POWER PITCH SHIFTING

by J. Arif Verner

No, it's not about baseball-it's all about how to get the most from your effects processors.

For some people, the whole concept of pitch shifting is shrouded in mystery; in other circles, it has developed a reputation for having poor performance. And let's ace it, there is nothing worse than bad harmonization (unless you're after the "Darth Vader Meets the Chipmunks" effect).

However, the quality has improved considerably over the past few years. The effect has also become far more widespread-pitch shifting algorithms are now included in lots of mainstream multi-effects processors.

In this article, we'll take a look at the mechanics, musical applications, and the ins and outs of the pitch shifting process. We'll also address pitch shifting acoustic instruments, electronic instruments and, of course, the voice.

For the sake of simplicity, let's dissect pitch shifting into four main applications: detuning, pitch correction, harmonization, and sound effects.

The basics

Sampling in multieffects processors is every bit as critical as in a sampler or digital recorder-the better the sampling quality, the better the audio quality. And two ingredients that affect quality are the sampling rate and bit resolution.

The sampling rate is measured in thousands of cycles per second-kilohertz (kHz). For benchmark comparisons, something called "CD quality" is sampled at 44.1 kHz. This means that 44,100 samples are taken per second. 48 kHz is also a standard in many studio applications. All things being equal, the higher the sampling rate, the better the audio quality, although there are many situations in which one would prefer to use 44.1 rather than 48 to avoid sample rate conversions.

The other key ingredient in the sampling process is the bit resolution. This affects the dynamic range and quality of the material. Again, 16 bits is the CD and DAT standard; as technology moves on, we're seeing 20 and 24 bits enter the picture (nothing like planned obsolescence).

So the chain of events goes something like this: audio is routed into the signal processor via the analog­to­digital (A/D) converters. Thousands of samples are recorded per second, and each sample is saved as a specified number of bits. When the music comes out, the procedure is reversed: the signal is converted from digital to analog through the D/A converters. Since we live in a stereo world, there are two channels.

To raise the pitch of a sample, audio data is compressed. This makes the sample play faster, similar to speeding up a tape recorder. The compression also makes the audio segment smaller. To compensate, the processor takes part of the signal and adds it back into that audio section. That accounts for the difference in size. Unfortunately, when new audio is added, the various parts do not always connect perfectly. This creates a problem with digital artifacts (unwanted byproducts).

Conversely, pitch shifting downwards expands the sample. Therefore, portions of the sound need to be removed. And likewise, the reconnected parts are also subject to artifacts. It's the same process, only in reverse.

The problem can also be compounded by other factors. The com plexity of the material plays an important part. A single instrument or voice will have less artifacts than pitch shifting your big jazz band or a cappella choir. Also, the size of the transposition is critical. The larger the pitch shift, the more the data is manipulated, hence the greater the artifacts. So pitch shifters have a lot of thinking to do (in a split second) to get it right. This explains why the process can be so finicky.

Detuning

Pitch shifting plays a subtle role in a process called detuning. Effects like chorus and flange use detuning as a key ingredient. If you look at the signal flow of a stereo chorus, you'll see an initial input signal being sent into the processor. The signal is then split into three parts: a dry signal and two processed signals. The two processed signals are routed as left and right channels. The chorusing is achieved by taking the left and right channels and adding a slight delay with a small amount of detuning. One channel is usually detuned slightly sharp while the other is slightly flat.

Added to this recipe is an LFO (low frequency oscillator) that modulates the detuned signals (modulates = delays by a moving amount). This creates variation in the effect. At the output, the dry signal is mixed back in with the processed signals. The final chorus effect makes the music sound thicker and richer. There are other parameters used to enhance chorusing, but this is the basic model.

To designate detuning values, many signal processors use the term "cents." Mathematically, a cent equals 1/100 of a semitone. So 100 cents would be a half step and 1200 cents an octave.

Detuning is not only used in signal processing but also in synthesis. One technique useful with synthesizers is to layer two sounds together (either the same sound or two different ones) and detune each of them slightly. For example, increase one sound by +3 cents and decrease the other by ­2 cents. This helps create a fatter patch. As we'll see, pitch shifting is useful in both acoustic and electronic music-the concepts are the same.

Pitch correction

How often have you recorded a perfect vocal track with the exception of one or two flat notes? Pitch shifting to the rescue. If you're using a tape­based system (digital or analog), simply route the signal through the pitch shifter and re­record the vocal line onto another track. Use the pitch shifter's bypass switch to pass the signal through it unaffected; when you get to the notes that need to be corrected, hit the bypass switch to engage the pitch correction.

On many hard disk systems, it's even easier: you simply highlight the beginning and ending points of the regions that need to be corrected and then specify the amount of transposition necessary; the computer does the rest.

Another problem you may encounter is correcting pitches of instruments that are not quite in tune. How many great guitar players do you know who can't tune their axe? (Okay, how many drummers does it take to change a light bulb...) And what about that kazoo you bought in the Wala Wala Islands during your vacation-only to find it's tuned to A=439.79?

In these cases, run the instrument or track through the pitch shifter in the pitch­correction mode. This should take care of any pitch idiosyncrasies that are otherwise impossible to deal with.

If these are the kinds of pitch corrections you need, hopefully the problems are minor. If that's the case, detuning the sound by a few cents should do the trick. If the track is in the background, there should be no problem. However, the more up­front the track is in the mix-as with a lead vocal-the more the processing becomes noticeable. Realize that the larger the pitch discrepancy, the less convincing the results.

Harmonization

This broad category is more in-your­face than simple detuning. The idea is to add specific notes to the harmonic range of the source material. Keep in mind that the ear hears higher frequencies first. So your best bet might be pitch shifting downwards. In this way, the original material will be heard first with the pitch shifted harmonies underneath. However, if you need to pitch shift up, try pulling the level of the processed track down in the mix. Remember, pitch shifting should play a supporting role in the music. This will make your sound more realistic.

Electronic instruments like sample playback synths can also be detuned and pitch shifted through menu parameters. Of course this is done internally in the digital domain, which simplifies the process. Because of their timbral characteristics, synths can often be pitch shifted several octaves with convincing results. For voice and acoustic instruments, an octave stretch is probably the limit. Also, different effects processors provide different pitch spans. Some units can stretch the sound as much as eight octaves-four octaves below and four above the original pitch.

Depending upon the sophistication of your unit, you should be able to do multiple harmonizations. While some can handle three pitches, others wil1 do five or eight different harmonies. This may seem like candy to the ears, but the results can be more like dill pickles if not handled properly. With pitch shifting, the axiom, "less is more" often applies.

Try detuning notes slightly and combining one or two harmonic intervals. A good place to start is with the harmonic third and fifth.

Keep the original signal, like a vocal track, panned in the center. The pitch shifted harmonies can be panned left to right. Put the level of the main track up front. The supporting harmonies should be mixed at a lower level. This will add realism and depth to your tracks, while accentuating the stereo mix.

Let's look at some more real world applications for pitch shifting. Doubling is a basic technique. This can be handled in several different ways. The most obvious trick is to run your original sound into the shifter and set it to unison (or O cents). This will produce a mirror image of your source material-great for strengthening tracks.

You can layer the double under the original track or pan as needed. Add a bit of delay, somewhere around 20 milliseconds. This will fatten your tracks, producing a sound similar to a vocal doubling. Also, try setting two voices to unison. And as with the previous example with synths, detune each voice a few cents in opposite directions. Pan accordingly. You also might want to experiment with a bit of modulation.

As for that ultimate bass sound, try shifting the pitch down an octave and mixing it in with your instrument. This works well with bass guitar or a bass synth patch. Drums and percussion also respond well to detuning and transposition. Shifting upwards will brighten your sound while downwards will add more punch. With guitar, set the shifting up an octave, add a touch of delay, and you have a 12 string guitar.

Synths and samplers can use any of the above techniques. As mentioned before, parameters are controlled internally. However, you might want to try a combination of an external processor with the synth's internal settings. Each type of pitch shifting produces a different kind of sound.

Another point to consider is the use of pitch shifting in a live situation. If you plan to harmonize your vocals, make sure the other instruments don't bleed into your mic, or you're going to get a pitch shifted band!

For instruments like guitar, there are several choices. You can plug directly into the pitch shifter via your preamp or route it through the board or mixer. If you go direct, reduce the amount of the wet/dry mix. And conversely, you may have to increase the mix ratio through the board. This is generally true for most signal processors, regardless of the effect.

Other cool features on some pitch shifters include the ability to specify a harmonic range or individual scales. This has been tagged "intelligent pitch shifting."

Another interesting effect is gender­bending. This allows you to choose vocal characteristics of male, female or neither. And don't forget about MIDI. By using external continuous controllers, you can do things like fade in pitch shifts, control LFO speed, regulate the wet/dry mix, change harmonies, etc. These MIDI events can be recorded into a sequencer via SMPTE synced to a tape recorder. So the next time you work on the mix, all of those moves will be automated.

Sound effects

This is where things get interesting. Depending upon the complexity of your box, different parameters can be bent to create some interesting results. But first, let's examine some of the tools of the trade. In addition to pitch settings, processors usually have adjustments for volume levels, pan positions and delay times. Some of the other toys include LFO waveforms, speed, depth, delay, and effects regeneration. Again, each voice can occupy its own space in time with different parameter settings.

For starters, try out the "crystal" effect. The key here is a high regeneration level of 60-70%. Regeneration (sometimes called feedback) feeds the output back into the input. The interesting thing about this parameter is that the sound keeps transposing itself with every new cycle. Set your pitch to 700 cents and a delay of 125 ms. Adjust the mix setting to about 75% wet. Play a note and listen to the sound cascade upwards into a spiraling cluster of echoes. It produces a beautiful crystalline effect. You can also set the pitch downwards to ­700 cents. This creates a downward spiral.

Here are some other ideas to try out. With a stereo chorus or flange, detune each voice to a different pitch. A good range might be from ­10 to +10 cents. Set the feedback regeneration to +95% on one channel and ­95% on the other. Adjust the LFO frequency to 3.00 Hz. Hit a fat bass pad on your synth and you'll be ready to score a sci­fi soundtrack. The result should sound like a metallic tube swirling between your speakers. If you need to create an echo effect, try setting your desired pitch range with a delay of about 250 ms. Pan the original signal hard right and the echo hard left. This pitch shifted echo should bounce across your speakers instantaneously.

To take it a step further, you might try constructing an octave scale arpeggio. With an eight voice harmonizer, set the pitches to the following cents: 1 = 0, 2 = 200, 3 = 400, 4 = 500, 5 = 700, 6 = 900, 7 = 1100, 8 = 1200. Set the delays to 50 ms intervals from 50 ms to 400 ms. The result should be an arpeggiated scale. Of course, you can plug in any harmonic interval that you want.

Yet another variation on the theme is a simple major chord-with a twist. The basic tuning with four voices would be: 1 = 0, 2 = 400, 3 = 700, 4 = 1200. Set the delay intervals at 200 ms from 0 ms to 600 ms. Pan voice 1 to the far left, 2 at left center, 3 at right center, and 4 at far right. As you can imagine, when you hit a note, it will change into a chord slowly panning across the sound stage. Keep in mind that different instruments will give different results with the pitch shifting process. And don't forget to try these parameter settings with vocals.

Combining the pitch shifting of an effects processor and a synth can create some bizarre results. How many times have you heard the rule that you should only play samples within their appropriate range? Yes, it's good advice unless you really want to break the rules. Okay, here goes. Pick an orchestral sample like an oboe or slow cello. Set its coarse tuning to ­48 semitones. Go to your pitch shifter and transpose it up to +2400 cents. If your results are anything like mine, it should sound like a herd of electronic dinosaurs tearing through your studio! [Check out the D70 samples on Playback 8 for examples of this effect, particularly "Largo Tree," a drastically pitch­shifted bell­tree sample-MM]

One last cool idea is to sample your pitch shifter. Play or sing a note into your pitch shifter and record it into your sampler. Edit the sample as need be. Then connect your sampler back into your pitch shifter. Play the sample from your keyboard and process it again through the pitch shifter. Don't be afraid to go into unknown territory. Guaranteed, you'll come up with some amazing results. If this turns out well, then create a multi­sample up the keyboard.

If your effects unit doesn't have all the parameters discussed above, then use what's available. Every manufacturer has its own bag of tricks embedded into their. gear. The best way to get familiar with the power of pitch shifting is to experiment. I would venture to say that the 80/20 rule also applies to most processing equipment-that is, 80% of the users only use 20% of the power. So get your money's worth and stretch your effects to 100%.

What's available

New developments in digital recording have made the world of pitch shifting quite exciting. With many hard disk systems, you can pitch shift your audio entirely in the digital domain. This reduces artifacts and noise that normally creep into the signal chain during the conversion process.

Another great feature is the ability to change tempo maps while maintaining accurate pitch intonation- often called "time stretching." So if you need to reduce a 30­second spot to 25 seconds, no problem. Other new applications include the integration of MIDI and digital audio. Most professional sequencers now offer this feature. This makes transposition a breeze with both audio tracks and MIDI data.

As you know, chorus and flanging are part of the bread­and­butter effects found in most signal processors. However, a quick look at the marketplace shows many companies including harmonization within their multi­effects processors. That's the good news.

The bad news is that quality can vary considerably. However, several manufacturers have made a dedicated commitment to advance this technology. These folks include Eventide with their professional line of DSP Ultra­Harmonizers, Digitech with their popular Vocalist and DHP­33, and more recently, Lexicon with their PCM 80 Pitch Shift card. There's also a software pitch shifting process built into Mark of the Unicorn's Digital Performer 1.7 that does a good job with vocals, and Emagic Logic Audio 2.6 will have a similar process when it comes out.

If you're looking to add pitch shifting to your musical arsenal, take some time to audition the gear. There are a lot of options to consider. Unfortunately, no one piece of equipment is perfect for all situations. Most importantly, let your ears be your guide...but also check the specs. What is the sampling rate, the bit resolution, and the signal­to­noise ratio? And of course compare all the features to your individual needs- and to the size of your wallet.

As technology moves forward, the harmonic mystique wil1 slowly disappear. But until then, get your sonic paint brushes ready. The versatility of pitch shifting will keep you busy for a long, long time.