Aphex Feel Factory
American company Aphex bring us a new concept in the "humanisation" of machine-generated music, in the form of their Feel Factory. Robert Rich finds himself in the world of grooves and Feel Algorithms.
The latest revolution to come from Aphex is a box that allows you to adjust the timing and dynamics of MIDI sequences to alter the music's feel.
WHAT'S THE DIFFERENCE between good musicians and great ones? Michael Stewart thinks that the answer to this question lies in their "feel". Michael Stewart has spent many years trying to capture the essence of feel, and he's invented several gadgets to help us capture that elusive groove. All his MIDI processors share a common goal: to breathe life into sequenced music. Stewart first brought us the Kahler Human Clock which synchronised a sequencer to a live drummer. His latest creation is Feel Factory, a "groove laboratory" that offers control over the "feel" of sequenced MIDI notes.
SO JUST WHAT is feel? Well, here's the basic theory behind Feel Factory: feel is a combination of timing and dynamics. Imagine a drum pattern in which the kick drum keeps a steady pulse and the snare defines the feel. Snare hits that occur slightly ahead of the beat tend to "push" the rhythm, imparting "drive", "snap", or "nervousness" at the extreme. When the snare hits fall slightly behind the beat, the rhythm tends to sound heavier, "in the pocket", more like an R 'n' B groove.
One factor that probably contributes to feel is called the "precedence effect". When two sounds occur very close together in time (less than about 70 milliseconds apart) we tend to suppress the later sound, hearing the first sound as being louder. The timing of events affects the perception of loudness. It seems reasonable that we could alter the perceived dynamics of a rhythm by changing its timing, and perhaps alter the perceived timing by changing the dynamics. Feel Factory specialises in just this sort of manipulation.
Most musicians probably don't intellectualise too much about feel - it's just one of those mysterious parameters that contribute to a musical performance. Some musicians may even feel threatened at the prospect of analysing such musical subtleties, but whether we like it or not, electronic music gives us control over all sorts of parameters that we've never had to think too hard about before. This degree of control formed one of the founding ideals in early electronic music experiments. (You could argue that MIDI has caused us to take a backward step from these ideals, inviting laziness through its limited assumptions about the nature of musical information.) Alas, it's very easy to create lifeless electronic music. By thinking a bit about this question of feel, we can hope to regain some of the control hidden within our technology.
I must admit that, at first, I wasn't convinced that feel was such a predictable and quantifiable quality. So when I first came up against Feel Factory I tried a little experiment. I entered the values provided in the manual for a 'reggae groove". Instead of feeding it a reggae rhythm however, I tried a quantised rock pattern. To my amazement, the groove of my pattem shifted. It sounded more relaxed, with a definite, though subtle, reggae touch. When I switched Feel Factory to an "R 'n' B groove", the reggae feel disappeared. Hardly a controlled experiment, but it was enough :c change my attitude about the validity of Feel Factory's approach.
SO WE TURN from theory to practice. Feel Factory is a flat metal box designed to sit on a table top, with eight short sliders, six programming buttons and a backlit 32-character LCD. The rear panel of the box holds all the communications connections: One MIDI In, four MIDI Outs, SMPTE in and out, MIDI clock out, Macintosh serial interface, power switch and 9V power input. There are also two extra ports labelled Mac data and MIDI data that are "reserved for future Feel Factory products".
This list should give some hints about the capabilities of Feel Factory. At its most basic level, it can accept incoming MIDI data, process it and spew it out of its four MIDI Out ports. The Mac serial port lets the unit double as a MIDI interface for the Apple Macintosh computer. Unfortunately, this wasn't working on the review sample, although Aphex assure me that it was working when it left the factory.) Feel Factory can control an external sequencer using MIDI Sync and Start/Stop messages, with an internal tempo of 60-200bpm, and it can also stripe and read SMPTE at 24, 25, 29.97 (drop frame) and 30 frames per second.
There were still some bugs in the user interface of the review sample, but ROM upgrades are promised after the initial release of the unit. However, I doubt that Feel Factory will ever satisfy those in need of instant gratification. The learning curve is fairly steep, and it took a good few hours before I could get it to do anything useful.
USING FEEL FACTORY involves two steps. First, you have to set up "feel algorithms" and apply them to the appropriate note ranges and MIDI channels. That's the hard part; once you've defined a few presets, recalling and using them is pretty intuitive.
The algorithms themselves provide timing and/or velocity shifts from note to note within a track. Each algorithm provides up to eight values that are used to perform these shifts. These values range from -9 to +9, with each step representing a number that is three times the value. For example, a value of 2 will boost a note's velocity by six, or shift its timing forward by six resolution steps (with an internal resolution of 0.832 milliseconds, this translates into about 5msec).
Different types of algorithm apply these values to the timing or velocity of incoming notes in different ways. The three types of algorithm are called Placement, Frequency and Randomisation. The Placement algorythm acts on different points within a bar of music. Each of the eight values affects a different quaver (8th note) within each bar. Value No. 1 affects the first quaver, value No. 2 affects the second quaver, and so on. The Frequency algorithm measures the time between incoming notes. In this algorithm, the first value affects notes separated by a quaver length, the second number affects notes separated by a crotchet (quarter note) length, and so on up to the eighth value that affects notes separated by a semibreve (whole note) length. The Random algorithm simply cycles through its values, applying the next number in its list to the timing or velocity of the next note it sees.
By applying all eight values in a Placement algorithm, you are by definition processing a 4/4 rhythm. You can reduce the list to correspond to smaller time signatures, or to achieve compound rhythms. For example, applying six Placement algorithm values to a 4/4 rhythm could create a three-against-four accent. I wish that these algorithms could cycle through more than eight values, however. The eight value limitation makes it impossible to add non-random feels to rhythm with more than four beats to the bar. This restriction is unlikely to be a real problem to anyone, however, because the Random algorithm can add life to just about anything.
The eight sliders on the front panel of Feel Factory can be assigned independently to timing or velocity, letting you slide notes forward or backward in time or change their dynamics in real time. You can assign a slider to a particular MIDI channel or a range of notes. These ranges can be applied to continuous notes or separate notes scattered across the keyboard.
A slider can control a range of 127 values, from -63 to +63, but it can only access a range of 64 values at any time. Feel Factory determines these ranges on power-up by defining each fader's current position as zero (this feature is called "auto null"). If a slider is at the bottom of its range when you turn the power on, it will span 0 to +64. If it's in the middle, it will span -31 to +31. You can adjust these ranges in a utility screen or by turning the power off, moving the slider and turning the unit on. Although a little confusing at first, the auto-null feature makes sense once you get used to it. I wish the faders were longer, though, and covered the full range.
The sliders control global changes, like shifting the timing of an entire MIDI channel. For more microscopic changes, you assign one of the three types of algorithm to each of the eight Feel Algorithm locations. You can assign two of these algorithms to each of the sliders (their effects can be additive - a very useful touch), although you can't really control the algorithms from the slider. What you're really doing is assigning two algorithms and one slider to the same MIDI channel or note range. This does restrict your assignment possibilities, but it also makes the assignments a lot easier to understand.
Feel Factory has three kinds of memory. Online memory chews on MIDI data, with eight slider assignments, eight range definitions, and eight algorithms. Recorder memory saves every change that you make to the Online memory. Thirty-two Preset memories let you store your setups. Rather than saving the complete set of machine parameters, a Preset saves only the changes that are stored in the Recorder memory - all the changes you've made to Feel Factory's power-up state. When you recall a Preset, its contents are combined with the Online memory, allowing you to combine the effects of different algorithms. The parameters not replaced by the recalled preset remain unchanged. (By the way, you can recall Presets with MIDI program change commands on channel 16, and you can adjust all Online parameters with SysEx.)
This memory scheme allows you to process all 16 MIDI channels despite having only eight sliders. If you set timing values for channels 1-8 then recall a preset that affects channels 9-16, the values for channels I-8 will still operate, although your real time control of the faders will only affect channels 9-16. The intricacies of this memory overlay structure can make the mind swim, but it does have a certain elegance.
YOU MAY HAVE wondered how this box can make a note play before it was supposed to. The answer is deceptively simple - it delays everything. Its nominal processing delay is around 32msec, which is roughly equivalent to a single SMPTE frame. With a time-shift range of +/- 32msec from its 32msec "centrepoint", feel Factory actually creates delays from nearly instantaneous (with a 1msec processing lag) to 64msec. In effect, Feel Factory is able to push notes ahead of the beat by pushing them ahead of their delayed neighbours.
This delay raises the following question: if you have a tape already recorded with music and SMPTE, how do you "feel factor' a sequence that syncs to this tape via SMPTE without introducing a noticeable time lag? The trick is to nudge your SMPTE start points. Just follow these steps. First, record the sequence without processing your notes with Feel Factory. You can still use the unit as a SMPTE reader to sync the sequencer to tape while recording, but if you monitor your playing through its MIDI connections, the delays will wreak havoc with your own natural feel. Then for playback, if you're using a SMPTE box other than Feel Factory, set the sequencer to start one frame earlier than when you recorded the track. If you're using Feel Factory as a SMPTE reader, it will automatically start one frame earlier than its setting states. As long as you remember that Feel Factory always reads SMPTE one frame early, but processes MIDI One frame late, you can keep track of your delays and avoid confusion. Note that you won't run into delay problems if all your tracks are sequenced and you run them all through Feel Factory processed or not. Its delays are at least consistent and predictable.
Because so many of Feel Factory's applications involve SMPTE, its internal SMPTE read/write functions will prove quite handy for those without a SMPTE interface. But before you get too excited about its SMPTE capabilities, I should point out that it doesn't actually keep tempo maps or perform some of the other handy housekeeping chores that we have come to expect from a fully-featured SMPTE box. Feel Factory takes a bit more time than usual to catch up to a stripe, and it's a bit finicky about poorly recorded signals. To avoid the slow catch-up time, the manual recommends striping for each song individually and giving at least five seconds lead time before the song starts. I didn't actually encounter any problems while using SMPTE until I tried pushing the limits. Michael Stewart explained that Feel Factory was not intended to be a fully-featured SMPTE interface and to include features like tempo maps would have required extra memory and therefore a higher price tag.
On the subject of timing, I should mention again that Feel Factory has a maximum internal resolution of 832 microseconds (equivalent to two SMPTE bits), about as fine as anyone should need. The sliders give full access to this resolution, while the algorithms work at a coarser 3:1 resolution.
FEEL FACTORY HAS an interesting future. Aphex, and Michael Stewart in particular, want to see it become the springboard for a new area of musical development. The algorithms themselves may grow more sophisticated with future ROM updates, possible additions including programmable mod-wheel variations and algorithms that modulate each other. As for peripheral developments, the first one I'd like to see would be a screen editor to provide relief from endless button-pressing.
Stewart also says he hopes that people will use Feel Factory not only as a production tool, but as a tool for researching the whole question of "feel". He wants to see people sharing their discoveries with others, and trading Feel Factory presets. If nothing else, the ideas behind Feel Factory may inspire musicians to listen more closely to some very subtle musical qualities.
FEEL FACTORY IS not your simple "plug in and jam" sort of gadget. It takes some time to learn. It invites an analytical approach to timing and dynamics. Some might complain that feel should stay in the realm of intuition, and that Feel Factory requires too much pre-planning before you can use it correctly.
However, once you've climbed the steep learning curve required to set up your own feel algorithms, the actual "performance of a feel" is fairly intuitive. Once you have a good setup, you can push sliders back and forth while listening to the subtle (or drastic) timing effects in the music. No other piece equipment that I know of offers such immediate feedback with this degree of subtle control. Of course. many sequencers will let you do what Feel Factory does, but they won't usually let you hear the changes as you make them. To replicate some of the effects of Fed Factory's algorithms could take hours of individual note editing. Finally, the unit's timing resolution far surpasses that of almost any sequencer.
I would not recommend this device for every starving MIDI musician. Get the essentials first - a good synth, mixer, reverb. multitrack or whatever. If you're a good musician, you can get a good feel without Feel Factory. But if you're looking for that extra something, check it out. The is a learning tool, a research tool, a post-production tool, and now that I've heard the difference, I just might become addicted.
Price £649 including VAT
Review by Robert Rich
Previous article in this issue:
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!
Please note: Our yearly hosting fees are due every March, so monetary donations are especially appreciated to help meet this cost. Thank you for your support!