by Ben Denckla
The auditory effects associated with the use of delay are already well documented, so this paper will focus on the compositional implications of delay and some ideas about its future use, especially as related to parametric control of delay devices in performance.
The modern delay scene is composed of digital audio delay devices and a few MIDI delay devices. Old analog delay devices (most of which I assume were built around the bucket-brigade design) are made all the more valuable today because of their scarcity. The continued value of analog delay devices is not due to any desirable operational characteristics that they possess in terms of the customary criteria of fidelity, maximum delay time, and flexibility of application (multiple tap points, MIDI controllability, etc.). It is their unique quirks and very lack of fidelity that lend them unique sound-processing capabilities not easily attainable by other means. I base the above statements on my experience with the Wilson Analog Delay produced by Serge Modular, but I feel that my statements might well apply to analog delay devices as a whole.
Tape delay was critical part of experimental and popular electronic music until the advent of inexpensive digital delays with large memories. Still, there may be reasons to use tape delay even now. For instace, the modern use of tape delay has meaning as a historical reference. In addition, the loss of fidelity in the regeneration process of tape delay may be a desirable effect, though I imagine that the use of a low-pass filter in the feedback path of a digital delay would achieve a similar effect of gradual softening the sound into nothingness. Such filtering capabilities are almost universally available on digital delays. A tape delay effect not easy to duplicate is the progressive distortion induced by using a gain greater than unity in the feedback path. The sound of tape saturation that results is quite unique. Of course gains of greater than unity are typically used only for short periods of time since the system typically reaches some kind of steady state roar after a few delay periods.
One might want to reinvestigate the practice of two-machine long tape delays. These are best achieved using low tape speeds to decrease the displacement needed between the two tape machines. Using this technique, a separation of 4 feet can achieve a delay of 6.4 seconds at 7.5ips, perhaps at fairly good fidelity if no regeneration is used. A large separation like this may be difficult to accomplish for technical reasons and even if successful provides delay times that are well within the reach of all modern digital delays. The technique is prohibitively cumbersome and involves the use of two reel-to-reel tape machines (preferably identical). At a certain point the separation distance between the two machines will introduce tension inconsistencies in the tape path that will render the process infeasible if high fidelity is to be pursued. My point in mentioning this beast of a technique is that the fluctuation in tape tension and even the failure of the tape delay due to excessive distance could quite reasonably be a part of a piece. Continuing along this vein, the theatrical aspects of tape delay provide some unique opportunities. Tiny digital delay boxes have a magic of their own, but the concreteness and easy comprehensibility of the tape delay process could add considerable audience interest to a performance piece.
Digital delay devices have opened up new possibilities in processing: it is not as if digital delays are hopelessly sterilized versions of analog or tape delays. The decreasing price of RAM makes more and more delay time available for lower prices, opening up new compositional techniques. Long delay times (which I would arbitrarily define as greater than 10 seconds) can participate in the formation of large structural elements of a piece. Before long delay times were available, delay always functioned on a local temporal level.
The expense of high-fidelity long delays is still prohibitive for many users. This leads my next topic, MIDI delay. One of the most important properties of MIDI delay devices is their ability to deal with long delays at low cost. I am not familiar with dedicated MIDI delay devices so my discussion is based on my experience using general-purpose computers as MIDI delays. Since the MIDI data stream contains far less information than that of digital audio, both in theory and especially in practice, one can see why long MIDI delays are easier to create. They can be made using any general-purpose computing platform with a MIDI interface. The implications of the use of MIDI delay go much further than long delays. They allow exploration of what I would call ultra-long delays: delays measured in minutes or even hours. The use of ultra-long delay times is made possible not so much by MIDI's low maximum bit rate as by its bit rate found in practical use, which is very low, especially discounting sysex and continuous controller data.
The MIDI data stream contains irregularly spaced data, so continual sampling and storage as in the case of digital audio is unnecessary. This property of practical MIDI data streams allows for the storage of time-stamped data over long periods of time. Time stamping is a process in which a piece of data is received and stored along with some timing information. This timing information could be a record of when the data was received, or a record of when it should be re-transmitted. To make a delay, a queue of received data is maintained and then re-transmitted according to the time stamp data it contains. This sort of delay scheme is not applicable to digital audio because of the regular sampling rate needed. (One exception to this is that if the program material to be delayed is known to contain many silences, only the information-carrying samples can be saved and they can be time-stamped like MIDI data. I know of no devices that use this scheme to accomplish digital delay. It is not surprising that this technique has never been implemented because it applies only to a very uncommon type of program material.) The use of ultra-long MIDI delay has important implications for the overall structure of a piece and even the piece's length.
In addition to the use of static delays, the live parametric control of delay devices opens up many avenues for compositional exploration. Digital delay devices usually have the ability for parameter control via MIDI. This control can take two forms: bulk sysex dump of an entire delay "patch" or dynamic control of a single parameter. Delay parameters might include delay time, clock rate, internal LFO rate, feedback gain, filtering, etc. Dynamic parameter control has compositional implications in that it changes the role of the performer from a producer of sound to a modifier of sound. The only analogous traditional performance role is that of a conductor.
Any MIDI data message could be used to control the dynamic delay parameters but an intermediary MIDI processor might be necessary to accomplish this. As an example, consider a delay processor which allows control of delay time by Mod Wheel messages. If one desires to control delay time by Note On values, it would be necessary to insert a MIDI processor in between the controller and the delay device. One good way to accomplish such a MIDI processing task would be using Opcode's Max. This sort of message conversion is a trivial programming task in this environment. Hardware devices may in many cases be able to convert messages as well; the Lexicon MRC comes to mind as a device which can probably do this with ease.
New directions in MIDI controllers is a subject which deserves a separate paper for itself, but I will mention some topics here. The dominant MIDI controller of the past and present is the unweighted, organ-like keyboard equipped with a spring loaded, center-return wheel ("pitch bend") and a wheel of approximately 100 degree rotational path ("mod wheel"). Percussion controllers such as the Roland OctaPad and the various DrumKat products have also gained some popularity. Slider control banks like that of Peavey and the Lexicon MRC are also becoming more popular as MIDI-controlled, continuous parameter devices like effects units and mixers proliferate.
In acoustic music, it is obvious that the choice of instruments has a profound effect on the sound of a piece. In MIDI music, the fact that any MIDI controller can control any sound module may tempt one to think that the controller can now "sound like any instrument." This is a seriously flawed conclusion because the physiological aspects of the controller impose very specific limitations on the data stream which it will emit under human control. One of the most obvious examples of this effect can be seen in the differences between what might be considered the two fundamental controller types: continuous and discrete. The keyboard is a discrete controller, in terms of its Note On data stream, but a mod wheel is a continuous controller. I refer to continuous and discrete not as descriptors of the MIDI data stream but rather the physical actions which are used to bring about this data transmission. As one can easily imagine, the use of a mod wheel as opposed to a keyboard as a controller will have profound effects on the final sound of the piece. This hypothetical case makes the assumption that the two controllers would be used in such as way as to transmit the same type of messages.
The purpose of my comments above is to suggest that the current homogeneity of MIDI controllers has unnecessarily decreased the variety of live MIDI music. Therefore the investigation of alternate controllers is in my opinion an important new direction in MIDI music. I might add here that alternate controllers need not necessarily take the form of physically different instruments. I believe that the innovative application of an extant controller is tantamount to the use of a physically new controller. A controller's MIDI data stream can be processed in such a way as to render a new, virtual controller. A simple example of this is keyboard remapping in the context of the control of pitched sounds. The use of a keyboard as a controller influences the music strongly in terms of what intervals appear in the music, both simultaneously and sequentially. If a keyboard's output is processed in such a way as to make the steps between consecutive keys greater than a half-step (and hopefully non-uniform!), I propose that this would create a new virtual controller which can be used to produce a very different type of music than would be possible using the traditionally configured keyboard. Again, Opcode's Max presents and ideal platform for such MIDI processing.
A significant hurdle to be overcome in the use of new controllers, be they virtual or physical, is the fact that it is difficult for performers to learn to play a new instrument. This is a difficult problem for which I can suggest three partial solutions which by no means sum to a whole solution. First of all, performers will be more willing to put extra work into learning new instruments if they are convinced that the use of these instruments plays an indispensable role in the piece. In other words, it is important to establish that extant instruments could not be used to achieve the same effect. Second of all, the composer [him/her]self could perform on the new instrument, thus avoiding the potential recalcitrance of performers. Third, performance difficulty can be made less of an issue through the use of new types of composition. In the traditional music world, the notions of virtuosity and accuracy in the execution of a piece are central to the value system that has successfully been inculcated in generation upon generation. I suggest that these values can be successfully challenged while producing music of value. The challenging of these values does much to reduce the formerly traumatizing worries about "wrong" notes, missed entrances, and botched passages. Once the notions of "right" and "wrong" are either abolished or redefined, there is less reason to view the use of a new instrument as a burden. The whole topic of reconsideration of musical values is a fascinating one, but at present the subject must stay that of delay.
Some older digital delay units have only front-panel, knob-based parameter control and perhaps a voltage-controllable clock rate. This makes these devices no less viable as performance instruments. In fact there is very little difference between controlling clock rate on a PCM 70 via a mod wheel and controlling the clock rate on a PCM 42 via the front panel knob. Actually, there is very little difference in terms of end musical result, but there is considerable difference in that the PCM 70 setup is far more complex and expensive. The use of voltage control for clock rate is particularly well suited to experimental applications since innovative voltage sources are not as hard to make as innovative MIDI controllers. One reason for this is that a physical MIDI controller is a complex piece of electronics to build. Another reason is that the historically, voltage control has been more closely tied to experimental music than MIDI has. Therefore there are many voltage-control sources left over from "the old days." An important area in the development of alternate MIDI controllers is the use of voltage controllers attached to CV to MIDI converters. Using such a device, extant or user-built voltage sources can be easily interfaced with the MIDI world. Photosensitive MIDI controllers, MIDI envelope followers, and brainwave to MIDI conversion come to mind as a few possibilities in CV to MIDI conversion. A CV to MIDI converter is a relatively simple piece of electronics which is not prohibitively expensive; for instance, PAiA Electronics offers a device which incorporates 8-channel CV to MIDI conversion for $400.
One exciting area of delay is the use of a controller which controls not only a delay but a sound source as well. The idea of parallel control is one which has strong precedent in experimental voltage-controlled electonic music. My piece Time Distortions involves the use of a MIDI delay whose delay time is controlled by the value of note-on messages received. At the same time, these note-on messages are controlling a sound source. In this way, high notes can be given long delay times and low notes short delay times. A striking consequence of this technique is that block chords come back as arpeggios with slight rhythmic irregularities.
I hope this article has shown that although delay appears to be a simple effect, this appearance is due more to the current lack of creativity in its application than to any property it inherently possesses. There are a variety of effects which may be accomplished using delay. It is a frequency domain effect if the delay times are kept short and there is a high degree of regeneration. The well-known technique of flanging is part of this frequency domain processing capability of delay. At longer delay times, it becomes a time domain effect. Here slap echo is its most familiar form.
Changes in clock rate can be used to create a delayed, temporary pitch shift. If the clock rate is changed smoothly, the resultant pitch shift is "self-correcting": the pitch will eventually slide back to its unshifted frequency. To understand this phenomenon, assume we have a delay time of ten seconds. If the clock rate is doubled over a period of 1 second, a 1 second upward octave slide in pitch shift is heard, followed by 8 seconds of octave-shifted material, followed by a 1 second downward octave slide.
Short delay times can be used in combination with stereo sound representation to create spatial effects. Long delays, especially MIDI delays, have important uses in the creation of macro-structural elements of a piece. The parametric control of delay makes new roles possible for performers.
The situation with delay is typical of electronic music today: tremendously powerful and relatively inexpensive devices are at now our disposal. We are left with the challenging and exciting task of finding ways to employ these technologies in creative ways.