″I′ll tell you something.
I am a wolf,
But I like to wear sheep′s clothing.″
— ″TEMPTATION WAITS,″ GARBAGE, VERSION 2.0 (ALMO SOUNDS, 1998)
Use a playback sample rate that is higher than the original sample rate used when the audio was recorded, and the pitch goes up. Play an audiotape at a slower speed than intended and the pitch of the recording goes down. Somewhere in this simple principle lies an opportunity for audio exploration and entertainment.
This chapter explains the fundamentals of pitch shifting and initiates an inventory of the production potential.
The principle of shifting pitch is straightforward. The basis for the effect is the delay. The same device stretched to such creative limits in Chapter 9 has still more applications.
To appreciate the elegance of pitch shifting, a bit of math is in order (follow along with Figure 10.1). Figure 10.1(a) shows one cycle of a simple sine wave with a chosen period of 4 milliseconds (ms) and, therefore, a frequency of 250 Hz. This sine wave is seen to complete exactly one cycle every 4 ms, and the frequency calculated courtesy of the following familiar equation:
where f is the frequency (hertz), and T is the period (seconds), which is the time to complete precisely one cycle (see Chapter 1). Using Equation 10.1 for Figure 10.1(a):
Figure 10.1 Pitch shifting through a variable delay.
Figure 10.1(a) labels some key landmarks during the course of a single cycle, using the letters V, W, X, Y, and Z. It all starts at point V. Here, at time equals zero, the sine wave has an amplitude of exactly zero and is increasing. It reaches its positive peak amplitude at W, taking exactly 1 ms to do so. It returns again to an amplitude of zero at the halfway point (time equals 2 ms), labeled X. The amplitude looks at this instant a lot like point V. While the amplitude is the same, notably zero, it is decreasing through point X. Y is the point of maximum negative amplitude and occurs 3 ms after the beginning of the cycle. At Z, the sine wave returns 4 ms later to what looks exactly like the starting point: the amplitude is zero, and increasing. Follow these points as some signal processing is applied.
Run this sine wave through a fixed delay of 1 ms, and the situation described in Figure 10.1(b) results. Visually, one might observe that the sine wave appears to have slipped along the horizontal time axis by 1 ms. Looking point by point, Table 10.1 shows what happens. Point V originally occurred at a time of 0 ms. Inserting a 1-ms delay on this signal, point V now occurs at time equals 1 ms. The other points follow. Point Y, for example, occurred undelayed at a time of 3 ms. After it has been run through a fixed, unchanging delay of 1 ms, point Y is forced to occur at a time of 4 ms.
Here′s the mind bender. What happens if the delay isn′t fixed? What if the delay sweeps from a starting time of 1 ms and then increases, and increases, and increases? Table 10.2 summarizes. Here the delay changes at a rate of 1 ms per millisecond. That can seem a bit confusing at first. For every millisecond that ticks by during the course of this experiment, the delay gets longer by 1 ms. At one instant the delay might be 10 ms. Whatever audio is fed into the delay at this instant will appear at the output in 10 ms; 5 ms later, the delay time has swept up to 15 ms. Audio entering the delay unit now won′t appear at the output for 15 ms.
The simple act of increasing the delay by 1 ms each millisecond has a surprising result. The pitch of the audio is changed. Table 10.2 shows the time location of the sine wave landmarks both before and after the introduction of this steadily increasing delay. Point V initially occurs at a time of zero. At this time, the delay time setting within the delay line is also zero. Point V then remains unchanged and occurs at time zero. Skip to point X. It originally occurs at a time of 2 ms. By this time, the delay has increased from 0 ms to 2 ms. This delay of 2 ms leads point X to finally occur 4 ms after the beginning of this experiment. Doing the math point by point leads to the sine wave of Figure 10.1(c).
The key landmarks are identified. The result is clearly still a sine wave. But because it takes longer to complete the cycle, the pitch is known to have changed. Return to Equation 10.1. Looking at the new sine wave in Figure 10.1(c), one can calculate its frequency. The sine wave in Figure 10.1(c) takes a full 8 ms to complete its cycle:
The constantly increasing delay caused the pitch of the signal to change. A 250-Hz sine wave run through a delay that increases constantly at the rate of 1 ms per millisecond will be lowered in pitch to 125 Hz. An increasing delay lowers the pitch. It is also true that a decreasing delay raises the pitch. That is, start with a delay of 50 ms and decrease it by, say, 0.5 ms per millisecond, to 49.5 ms, 49 ms, etc. The result is a waveform exactly one octave higher.
In addition, the case studied above found that increasing the delay at a rate of 1 ms per millisecond raised the pitch by an octave. It is also possible to change the pitch by two octaves, or a minor third, or a perfect fifth — whatever is desired. One need only change the delay at the correct rate.
The underlying methodology of pitch shifters is revealed. A pitch shifter is a device that changes a delay in specific controlled ways so as to allow the user to affect the pitch of the audio.
Naturally, there are some significant details that must be addressed. Return to the example in which the 250-Hz sine wave was lowered by an octave through a steadily increasing delay. If this effect were applied to an entire 31/2-minute tune, not just a single cycle of a sine wave, then one would find it necessary to increase the delay from a starting point of 0 ms to a final delay time of 210,000 ms (31/2 minutes equals 210,000 ms). That is, from the start of the tune an increasing delay is needed: 1 ms, then 2 ms, and so on. By the end of the tune, the delay time has reached 210,000 ms. This highlights two problems.
First, it appears a delay device capable of a very long delay time is needed. Many hardware delays start to run out of memory closer to a 1-second delay (1,000 ms). The more capable, more expensive delay lines might go up to maybe 10 seconds of delay. But a delay of hundreds of thousands of milliseconds is a lot of signal processing (i.e., memory) horsepower that is not typically available outside of a computer.
Second, the song that used to be 31/2 minutes long doubles in length to 7 minutes as the pitch is lowered by one octave. Consider the last sound at the very end of the song. Before pitch shifting, it occurred 31/2 minutes (210,000 ms) after the beginning of the song. By this time, the pitch-shifting delay has increased from 0 ms to 210,000 ms. Therefore the final sound of the pitch-shifted song occurs at 210,000 ms (original time) plus 210,000 ms (the length of the delay). That is, the song now ends 420,000 ms (that′s 7 minutes) after it began. The 31/2-minute song is lowered an octave, but doubled in length.
Simply increasing the delay forever as above is exactly like playing a tape back at half the speed it was recorded. The pitch goes down and the song gets longer. Pitch-shifting signal processors differentiate themselves from tape speed tricks in their clever solving of this problem. Digital delays can be manipulated to always increase, but also to reset themselves. In the sine wave example, what happens if the digital delay increases at a rate of exactly 1 ms per millisecond but never goes over 50 ms in total delay? That is, every time the delay reaches 50 ms, it resets itself to a delay of zero and continues increasing from this new delay time at the same rate of 1 ms per millisecond. The result is pitch shifting that never uses too much delay, and never makes the song more than 50 ms longer that the unpitch-shifted version. After all, the analysis above showed it was the rate of change of the delay that led to pitch shifting, not the absolute delay time itself. Any delay time that increases at rate of 1 ms per millisecond will lower the audio by one octave. Any delay time. So why not keep it a small delay time?
The devil is in the details. Getting the pitch shifter to reset itself in this way without being noticeable to the listener is not easy. It is a problem solved by clever software engineers who find ways to make this inaudible. Older pitch shifters ″glitched″ as they tried to return to the original delay time. Today, those glitches are mostly overcome by intense signal processing. Software algorithms can evaluate the audio and find a strategic time to reset the delay time, applying cross fades to smooth things out. Another approach might be to use two delay lines, sweeping them at the same rate but resetting them at different times. If one cross fades between them at opportune times, one avoids the moment when either delay is reset. Digital signal-processing approaches have become more advanced still, with the result that convincing pitch shifting is a staple effect available in any well-equipped studio.
Pitch-shifting effects are common in multitrack production — sometimes subtle, other times obvious; sometimes accidental, other times deliberate.
Before any discussion of effects deliberately using pitch shifting, it is worth noting that pitch shifting is a natural part of some effects already studied in this book. Recall the chorus effect that comes from adding a slowly modulated delay of about 20–50 ms >(see ″Medium Delay,″ Chapter 9). Chorus is difficult to describe in words. To study it, one must listen to it. A careful, critical listen to the richness that the chorus effect adds to a vocal or guitar reveals a subtle amount of pitch shifting. Beyond that blurring of things in time through the use of perhaps several medium delays, pitch shifting is a fundamental component of that effect known as chorus. Since a chorus pedal relies on a modulating delay, it introduces a small amount of pitch shifting. As the delay time sweeps up, the pitch is slightly lowered. As the delay time is then swept down, the pitch is then raised, ever so slightly. One can not have chorus without at least a little pitch shifting.
In Chapter 12, use is made of a common effect built, in part, on pitch shifting: the spreader. Here is a quick summary of this effect. The spreader is a ″patch″ that enables the engineer to take a mono signal and make it a little more stereolike. A single track is ″spread″ out by sending it through two delays and two pitch shifters, each hard panned left and right. The delays are kept short, each set to different values somewhere between about 15–50 ms. If they are too short, the effect becomes a range/comb filter; if they are too long, the delays stick out as distinct audible echoes. Delay left might be 17 ms, while delay right is 22 ms.
In using a spreader, the return of one delay output is panned left while the other is panned right. The idea is that these quick delays add a dose of support to the original monophonic track. In effect, these two short delays simulate some early sound reflections that one would hear if the sound were performed in a real room. The ″spreader″ takes a single mono sound and sends it to two slightly different, short delays to simulate reflections coming from the left and right.
That is only half the story. The effect is taken to the next level courtesy of some pitch shifting. Shift each of the delayed signals ever so slightly, and the mono source material becomes a much more interesting loudspeaker creation. Detune each delay a nearly imperceptible amount, maybe 5–15 cents. This is not a significant pitch change. An octave is divided into twelve half steps, representing adjacent keys on a piano or adjacent frets on a guitar. Each half step is further divided into 100 equal pitch increments, called cents. The pitch shifting called for in the spreader, then, is just 5 to 15% of a half step — all but imperceptible except to the most trained listeners. The goal of the spreader is to create a stereo sort of effect. As a result, one seeks to make the signal processing on the left and right sides ever so slightly different from each other. Just as unique delay times are selected for each side of this effect, choose different pitch shift amounts left and right as well — maybe the left side is shifted down 8 cents while the right side is shifted up 8 cents.
Like so much of what is done in recording and mixing pop music, the effect has no basis in reality. When adding delay and pitch shifting, the engineer is not just simulating early reflections from room surfaces anymore. The spreader makes use of common studio signal-processing equipment (delay and pitch shifting) to create a wide stereo sound that only exists in loudspeaker music. This sort of thing does not happen in symphony halls, opera houses, stadiums, or pubs. It is a studio creation, plain and simple.
Take this effect further and the sonic result might be thought of as more of a ″thickener.″ There is no reason to limit the patch to two delays and two pitch changes. Ample signal-processing horsepower in most digital audio workstations makes it trivial to chain together eight or more delays and pitch shifts. Strategic selection of unique delay times, pitch shift increments, and pan locations control the fullness, width, and coloration of the effect. It is likely to sound unnatural when used heavily, but a light touch of this effect on vocals, guitars, or keyboard parts can help those tracks sound larger, fuller, and more exciting. Modulate each of those delays like a chorus, and more complex pitch shifting is introduced. Added in small, careful doses, this densely packed signal of supportive, slightly out-of-tune delays will strengthen and widen the loudspeaker illusion of the track.
Pitch shifting is also used to zoom in and fix a problematic note. Maybe this sounds familiar: It is 5 a.m. It′s the fiftieth take of the song. It′s a great take. Then on the fifth repeat of the last chorus, the singer — tired from working all night long — drifts flat on the key word, the title word of the song, held for two bars: ″Gaaaaaaaarlic.″ The song simply does not work if the last time the listeners hear the title of the song, ″Garlic,″ the line is flat. No problem.
In the old days of multitrack production (and the reader is encouraged to try this approach), the sour note was sampled. It was then manually tuned using a pitch shifter, and the engineer′s sound musical judgment. It was raised or lowered to taste. Finally, the sampled and pitch shifted note was re-recorded back onto the multitrack. With the problematic note shifted to pitch perfection, no one was the wiser.
Alternatively, and more frequently, the engineer reaches for clever digital signal processes that can automatically shift pitch into tune. Such an effects device can monitor the pitch of a vocal, violin, or didgeridoo. When it detects a sharp or a flat note, it shifts the pitch automatically by the amount necessary to restore tuning. The engineer typically watches over this process, choosing when the pitch shifting is used and when the original performed pitch is to remain. These pitch-correction tools are very effective, but one has to be careful not to overuse these devices.
First, engineers should not overpolish their productions. Pitch shifting everything into perfect tune is not always desirable. Vibrato is an obvious example of the musical detuning of an instrument on purpose. And, if Bob Dylan had been pitch shifted into perfect pitch, where would folk music be now? There is a lot to be said for a musical amount of ″out-of-tuneness.″ Remove all the bends and misses, and we risk removing a lot of emotion from the performance.
Second, producers should not expect to create an opera singer out of a lounge crooner, or a pop star out of a karaoke flunky. There is no replacement for actual musical ability. If the bass player can not play a fretless, give them one with those pitch-certain things called frets. If the violin player can not control their intonation, hire one who can. Do not expect to rescue poor musicianship with automatic pitch correction. Use it to add to a stellar performance, not to create one. Musical sense and good judgment must motivate everything that is done in the recording studio. People generally want to hear the music, not the effects rack.
That said, Cher′s title track from the album Believe puts wholly unnatural, machine-driven pitch shifting front and center and makes it one of the hooks of this pop tune. The lifespan of this effect is likely short, but variations on this effect are there for the creative engineer to explore.
Pitch shifting need not be subtle, as shown in the many examples that follow.
Leslie
Hammond B3 organs, many blues guitars, and even vocals are often sent through a rather unusual device: the Leslie cabinet. The Leslie sound is a hybrid effect built on pitch shifting, volume fluctuation, and often a good dose of tube overdrive distortion. The Leslie cabinet can be thought of as a guitar or keyboard amp in which the speakers sound as if they rotate. A two-way system, the high-frequency and low-frequency parts work in slightly different ways. The high-frequency driver of a Leslie is horn loaded. The driver is fixed, but it fires into a rotating horn. The perceived location of sound is the end of the horn, which moves toward and then away from the listener as the horn spins.
It would be very difficult to spin the large, low-frequency driver to continue the effect at low frequencies. Instead, the woofer is enclosed inside a drum. The drum has a few large holes in it. While the woofer conveniently remains fixed, the drum rotates, opening and closing the woofer sound. The result is a low-frequency approximation of what the Leslie is doing with the horn at higher frequencies.
In addition, the spinning system has three speeds. The rotating horn and drum may be toggled between off (the amp stays on, but the rotating mechanisms stop), slow, and fast during the performance.
The sound of the Leslie is fantastic. With the drum and horn rotating, the loudness of the music increases and decreases — tremolo or amplitude modulation(see Chapter 7). With the high-frequency horn spinning by, a Doppler effect is created: The pitch increases as the horn comes toward the listener/microphone and then decreases as the horn travels away.
The typical example used in the study of the Doppler effect is a train going by, horns ablaze. That classic sound of the pitch dropping as the train passes is based on this principle. Sound sources approaching a listener with any appreciable velocity will increase the perceived pitch of the sound. As the sound source departs, the pitch similarly decreases.
The high-frequency portion of the Leslie sound is heard through a horn. The perceived location of that sound source is the bell of the horn. So while the driver sits fixed within the Leslie cabinet, the high-frequency sound source is moving. The Doppler effect results.
The low-frequency driver, housed within a spinning drum, does not experience a pronounced pitch shift. As the holes within the drum rotate by, the low-frequency signal gets slightly louder. Continued rotation of the drum causes the sound of the woofer to be again attenuated, when holes have not opened up the woofer to the listener/microphone. Amplitude modulation without pitch bending is the signature low-frequency sound of a Leslie. The net result of the Leslie system then is a unique fluttery and wobbly sound.
The Leslie effect is common wherever B3s and their ilk are used. Switching between the three available speeds of rotation, one can create a fast Leslie and a slow Leslie effect or no spinning Leslie effect, as well as the acceleration or deceleration in between. Listen to the single note organ line at the introduction to ″Time and Time Again″ on The Counting Crows′ first record, August and Everything After. The high note enters with a fast rotating Leslie. As the line descends, the speed is reduced. Listen carefully throughout this song, this album, and other B3-centric tunes, and the Leslie pitch-shifting vocabulary that keyboardists love will be revealed. Of course, sound engineers can apply Leslie to any track they like — guitar, vocal, and oboe — if you have the device, or one of its many imitators or simulators.
Big Shift
The straightforward pitch shifting that is the basis for the spreader and the thickener can be used in a more forward, don′t-try-to-hide-it way. The hazard with an obvious pitch shift is that it can be hard to get away with musically. Special effects — in movies and on some records — where a vocal is shifted up or down by an octave or more can have a comedic effect. If it is too low, the pitch-shifted vocal conjures up images of death robots invading the mix to eat entire villages. If too high, the singer becomes a gerbil-on-helium sort of creation.
In the hands of talented musicians, aggressive pitch shifting really works. Prince famously lowers the pitch of the lead vocal track and takes on an entirely new persona in the song ″Bob George″ from The Black Album. The effect is obvious, and an incredible story results.
No effort was made to hide the deeper than typically found in nature bass line of ″Sledgehammer″ on Peter Gabriel′s classic So. The entire bass track seems to include the bass plus the bass dropped an entire octave. The octave-down bass line is mixed right up there with the original bass. There is nothing subtle about it.
The pitch-shifting effect can be used to add two-, three-, or four-part harmony if the engineer is so inclined. Get out the arranging book though, because the pitch shifter makes it easy to inadvertantly add a dissonant interval. Each pitch-shifted note is a fixed interval above whatever note occurs in the source track. Only the octave will stay in tune with the harmony of a song. The perfect fourth and perfect fifth, tempting as they are, lead to a couple of nondiatonic notes when applied as a fixed interval above all the notes of the scale. The major third, applied rigidly to every note in a scale, will lead to dissonance and confusion; many notes will be out of tune with the key of the song.
A bit of additional signal processing is needed in the pitch-shifting device if it is to use minor thirds or major thirds and stay in the appropriate key. Digital signal processors offer this ability. The pitch-shift interval is variable, as needed to add harmony to a line.
The pitch shifting can be tied to Musical Instrument Digital Interface (MIDI) note commands enabling the engineer to dictate the harmonies from an MIDI controller. The pitch shifter is processing the vocal line on tape or disk according to the notes played on the keyboard. The result is a harmony or countermelody line with all the harmony and dissonance desired.
This production tool can reach beyond harmonies. One can use pitch shifting to turn a single note into an entire chord. String patches can sometimes be made to sound more orchestral with the judicious addition of some perfect octave and perfect fifth pitch shifting (above and/or below) to the patch.
It does not stop with simple intervals. Chords loaded with tensions are okay too when used well. Progressive rockers, Yes, put it front and center in ″Owner of a Lonely Heart″ on the album 90125. Single-note guitar lines are transformed into something more magical and less guitarlike using pitch shifters to create the other notes.
A final obvious pitch-shifting effect worth mentioning is the stop tape effect. As analog tape risks extinction, this effect may soon be lost on the next generation of recording musicians. When an analog tape is stopped, it does not stop instantly; it takes an instant to decelerate. Large reels of tape, like two-inch 24 track, are pretty darn heavy. It takes time to stop these large reels from spinning. If one monitors the tape while it tries to stop (and many fancy machines resist this, automatically muting to avoid the distraction this causes during a session), one will hear the tape slow to a stop. Schlump. The pitch dives down as the tape stops. This sometimes is a musical effect. It is not just for analog tape as Garbage demonstrates via a digital audio workstation effect between the bridge and the third chorus of ″I Think I′m Paranoid″ on their second album, Version 2.0.
Start Tape
The stop tape effect can be turned around, at least in the analog domain. Have the performer start playing before the tape recorder is rolling. Go into record immediately as the tape machine gets up to speed. The result, on playback at full speed, is a high-pitched descent into proper pitch. While the tape machine was coming up to speed, the signal was being recorded at an improperly slow speed. Playback at proper speed raises the pitch of that portion of the signal, and a unique pitch shift results.
The intro to ″Synchronicity II″ on the album, Synchronicity by The Police demonstrates this effect quite clearly. The squealing electric guitar in feedback pops into the mix with a sharp pitch bend courtesy of some coordinated start tape effects.
Artist: Cher
Song: ″Believe″
Album: Believe
Label: Warner Brothers Records
Year: 1998
Notes: The reference for just how far blatant pitch shifting can be pushed.
Song: ″Time and Time Again″
Album: August and Everything After
Label: Geffen Records
Year: 1993
Notes: This album abounds in good Leslie examples, but the intro of this song isolates the Leslie on the B3 on the right channel. It enters with fast rotation, and slows before the lead vocal enters.
Artist: Peter Gabriel
Song: ″Sledgehammer″
Album: So
Label: Geffen Records
Year: 1986
Notes: Pitch shift that bass, one octave down. Don′t hide it. Mix it in with the original bass. Larger than life.
Artist: Garbage
Song: ″I Think I′m Paranoid″
Album: Version 2.0
Label: Almo Sounds
Year: 1998
Notes: The pitch dives as if analog tape were slowly stopped between the bridge and the third chorus, though this is most likely a digital effect simulating the tape stop.
Artist: The Police
Song: ″Synchronicity II″
Album: Synchronicity
Label: A&M Records
Year: 1983
Notes: Intro reveals start tape effect. Electric guitar is made to feedback first. Then the tape machine is punched into record from a standstill. It ramps up to speed while recording. Played back at speed, the formerly slow bits are now pitched up. The sound of the machine achieving full speed in record becomes a pitch dive when played at uniformly correct speed.
Artist: Prince
Song: ″Bob George″
Album: The Black Album
Label: Warner Brothers Records
Year: 1983
Notes: Tracked at a fast speed, the protagonist vocal is low and ominous when played back at regular speed. This all-analog pitch-shifting effect supports a convincing performance by Prince.