Antares ATR-1
Real-time pitch correction that really works.
by Rob Shrock, Electronic Musician, February 1999
Most electronic musicians keep their eyes and ears open for that magic "black box" that can perform miracles on lackluster audio. We've seen several such devices over the past few years: aural exciters, harmonizers, subharmonic synthesizers, amp simulators, even devices that let you remove the lead vocal from a favorite classic rock song and make a fool of yourself at a family reunion.
Of course, the "magic" of these devices is really innovative design and implementation, with intelligent compromises. Such is the case with the Antares ATR-1 intonation processor. Known for its excellent Auto-Tune software plug-in (see the sidebar "Auto-Tune"), Antares has developed a hardware version of the same pitch-correction technology and packaged it in an easy-to-use, single-rackspace black box that operates in real time. And it is indeed magical.
The front panel includes a two-line, 40-character alphanumeric LCD, a large data-entry knob, and two cursor-navigation buttons. Also on the front panel are buttons for selecting Program, Song, and System Edit mode and another button to cycle through pages within the selected Edit mode. A Bypass button does exactly what you'd expect. A vertical row of six LEDs indicates the input-signal level, and a horizontal row of ten LEDs indicates the amount of pitch correction at any given moment.
The back panel sports a 7-pin DIN AC power-cord receptacle, a MIDI In port, and a 1Ú4-inch footswitch input. The mono, balanced, line-level audio inputs include 1Ú4-inch TRS and XLR connectors, but they can't be used simultaneously to mix two signals. The line-level output includes 1Ú4-inch TS (unbalanced) and XLR (balanced) connectors. A grounding switch lets you select circuit-board or chassis ground, which is a nice touch.
Defining the Pitch
The 52-page manual covers the basics of pitch detection and provides thorough information on the operation of the ATR-1. I strongly recommend that you read the manual; it's a short read, and it covers some concepts you must understand to operate the ATR-1.
The ATR-1 introduces a short processing delay (1 to 4 ms) that allows it to calculate the pitch of an incoming periodic waveform. Often, the algorithm begins calculating the pitch of the waveform before the amplitude reaches an audible level, but the delay ensures a stable and accurate pitch-detection process, which operates continuously. This delay is unnoticeable unless you run the ATR-1's output alongside the unprocessed input signal, in which case you hear a small amount of phasing. In practice, the delay is no problem.
However, the ATR-1 can process only certain types of signals. The input waveform must be periodic and free of excessive noise. Obviously, this limits the pitch correction to monophonic instruments and solo voices. Unison ensembles (such as a unison choir or violin section) cannot be processed because they have no clearly defined fundamental due to the minor pitch discrepancies within the ensemble. As long as you are dealing with signals like a lead vocal or solo sax, the ATR-1 can work wonders to correct intonation problems.
Some monophonic signals, such as breathy vocals, can cause pitch-tracking problems. Fortunately, the Sensitivity control helps in this regard; at the correct setting, this control allows even signals with quite a bit of bleed from other instruments to be pitch-corrected. For anything other than a clean, isolated track, the Sensitivity control is one of the crucial parameters to have at your fingertips.
Program Mode
The ATR-1 continuously tracks the incoming pitch of a live or recorded signal and compares it to the notes in a user-defined scale. If the signal matches a pitch in the scale, it is left alone. If there is a discrepancy, the pitch is shifted to the nearest scale tone. You can also specify certain notes in the scale to be bypassed, in which case nearby pitches in the input signal are not altered.
Scales are stored in Programs, of which there are 50 in the ATR-1's memory. The first 13 Programs are preset with the 12 major (and relative natural minor) scales and the chromatic scale, but you can save different scales in these locations if you want to. Unlike the plug-in versions of Auto-Tune, the ATR-1 can deal only with 12-tone equal-tempered scales; it has no microtonal capabilities.
Each Program shifts all pitches that are not in the selected scale, so "accidental" notes will be altered as they pass through the device (except notes that are set to be bypassed). I typically used the chromatic setting because most of the tracks I processed weren't strictly diatonic, and they had relatively good pitch to begin with. If you're working with a diatonic melody that stays in one or two keys, the corresponding Program might work well for you.
Another vital Program parameter is the Speed setting, which determines how quickly the pitch is adjusted toward the scale tone. Fast Speed settings work well on instruments that change pitch quickly without a lot of inflection, such as oboes. Slower settings allow more expressive pitch gestures, such as vibrato and slides. You can also program the ATR-1 to change the Speed in response to a user-selectable MIDI Control Change message.
Setting this parameter correctly is crucial. If it's too fast, any natural vibrato will be completely removed, making the part sound mechanical; if it's too slow, the pitch will not be corrected to the target note before the next note begins. A value in the range of 6 to 12 generally corrects the pitch without tampering with the vibrato and other gestures, leaving a natural-sounding and in-tune performance. When you find the right setting, the result is gorgeous. However, the normal tendency is to use too fast a setting in an attempt to maximize the pitch correction (more on this later).
The ATR-1 also lets you introduce artificial vibrato into the signal with control over the waveform, rate, depth, and onset delay; the rate and delay can even be controlled in real time with user-selectable MIDI Control Change messages. If you must squeeze the life out of a part for the sake of pitch, you can at least add some modulation back to the signal. But unless you're going for an unusual effect, this isn't something you would do routinely, because artificial vibrato never sounds as good as the real thing.
If you need substantial pitch correction but don't want to remove the character of the performance, a viable alternative is to process the real trouble spots of the track in a second pass, using fairly slow settings for both passes. In many cases, this will do the trick without ruining the overall performance and removing all the desired inflections. Of course, this works only for recording applications; you can't do two separate passes in a live performance.
Each Program has its own Scale, Speed, and Vibrato settings. You can step through the Programs with a footswitch or by using the MIDI Data Increment message. This mode is intended primarily for applications in which only one or two scales are required, as is the case with highly diatonic music.
Song Mode
In addition to Program mode, the ATR-1 provides a Song mode. Up to 20 Songs can be stored in memory, and each Song can include up to 15 Steps. Each Step contains a Program or one of several other navigation items (more in a moment), along with its own Speed and Vibrato settings, which override the associated Program's settings. This lets you use the same scale in different Songs without having to save multiple copies of the Program with different settings.
Song mode is designed for live situations, where you might need to step through several Programs in a performance. You can cycle through the Steps in a Song with a footswitch or MIDI Data Increment messages. (You can also step through the Songs in memory using the same techniques.)
As mentioned earlier, each Step can include one of several navigation items instead of a Program. For example, when you activate a Step set to Bypass, the signal passes through the ATR-1 without correction. If the active Step contains the Loop item, the Song immediately jumps back to Step 1; this is useful when you want to repeat a harmonic structure several times. (To exit the loop, you must select the next Song with the footswitch or a MIDI Data Increment message.) The Link item causes the ATR-1 to jump to the next Song in memory, which lets you create a sequence of more than 15 Steps.
A bit too perfect
My first ATR-1 experience was with a prototype loaned to me by Ted Perlman, who owned one of the five that existed at the time. (With the exception of a few parameter tweaks, the production model I later received for review is identical to the prototype.)
I was producing a song for Dionne Warwick's current album, and I was faced with a tight deadline to finish the master and hand it over to the mix engineer. Warwick and I had tracked her lead vocal and 12 background parts (triple-stacked four-part harmony) in one very late night at the end of a long string of recording and concert dates.
Seasoned professional that she is, Warwick got all her vocals down in one evening, but we did not have the opportunity to touch up any parts later with fresh ears. After sorting through the various background parts the next morning, I found some places in the tripled background vocals where the deviations between like parts were too wide for my taste. They weren't necessarily out of tune, but rather out of focus.
I didn't have time to get involved in a Pro Tools session for vocal editing and comping because I also had some keyboard overdubs to do and a plane to catch that evening. The vocals had to be right, but they also had to be dealt with quickly. Enter the ATR-1.
Working with David Crigger at his project studio in North Hollywood, California, I ran each vocal track through the ATR-1 and made pitch corrections while transferring from ADAT tapes into StudioVision for comping and editing. Unfortunately, there is no digital input or output on the ATR-1, so we had to run each track through several conversions: from the ADAT's D/A converters into the ATR-1's A/D and out of the ATR-1's D/A into another set of A/D converters on the way to a Power Macintosh 7100. Another drawback is that it has no I/O level controls. But the show had to go on.
A quick read through the manual got us up and running in about 15 minutes. Once we set the main parameters–Speed, Sensitivity, and Scale–we were in business. As mentioned earlier, my natural tendency was to overprocess the signal using a Speed setting that was too fast on the first couple of vocal tracks. These tracks had to be redone later in the session when it became apparent that the cumulative results were unnatural sounding.
We also noticed a low-level grounding hum polluting the ATR-1 output. The ground switch on the back panel did nothing to alleviate this problem, so we had to resort to using a two-prong ground-lift adapter. This eliminated the hum, and the ATR-1 output was virtually dead quiet. (The hum turned out to be a problem in the studio's electrical system, as subsequent experience demonstrated.)
Once we settled on a Speed setting that worked (8 to 10, depending on the track), it was easy to run a whole track through the ATR-1 and move on to the next take. If a section sounded fine, we just hit the Bypass button until we needed to use the pitch correction again. Ideally, I would have preferred to go bit by bit and process only certain sections of each track, varying the Speed parameter and diligently comparing it to other tracks. However, this would have taken many more hours, and pressing schedules demanded that we finish promptly.
I was dismayed that the ATR-1 has no digital I/O, but in spite of that, I was pleasantly surprised at the results. There was very little coloration in the sound from the ATR-1, although I could tell that each track had seen a few too many converters along the way for my liking. A minute amount of high-end noise and distortion was audible, but only when we mercilessly compared the processed sound to the original. Of course, no one listens to music this way, so the results were more than acceptable–they were brilliant! I was pleased, and the record company loved the track. End of story.
Well, almost. It's funny how pitch perception is relative. On the mix date, Warwick noticed that the background vocals had been polished, and she thought the tracks sounded a bit too "Pro Tool'ed." She asked the mix engineer to blend a little of the raw background tracks in with the pitch-edited parts, and then she was happy, too.
This brings up an interesting effect I've played with in the past: mixing a pitch-corrected signal with the raw signal can sometimes make a great doubling effect, especially with a little delay on one of the channels. It takes some experimentation to determine which channel (raw or pitch-corrected) should "lead" (higher volume, no delay), but the results are often stunning.
Happy holidays
While touring with Elvis Costello and Burt Bacharach in the month of October, I still had to complete a holiday CD I'd previously been hired to produce. This meant I was recording during off hours while on the road–not an ideal situation. As with the Warwick project, I had only one shot to record the vocals of three different guest singers. Then, I had only three days to completely mix ten songs in order to meet a mastering and manufacturing deadline.
All three singers gave excellent recorded performances, but under these conditions, just about any vocal track will have a few imperfections. However, recording several takes gave me some flexibility later.
I used the ATR-1 to fix a few spots in a vocal track, allowing me to go for the best performance. In other words, I didn't have to sacrifice a cool vocal passage with a few intonation problems for a more in-tune but less interesting phrase on another take. The ATR-1 let me get more mileage out of the limited amount of material I had, without being bogged down in a time-consuming process and missing the mix deadline. That, my friends, is the sign of a product that is truly productive.
The Moral of the Story
The ATR-1 is one of those products that can save your butt; in fact, it has saved mine twice. It is easy to use and extremely effective once you get a handle on its important parameters. It is very quiet, and it imparts few, if any, perceptible artifacts. Best of all, it can do its thing in real time without destroying the overall feel of the original performance, making it a live-performance tool as well as a studio lifesaver.
Of course, this box isn't perfect; nothing is. I wish it could handle stereo recordings of monophonic instruments and that it had digital I/O and analog level controls. In fact, digital I/O is one of the biggest arguments for the plug-in over the hardware version for serious professional applications.
The ATR-1 is less tweaky than the software plug-in, but it's less complicated to use, too. If a particular track has to be absolutely perfect and you have the time to scrutinize and experiment with every little phrase, then you should use the Auto-Tune plug-in.
In the end, it was the ATR-1's ease of use that won me over when the heat was on, even though I already had the software version. If my only option for vocal pitch correction had been to use a computer-based plug-in, I probably wouldn't have done anything at all; I simply couldn't afford the additional time it would have taken. As it was, my finished tracks were better because the ATR-1 let me clean up the best vocal performances with very little additional time spent. This is a great trade-off for the more extensive parameter control of the software version. If your need for pitch correction isn't strictly limited to a computer setup, the ATR-1 is the magic black box to have.
Auto-Tune (sidebar)
Antares Systems originally implemented its pitch-correction algorithm as a TDM plug-in called Auto-Tune ($599). This plug-in is currently available in a number of incarnations, including VST ($399), DirectX ($299), MAS ($399), and V8 ($599), as well as a stand-alone version for Macintosh that uses the computer's digital hardware ($399).
Auto-Tune operates in one of two basic modes: Automatic or Graphical. In Automatic mode, Auto-Tune functions in virtually the same way as the ATR-1. After a scale is specified, the Retune slider sets the speed of pitch correction, and the Tracking slider is similar to the ATR-1's Sensitivity setting. The Vibrato parameters are also the same as those in the ATR-1.
In Graphical mode, the Retune slider has the same effect as in Automatic mode–determining how quickly pitch correction is applied–but the target pitch is not a scale tone; it is determined graphically by the user. As the signal is played into Auto-Tune, a waveform display at the bottom shows the current audio section, whose pitch profile is displayed in the main window.
As a section of the waveform is highlighted, the Make Curve command creates a pitch curve that is based on the original performance. This curve can be globally dragged or reshaped as desired. Graphical mode allows you to maintain the desired amplitude and pitch fluctuations of a performance while adjusting the overall pitch center.
Graphical mode provides much more controlled and detailed pitch editing than Automatic mode, but it is potentially more laborious, and results vary with the skill of the user. Deciding which mode is best for a particular situation is a trial-and-error operation.
Antares ATR-1 intonation processor
PROS: High-quality pitch correction of monophonic sources. Simple to use. Preserves performance character and inflection while correcting pitch. Very little sonic coloration. Custom scales. Can be controlled via MIDI or footswitch.
CONS: No digital I/O. No analog I/O level controls. A bit expensive.
---
Composer and producer Rob Shrock is the musical director for Burt Bacharach. He has also worked with Elvis Costello, LeAnn Rimes, Luther Vandross, Chrissie Hynde, Stevie Wonder, and a host of others.