Last month, I discussed the basic characteristics of multi-effects processors, focusing on delay, phasing, and reverberation effects. In this month's column I'll continue my survey of different types of effects commonly featured in modern effects processors, including distortion, pitch-shifting, and amplitude modulation. I'll deal with dynamics processing another time.
Though the focus of this article is on hardware, many of the processes I'll describe can also be found in software plug-ins or stand-alone applications. Many of these plug-in effects offer real-time tweaking, but regardless of whether they work in real time, their underlying concepts are similar to those of their hardware counterparts.
A number of common effects are based on amplitude modulation (AM), a technique in which a modulator signal controls the amplitude of an input, or carrier, signal. Tremolo, for example, is a pulsating effect that has been popular since it was introduced in guitar amps in the 1950s. It's produced by a sine or triangle wave modulating the amplitude of an audio carrier. Autopanning sweeps a signal back and forth between stereo channels, under the control of an LFO. Usually, tremolo and autopan effects have similar parameters: rate, depth (amount of effect), and LFO waveform.
In AM synthesis, the modulator and carrier signals are both in the audio frequency range (approximately 20 Hz to 20 kHz). AM synthesis generates sideband frequencies in the output spectrum. Sideband frequencies are equal to the sums and differences of the input frequencies. Often, the sidebands are not harmonically related to each other, producing distinctive clangorous timbres. There are several forms of AM synthesis. One of the most popular is ring modulation. A true ring modulator produces only sidebands, suppressing the original input signals. For an in-depth discussion of amplitude-modulation techniques, see "Square One: Modulation Synthesis Methods" in the March 1999 issue of EM.
FIG. 1: Transfer functions can be visualized as plots on a two-dimensional graph. Pictured here are the functions that would be used for (a) a linear device; (b) a hard-clipping device; and (c) a soft-clipping device.
To understand how distortion devices work, it is useful to understand the concept of a transfer function. A transfer function maps a device's input amplitude values to its output amplitude values. By plotting inputs against outputs, you can visualize the transfer function of a given device. Fig. 1a shows a perfect linear transfer function, in which input values map to identical output values. A device with a linear transfer function may amplify, invert, or attenuate a signal (in Fig. 1a the signal is passed through untouched), but it doesn't modify the signal's waveshape. Therefore, a linear device doesn't introduce any spectral changes. An ideal amplifier would have a completely linear transfer function.
Fig. 1b shows the transfer function of a device that is linear through part of its input range, but that will not exceed certain positive and negative levels. The effect of this nonlinearity is that the signal is clipped when its value exceeds the linear range (see Fig. 2, top track). Clipping is a simple example of nonlinear waveshaping.
FIG. 2: Clipping will occur when a signal''s output exceeds the audio system''s dynamic range. The top track shows an example of hard clipping; the bottom shows an example of soft clipping.
This hard clipping introduces odd-numbered harmonics into the spectrum. Hard clipping is characteristic of solid-state devices, which is one reason why many guitarists abhor solid-state amps. (Paradoxically, many tube-loving guitarists buy fuzz boxes, which are nothing but simple hard-clipping devices, and put them in the signal path before their tube amps!)
Hard clipping is by no means the only type of nonlinear waveshaping. Fig. 1c shows a nonlinear transfer function that distorts the signal in a more subtle way. The output becomes more distorted as the input approaches maximum positive or negative amplitude. The transition from undistorted to clipped signal is gradual, so the sonic effect is much less jarring. Distortion produced by this type of "S-shaped" transfer function is often called soft clipping.
The bottom track in Fig. 2 shows a soft-clipped sine wave. Soft clipping is characteristic of tube devices and lends "warmth" to the signal. The voltage-controlled amplifiers (VCAs) employed in many Moog analog synthesizers had S-shaped transfer functions, although they were not tube devices. The famous "fatness" of the Moog sound was partially due to these VCAs.
In multi-effects processors, distortion effects usually involve some form of hard or soft clipping. But a distortion processor could use any imaginable curve as a transfer function. The best way to explore this is to use a software-based processor that lets you draw an arbitrary transfer function and apply the curve to your audio input. Digidesign's Turbosynth SC and Cycling 74's Pluggo package (both for Macintosh) include waveshaping processors, as does Sounds Logical's WaveWarp for Windows.
For decades, musicians and engineers have sought ways to alter the pitch of a signal without changing its duration or introducing significant distortion or other artifacts. This Holy Grail has yet to be found. However, technology has progressed to the point where you can expect a midrange effects box to achieve fairly clean pitch transposition over a limited range.
Real-time pitch-shifting is usually accomplished using some variant of the following technique. Samples are read into a memory buffer at your system's sample rate and read out at a different rate-faster to raise the pitch, slower to lower it. To preserve the duration of the signal, samples are either repeated (if the pitch is being raised) or dropped (if the pitch is being lowered). This process, needless to say, can introduce nasty discontinuities and other artifacts into the signal; these become worse as the pitch-shift ratio increases. Designers of pitch-shifting devices have devoted considerable ingenuity to reducing pitch-shift artifacts, devising algorithms that adapt to signal behavior.
Most pitch-shift processors offer a range of 1 or 2 octaves, but the clean range will often be quite a bit smaller. You should listen critically before buying: pitch-shift quality varies considerably from one algorithm and product to the next.
One of the most common pitch-shift applications is chorusing. Acousticians long ago observed the "chorus effect" that takes place when a group of similar instruments (such as a violin section) plays in unison. Each individual instrument plays at a slightly different pitch, but the ensemble does not sound out of tune. The beating and phase-shift effects produced by minute tuning differences give the group a massive sound that fluctuates pleasingly over time.
Chorusing is achieved by pitch-shifting a signal up or down by a few cents and mixing the shifted signal with the original signal. You can achieve a more massive effect if you pitch-shift several copies of the signal by different amounts. One variant of this technique is to route the copies to different stereo channels. Chorus effects are often enhanced by varying the pitch shift slightly with an LFO, by a small amount of feedback, or by a bit of delay before detuning.
A harmony processor generates one or more harmony lines from the input signal. The simplest harmony processors transpose all input by a fixed interval. This approach, considered crude nowadays, has two drawbacks. First, it quickly bores the listener. Second (and worse), it produces notes that are out of tune. (Harmonize the first few measures of Beethoven's Ode to Joy in straight major thirds to get an idea of how bad this can sound.)
"Intelligent" harmony processors let you specify a scale and key, and they follow the input melody at the diatonically (or modally) correct intervals. If your music modulates a lot, you can change the scale and key parameters via MIDI. Dedicated harmony processors, such as the DigiTech Vocalist series, can perform impressive feats of multipart harmonization and pitch correction. Some high-end multi-effects units, such as the Lexicon PCM 81, also offer sophisticated pitch processing.
Varispeed imitates the sonic effect of changing the speed of a good old-fashioned analog tape deck or turntable. In the digital domain, this is easy to do-just vary the playback sample rate. When the sample rate changes, so does the pitch, producing the classic "munchkinization" effect. Given the amount of R&D that has gone into avoiding this effect in pitch/time processing, it's a bit ironic that varispeed has retained its popularity. Varispeed is always good for an audience laugh, though; you'll hear it at some point in almost any episode of Ally McBeal.
READ ALL ABOUT IT
In my articles on multi-effects processing, I've only been able to touch upon the basics of this very extensive subject. In fact, any of the effects that I've mentioned could easily be the subject of an entire article. To explore further, you might want to read through some of the excellent reference materials on the subject, such as the signal-processing sections of Curtis Roads's Computer Music Tutorial (MIT Press, 1996).
Multi-effects open up a huge sonic universe, and the more you know about them, the better you can put them to use in your work. Take a look at the hardware and software in your studio and see what sort of effects are lurking under the hood!
John Duesenberry's electronic music is available through the Electronic Music Foundation. Check the EMF catalog at www.emf.org.