Antares AVP1 Hardware user manual

							5
Pitches are often described relative to one another as intervals, or ratios of
frequency. For example, two pitches are said to be one octave apart if
their frequencies differ by a factor of two. Pitch ratios are measured in
units called cents. There are 1200 cents per octave. For example, two tones
that are 2400 cents apart are two octaves apart. The traditional twelve-
tone Equal Tempered Scale that is used (or rather approximated) in 99.9%
of all Western tonal music consists of tones that are, by definition, 100
cents apart. This interval of 100 cents is called a semitone.
How Auto-Tune detects pitch
In order for Auto-Tune to automatically correct pitch, it must first detect
the pitch of the input sound. Calculating the pitch of a periodic waveform
is a straighforward process. Simply measure the time between repetitions
of the waveform. Divide this time into one, and you have the frequency in
Hertz. The AVP does exactly this: It looks for a periodically repeating
waveform and calculates the time interval between repetitions.
The pitch detection algorithm in the AVP is virtually instantaneous. It can
recognize the repetition in a periodic sound within a few cycles. This
usually occurs before the sound has sufficient amplitude to be heard. Used
in combination with a slight processing delay (no greater than 4 millisec-
onds), the output pitch can be detected and corrected without artifacts in
a seamless and continuous fashion.
The AVP was designed to detect and correct pitches up to the pitch C6. If
the input pitch is higher than C6, the AVP will often interpret the pitch an
octave lower. This is because it interprets a two cycle repetition as a one
cycle repetition. On the low end, the AVP will detect pitches as low as 42
Hz. This range of pitches allows intonation correction to be performed on
all vocals and almost all instruments.
Of course, the AVP will not detect pitch when the input waveform is not
periodic. As demonstrated above, the AVP will fail to tune up even a
unison violin section. But this can also occasionally be a problem with solo
voice and solo instruments as well. Consider, for example, an exceptionally
breathy voice, or a voice recorded in an unavoidably noisy environment.
The added signal is non-periodic, and the AVP will have difficulty deter-
mining the pitch of the composite (voice + noise) sound. Luckily, there is a
control (the Sensitivity control, discussed in Chapter 4) that will let the
AVP be a bit more casual about what it considers “periodic.” Experiment-
ing with this setting will often allow the AVP to track even noisy signals. 
						

							6
How Auto-Tune corrects pitch
Auto-Tune works by continuously tracking the pitch of an input sound and
comparing it to a user-defined scale. The scale tone closest to the input is
continuously identified. If the input pitch exactly matches the scale tone,
no correction is applied. If the input pitch varies from the desired scale
pitch, an output pitch is generated which is closer to the scale tone than
the input pitch. (The exact amount of correction is controlled by the Speed
parameter, described below and in Chapter 4.)
Scales
The heart of Auto-Tune pitch correction is the Scale. The AVP comes with
25 preprogrammed scales. For each Scale you can define which notes will
sound and which won’t. And for each note that will sound, you can decide
whether the AVP will apply pitch correction to input pitches near that
note or leave those pitches uncorrected.
You can also edit any of the preprogrammed scales and save your custom
scale as part of a Preset.
Speed
You also have control over how rapidly, in time, the pitch adjustment is
made toward the scale tone. This is set with the Speed control (see Chap-
ter 4 for more details).
•Fast Speed settings are more appropriate for short duration notes and
for mechanical instruments, like an oboe or clarinet, whose pitch
typically changes almost instantly. A fast enough setting will also
minimize or completely remove a vibrato. At the fastest setting, you
will produce the now-infamous “Cher effect.”
•Slow Speed settings, on the other hand, are appropriate for longer
notes where you want expressive pitch gestures (like vibrato) to come
through at the output and for vocal and instrumental styles that are
typified by gradual slides (portamento) between pitches. An appropri-
ately selected slow setting can leave a vibrato unmodified while the
average pitch is accurately adjusted to be in tune. 
						

							7
An example
As an example, consider this before-and-after graphic representation of
the pitch of a vocal phrase that contains both vibrato and expressive
gestures.
In the original performance, we can see that although the final note
should be centered around D, the vocalist allowed the tail of the note to
fall nearly three semitones flat. The “after” plot is the result of passing
this phrase through the AVP set to a D Major Scale (with C# and B set to
”Blank”) and a Speed setting of 10. That Speed causes the pitch center to
be moved to D, while still retaining the vibrato and expressive gestures.
(Setting C# and B to ”Blank” is necessary to keep the AVP from trying to
correct the seriously flat tail of the last note to those pitches. See Chapter
4 for more details.)
Antares Microphone Modeling
If you’ve spent any time lately flipping through the pages of pro audio
magazines, you have almost certainly noticed the intense focus on micro-
phones. From the proliferation of exotic new mics to the almost cult-like
following of certain historical classics, never has the choice been greater.
But amassing a substantial collection of high-end mics is financially pro-
hibitive for all but the most well-heeled studios.
Now, using our patented Spectral Shaping Tool™ technology, we’ve
created digital models of a variety of microphones. Simply tell the AVP
what type of microphone you are actually using and what type of micro-
phone you’d like it to sound like. It’s as simple as that.
10.0 10.5 11.0 D3
B2 C

3ORIGINAL
PERFORMANCE CORRECTED 
BY AVP 
						

							8
With the AVP, you can record each track through a model of the type of
mic that will best produce that ideal sound you’re looking for. Or use it in
live performance to get the sound of mics you’d never consider using on
stage. You can even use it during mixdown to effectively change the mic
on an already recorded track. And for that final touch of perfection, you
can even add some tasty tube saturation.
About the technology
The models employed by the AVP are not derived from theoretical consid-
erations. They are generated by a proprietary analysis process that is
applied to each physical mic modeled. Not only the sonic characteristics,
but the behavior of other parameters such as low-cut filters or proximity
effects accurately reflect the specific performance of each microphone we
model.
Another advantage of our model-based approach is that there is essen-
tially no processing delay apart from the natural phase effects of the
microphones being modeled.
Finally, the quality and signal-to-noise characteristics of the processing are
pristine. Because of our commitment to model-based processing, there are
none of the limitations or distortions characteristic of FFT-based algo-
rithms. The quality of the output is limited only by the quality of the
input.
So what exactly does it do?
While there is a lot of fairly complicated stuff going on under the hood,
the essential functionality of the AVP’s Mic Modeling module is really
quite simple. Basically, audio originally recorded by a microphone is input
to the AVP where it is first processed by a “Source Model” which serves to
neutralize the known characteristics of the input mic. The audio is then
processed by a second “Modeled Mic” model which imposes the character-
istics of the modeled mic onto the previously neutralized signal. Finally,
the audio is passed through a model of a high-quality tube preamp
offering the option of classic tube saturation distortion.
Understanding Compression
Compression is probably the most widely used (and potentially confusing)
signal process used in today’s studios. Simply put, compression reduces the
dynamic range of a signal. That is, it reduces the difference in loudness
between the loudest and quietest parts of a piece of music. Another way
to think about this is that the compressor is acting as an automatic fader
which fades down when the signal gets loud and fades back up when the
signal gets soft. 
						

							9
Why reduce the dynamic range? Consider the problem of mixing the vocal
in a contemporary rock or pop song. Typically, pop music has a relatively
consistent level of loudness. If an uncompressed vocal track is added to a
typical pop mix, loudly sung words or syllables would jump out of the mix,
while quieter phrases would be buried beneath the instrumental texture.
This is because the difference between the loudest and softest sounds in
the vocal - its dynamic range - is very large. This same problem occurs for
any instrument which has a dynamic range larger than the music bed into
which it is being mixed. (For that reason, most instruments, not just vocals,
undergo some compression in the typical mix.)
By using a compressor to decrease the dynamic range of the vocal, the
softer sounds are increased in loudness and the loudest sounds are re-
duced in loudness, tending to even out the overall level of the track. The
overall level of the compressed track can then be increased (using what is
referred to as “make-up gain”), making the vocal track louder and more
consistent in level, and therefore easier to hear in the mix.
Threshold and Ratio
How is compression measured? What is a little compression and what is a
lot of compression?
The effect a compressor has on a track is determined by the settings of its
threshold and ratio. The threshold is the level above which the signal is
attenuated. The ratio is the measure of how much the dynamic range is
compressed.
The graph shown below shows the relationship between the input level of
a signal and the output level of the signal after compression. Notice that
signals that are louder than the threshold are compressed (reduced in
level) while those softer than the threshold are unchanged.
As the input signal exceeds the threshold, gain reduction (reduction in
loudness) is applied. The amount of gain reduction that is applied depends
on the compression ratio. The higher the compression ratio, the more gain
reduction is applied to the signal.
The graph shows the relationship between compression ratio and gain
reduction. Examine the 2 to 1 ratio curve. For signals above the threshold,
this setting transforms a range of loudness 2 units large into a range of
loudness one unit large (i.e., if the input signal gets “x” units louder, the
compressed signal increases by only “x/2” units). 
						

							10
Limiting
Examine the 99:1 curve in the above graph. This setting reduces all sounds
above the threshold to the same loudness. This is called limiting. Limiting
is usually employed to allow a dynamic signal to be recorded at a maxi-
mum level with no risk that transient peaks will result in overload. In this
application, the threshold setting (usually set relatively high) determines
the extent to which the peaks will be limited.
Dynamic Expansion and Gating
Sometimes, it is desirable to increase the difference between the quietest
signal and the noise in a recording by using a downward expander. A
typical application would be eliminating room noises and breath sounds
that can be heard between the phrases of a recorded vocal part.
The graph below shows the curveÉ
						

							11
When expanders use ratios higher than 1:10, sounds below the threshold
are faded out very rapidly. This effect is called gating and can sound very
abrupt. Adjusting the gate ratio can smooth out the abrupt change. The
graph below shows the input/output curve for a typical gate.
OUTPUT
LEVEL
INPUT LEVELLOUDER
LOUDER
THRESHOLD
1 TO 2 EXPANSION RATIO
1 TO 1 RATIO
OUTPUT
LEVEL
INPUT LEVELLOUDER
LOUDER
THRESHOLD
1 TO 99 EXPANSION RATIO
1 TO 1 RATIO 
						

							12
Sounds that are louder than the threshold get “through the gate” un-
changed. Sounds that are below the threshold are not heard. Gates can be
used to great effect in processing drum tracks where sounds from the
other instruments in the drum set leak through the mike of the instru-
ment being recorded. Gates are also used frequently to “gate off” a
reverb tail or the ringing from an insufficiently damped drum head.
Compression and Expansion Combined
The AVP allows you to use both compression and expansion simulta-
neously. This ability is useful in taming the typical problems that arise
when processing vocal tracks. The graph below illustrates the use of
compression with a downward expanding gate.
Using this setting, levels above the compressor threshold will be com-
pressed at a 4 to 1 ratio. Levels below the compressor threshold but above
the gate threshold will not be changed. Levels below the gate threshold
will be gated out completely.
Used on a vocal track, this setting will compress only hot peaks in the
voice, while gating out the room sounds, mike stand sounds, and breath
noises in the track. Precisely what gets compressed and gated is a function
of the compressor and gate threshold settings.
The graph below shows a dynamic expander. In this application, the gate
threshold and ratio are set to gently expand the program material at a 1.5
to 1 ratio. The compressor ratio is set to 1 to 1. The setting is useful for
repairing over-compressed material or for adding some punch to drums or
other percussive sounds.
OUTPUT
LEVEL
INPUT LEVELLOUDER
 GATE THRESHOLD
 COMPRESSOR THRESHOLD
1 TO 99 EXPANSION RATIO
4 TO 1 RATIO 
						

							13
Hard Knee/Soft Knee
The graphs shown above have what are described as “hard knees” in their
gain curves. This means that as the signal passes through the threshold,
the gain reduction it receives will begin abruptly. In settings where the
compression or expansion ratios have high values, the abrupt change can
be heard and often sounds artificial.
To make it possible to create settings where the dynamic effects are more
natural sounding, the AVP incorporates a Knee control which allows you
to soften the transition between sections of the gain curve. The graph
below shows a curve which has “soft knees,”making the dynamic transi-
tions more subtle.
OUTPUT
LEVEL
INPUT LEVELLOUDER
LOUDER
GATE THRESHOLD
1 TO 5 EXPANSION RATIO
COMPRESSOR
THRESHOLD
OUTPUT
LEVEL
INPUT LEVEL
SOFT KNEES
KNEE = 100  COMPRESSOR THRESHOLD
 GATE THRESHOLD 
						

							14
Attack and Release Times
The attack time of a compressor is how long it takes for the compressor to
react once the input level has met or exceeded the threshold level. With a
fast attack time, the signal is brought under control almost immediately,
whereas a slower attack time will allow the start of a transient or a
percussive sound to pass through uncompressed before the processor
begins to react.
For sounds without percussive attacks (voices, synth pads, etc.), a fairly
short attack time is usually used to ensure even compression. For instru-
ments with percussive attacks (drums and guitars, for example), a slower
attack time is typically used to preserve the attack transients and, hence,
the characteristic nature of the instruments.
The illustration below shows the effect of various the attack times.
The release time of a compressor is the time it takes for the gain to return
to normal after the input level drops below the threshold. A fast release
time is used on rapidly varying signals to avoid affecting subsequent
transients. However, setting too quick a release time can cause undesirable
artifacts with some signals. On the other hand, while slower release times
can give a smoother effect, if the release time is too long, the compressor
will not accurately track level changes in the input. Slow release times may
also result in audible level changes known as “pumping.”
UNCOMPRESSED INPUT COMPRESSED
1 mSEC ATTACKCOMPRESSED
10 mSEC ATTACK