-location automation. The most obvious technique is to linearly vary the location forward in time — using a ramp-up LFO, for example. That allows the grain pitch to control the playback pitch while the LFO rate controls the playback speed, giving independent control of pitch and time. Beyond that, using a different LFO shape or a retriggering envelope generator allows you to construct all kinds of loops, with completely independent control of grain pitch and loop length. Using a step sequencer to control grain location is another possibility.
Granular synths are not that common and are most often part of more technical applications, such as Native Instruments' Reaktor or Cycling '74's Max/MSP. If you have a granular synth, however, you have a potent tool for loop mangling. Granular DSP plug-ins are more common, but because they work in real time, they don't offer control of the most important loop-processing feature: grain location.
Fig. 5 shows Reaktor's Travelizer Ensemble, which comes in the Reaktor factory library. There are several other grain-player-based Ensembles in the Reaktor library, and you can, of course, build your own. The large x-y-controller on the right sets the grain location and size. LFOs are provided for modulating grain pitch and location. Resonator and feedback-delay effects are added at the end of the signal path. The audio example DalyTravels.mp3 is made entirely from loops created in Travelizer using a single speech clip (heard at the beginning).
Almost any software that lets you graphically manipulate audio files can be used to create new loops from old. Sample-editing software typically allows you to cut, copy, paste, merge, insert, reverse, fade, and otherwise munge any selection within an audio file. Many let you destructively apply effects plug-ins to selections (rather than the whole file). Independent time-stretching, pitch-shifting, and formant-shifting are also often provided. Use those tools on different parts of a loop or different copies of the same loop pasted end to end.
Even with a basic digital audio sequencer, you can get a lot of extra mileage out of your loop library with the techniques described here. If you have more advanced tools, the possibilities are limitless. The trick is to keep the basics in mind, start simple, and let things evolve from there.
Len Sasso can be contacted through his Web site at www.swiftkick.com.