 |
 |
 |
|
The Oxford Limiter has been developed
from decades of professional audio experience to provide a very high
degree of quality and facility in programme loudness control and
limiting functions. By employing highly accurate logarithmic side chain
processing, along with innovative adaptive timing functionality using
look ahead signal acquisition, the limiter provides exemplary
performance, whether one is seeking general transparent level control,
programme loudness maximisation or heavily applied artistic sound
effects.
Unique processing in the form of the Enhance function provides the
sample value limiting needed to reliably avoid overloads in digital
workstation environments and allows unprecedented volume and punch to
be applied to programme beyond that available from conventional
limiting functions.
Comprehensive metering is provided which displays not only conventional
peak sample value, but additionally allows the user to monitor the true
validity of the programme in order to avoid the generation of damaging
reconstruction overloads in the target equipment which are often
invisible during production (sometimes termed 'inter sample peaks').
A further function allows the user to dynamically correct for
reconstruction overloads in real time, thereby achieving maximum
possible modulation levels without the risks of producing illegal
signals often associated with compression and limiting.
Comprehensive dithering functionality with selectable and variable
depth noise shaping ensures first class mastering output quality in
either 24 bit or 16 bit modes.
Features
Unique Enhance section - provides sample value limiting and enhances perceived loudness
Ability to add harmonic distortion, where artistic style requires
Variable soft knee control
Attack and Release controls
Reconstruction Meter shows 'actual' signal rather than sampled signal
Auto compensate feature - corrects recon errors without losing overall programme loudness
16 and 24 bit dithering with 5 selectable types
User has control over variable depth noise shaping
16 presets - some quite extreme !
Platforms Supported
Pro Tools HD (TDM), Pro Tools LE (RTAS), M-Powered (RTAS), PowerCore, AU, VST
General
The limiter plug-in comprises of four cascaded processes in the order below.
Peak programme limiting section.
Programme enhancement section.
Reconstruction metering and compensation section.
Dithering and noise shaping section.
Pre-process Section
The Pre-process section is a musical programme levelling function. Its
primary purpose is to control programme level over a wide range in
order to provide optimum conditions for the following enhancement
stage. When the enhancement is disabled in normal mode (at 0% with Safe
mode deselected) the pre-process section can be used as a conventional
levelling section in it ’s own right. Enhancement Section
The purpose of the enhancement process is to provide sample value
limiting and overall programme loudness improvement. The process
follows the pre-process section in the signal path and is controlled by
a separate fader from 0% (no action) to 125% (maximum action). In
normal mode the range from 0% to 100% fades in the effect to full
level, at which complete sample value limiting occurs. Settings from
100% to 125% further modify the process to progressively increase
loudness and programme density at the expense of increasing potential
distortion artefacts.
Safe mode is provided to allow absolute peak level control without
excessive enhancer action, even when using slow attack settings. In
Safe mode the enhance process is set to run permanently and the enhance
slider modifies the action of the process (rather than the proportion
of the effect). Setting ranges from 0% to 100% control the degree of
programme loudness boost generated by the enhancer. The control region
from 100% to 125% works identically to normal mode. It should be noted
that in safe mode signals at all levels are being processed
permanently, therefore some minor changes to the programme dynamics can
occur even for a minimum setting of 0%.
The enhance process improves the perceived loudness and presence of the
programme by modifying the dynamic and harmonic content of the signal.
Since the method used is different from the pre-processing section, it
can further enhance the perceived volume of a previously processed
signal whilst suppressing all signal overloads. As the limiting action
does not involve conventional sample value clipping, harsh distortions
are avoided and programme detail and dynamic information is largely
retained. Attack timing
The addition of an attack timing control is a significant departure from conventional limiter applications.
The attack control provides a means of increasing the attack time to
achieve a favourable improvement in the sonic qualities of the peak
reduction process by allowing peak programme transient events to escape
hard gain reduction. Since the plug-in has internal headroom these
overshoot peaks are retained and not clipped. Meter operation
When the Recon Meter
is selected the meter is switched from conventional peak sample value
mode into reconstruction mode. In this mode peak reconstruction levels
will be displayed on the meter. Levels in the red overload range of the
meter represent the presence of potential reconstruction errors. Dither and Noise Shaping
The finite mathematical precision provided by digital audio systems and the
effects of dither have been a source of confusion in the audio community for
some years. When dither is applied, any signal related error caused by
finite word length limitation is turned into constant random noise with no
relation to the signal itself, so such dithering provides complete removal
of harmonic distortion due to precision limits. With the Oxford Limiter, we
provide conventional 16 and 24 bit TPDF dither options.
In addition, the Oxford Limiter also provides several types of Noise Shaping
dither. Noise shaped dithering is a mechanism that aims to reduce the
perceived loudness of the noise of a dithered signal by either forcing the
spectrum of the noise out of the audible range or placing it into frequency
ranges to which we are less sensitive. In this way the noise at very low
levels may be reduced and even lost entirely if it is at the limit of our
hearing within ambient noise conditions.
System Requirements
Pro Tools
Pro Tools 6.0 or above
Approved Pro Tools CPU, OS and hardware configuration. More details: www.digidesign.com
Mac OSX 10.3 or later
(Leopard supported see Digidesign for details)
Windows XP
/ Vista32RAM 1GB minimum
iLok key with latest drivers
Audio Units
Audio Units compatible application (Logic, Digital Performer etc.)
Mac OSX 10.4 or later
(including Leopard)
RAM 1GB minimum
iLok key with latest drivers
VST
VST compatible application (Cubase, Nuendo, Acid etc.)
Mac OSX 10.4 or later
(including Leopard)
Windows XP
/ Vista32
RAM 1GB minimum
iLok key with latest drivers
PowerCore
TC PowerCore hardware
PowerCore version 3 drivers
AU or VST host application
Mac OSX 10.4 or later (including Leopard)
Windows XP / Vista32
RAM 1GB minimum
iLok key with latest drivers
|
roduce signals that are considerably in excess of maximum modulation.
Since the vast majority of metering within workstation environments
responds to sample value only, the above example would show a level of
around -3dB below clipping. However any further increase in the level
of the signal would result in a potentially illegal output level from
the system converter. As this error would not be reported on metering
within the workstation, in this particular case a possible 3dB overload
can result if the signal is increased to a maximum reading on the
workstation meters. This phenomenon is sometimes termed 'inter-sample
peaking'.
Although the above example is somewhat extreme and specific, there is
plenty of potential for this to occur within the mixing environment.
Combining a number of processed contributing tracks and limiting the
result to the maximum possible modulation level in order to satisfy
current industry trends, using only peak value metering, provides a
recipe for such hidden errors.
Since the very purpose of a limiting application is most often to
increase average modulation levels, a reconstruction meter with both
manual and automatic correction processing has been included in the
Sonnox Oxford Limiter plug-in, in order to provide the user with a
method to avoid or repair such errors. Back to Top
Meter operation
When the Recon Meter
is selected the meter is switched from conventional peak sample value
mode into reconstruction mode. In this mode peak reconstruction levels
will be displayed on the meter. Levels in the red overload range of the
meter represent the presence of potential reconstruction errors, as
illustrated below using the previous example.
Two methods are provided to correct for this.
Since the output level fader precedes the metering, errors may be
corrected manually by simply reducing the output level setting by the
same amount as the maximum error level reported on the meter. Auto Comp
Under normal circumstances errors are interspersed throughout the programme often restricted to certain specific events. The Auto
Comp
function is provided to address the situation where it may be
undesirable to reduce the level of the whole programme to avoid
transient errors. When Auto
Comp is
selected the level of the output is automatically controlled to repair
reconstruction errors by the minimum amount required and only for the
duration of the error. In this way the loudness of parts of the
programme unaffected by the errors remains as high as possible.
A combination of Auto Comp
and manual output level reduction can be used to strike a compromise,
if the action of the error correction becomes intrusive in the presence
of very large and intermittent error conditions.
Dither and Noise Shaping
The finite mathematical precision provided by digital audio systems and
the effects of dither have been a source of confusion in the audio
community for some years. Such discussion may lead to possible
misconceptions, which could prevent the user from achieving maximum
performance from systems in use. Therefore the dithering options
provided in the Sonnox Oxford Limiter warrant some prior explanation. Conventional Dither In both 24bit and 16bit output word
length selections, high pass TPDF (triangular probability density
function) dithering is applied to the output of the plug-in. Since any
signal related error caused by finite word length limitation is turned
into constant random noise with no relation to the signal itself, such
dithering provides complete removal of harmonic distortion due to
precision limits, which are an inescapable result of any numerical
signal representation. Dithering also suppresses any possibility that
the programme will suffer loss of harmonic signal resolution due to
word length restriction. The following plots illustrate this in action.
The next plot shows the exact same signal and truncation to 16 bits but with the HP TPDF dither applied.
It can be seen that all harmonic errors have been removed. Also since
the FFT analysis method provides an enhanced view of the signal below
the noise floor, it can also be seen that there is effectively no low
level floor below which a signal will fail to pass. To illustrate this
fact the following plot shows a 1KHz signal at -120dBr passing through
a dithered 16bit system. This corresponds to a signal 24dB below the
level of the least significant bit, the effective channel SNR is added
in blue for illustration purposes.
This shows that dither turns a quantised numerical signal conduit into
the equivalent of a naturally continuous (un-quantised) system, which
exhibits a finite signal to noise ratio with no practical limit to
harmonic signal resolution. In other words the inescapable presence of
quantisation in numerical systems does not forcibly lead to
'discontinuity' or 'resolution loss' in the signal. Misunderstandings
of this fact underpin many of the most damaging misconceptions
surrounding digital audio systems. It can also be deduced from the
above plots that any undithered digital representation of an audio
signal is effectively illegal.
Noise Shaping dither
If for some reason Signal to Noise Ratio (SNR) figures of 93dB at 16bits (or
143dB at 24bits) prove insufficient, noise shaping can provide an apparent increase
in SNR, but there are some potentially hidden costs. Noise shaped dithering is
a mechanism that aims to reduce the perceived loudness of the noise of a dithered
signal by either forcing the spectrum of the noise out of the audible range or
placing it into frequency ranges to which we are less sensitive. In this way
the noise at very low levels may be reduced and even lost entirely if they are
at the limit of our hearing within ambient noise conditions. The following plot
illustrates this process.
The red line shows the original 16bit dithered output with the -120dBr
signal passing. The blue line shows the effect of noise shaping (type 1
at 100%) on the same signal transfer.
It can be seen that the noise has been substantially reduced in the
regions up to around 8KHz where we are most sensitive, at the expense
of extra noise energy in the higher ranges above 10KHz at which we are
less sensitive. Such processing can have a dramatic effect on the
perceived intrusion of low-level noise.
However as is always the case, one cannot get something for nothing and
it can be seen from the above plot that the total noise power across
the whole range must remain constant to satisfy the dithering
requirement. This means the noise level necessarily increases in some
ranges of the spectrum. A level increase anywhere in the spectrum must
be accommodated by an increase in total peak noise level. The following
plot illustrates this in action with the previous test conditions.
The first plot shows a sample value plot of the conventional TPDF
dithered signal. The second shows the increase in values caused by the
application of the type1 noise shaping at 100%. Focussing the dither
energy into a more restricted range than would naturally occur causes
the level to increase.
From the effective level values highlighted in the plots it can be seen
that the application of noise shaping has increased the effective noise
level from around -93dBr to -80dBr, an increase of roughly 13dB.
From this it can be understood that the design of suitable noise
shaping frequency curve is a trade off between the perceived loudness
of the noise under certain conditions and the increase in overall level
of the dither signal, much of this trade off relies on what we can hear
(psycho-acoustics). Significant research has been carried out over the
years into various approaches to this issue and several accepted curves
are in use around the industry.
The Sonnox Oxford Limiter includes 4 noise-shaping curves. Types 1 and
3 are fifth order and types 2 and 4 are third order designs
representing a varied set of trade-offs to suit most programme types,
as illustrated below.
Whilst it is understood that the selection of noise shaping type is
largely a matter of user preference, generally speaking types 1 and 2
produce the most dramatic reduction in overall noise loudness, with
type 1 being the most effective of all. Types 3 and 4 provide gentler
responses, which under some circumstances may produce less intrusive
sounding spectrums, at the expense of higher audible residual noise.
Type 3 also provides greater noise attenuation in the range between
10KHz and 16KHz at the expense of higher noise levels in the mid
ranges.
Noise shaping Depth control From the previous section it can
be seen that noise shaping can potentially cause unwanted effects in
equipment and processes down line, particularly if the programme is to
be further modified, such as in mastering situations. Some unwanted
effects may include:
Marked increase in noise levels if the file is not transferred intact bit for
bit, (i.e. if further processing is done).
Premature meter readings in silence.
Premature peak level over loads (as increased dither levels add to peak signal
value).
Unwanted low-level behaviour in dynamics processing.
Disturbance caused to data reduction encoding processes such as MP3, WMA etc.
Increased audibility (unmasking) of various errors that may occur in play-out
systems.
Generally speaking high levels of noise shaping renders a signal that is more
fragile. Almost any change to the produced audio file after noise shaping could
potentially result in unwanted effects.
For these reasons the depth control is
provided to put you in charge of the degree to which noise shaping is applied.
When any of the noise shape curves are selected, the depth control
varies the degree of noise shaping from 0% to 100%. At 0% the dither is conventional
HP TPDF dither (as if noise shaping were not selected), at 100% full noise shaping
is applied. All control positions within the range produce legal proportions
of dither.
The action of the depth control is illustrated below with type 1 selected.
There is no technical (or philosophical) advantage to noise shaping
above and beyond that which can be actually heard directly. Therefore
the decision to use noise shaping (and to what extent) is basically
determined by what might actually be heard in practice. If conventional
TPDF dither provides sufficient audible dynamic range such that noise
never intrudes within the programme, it is safest to avoid
noise-shaping altogether.
Because of the potential fragility of a noise shaped signal, it is better to ensure that it is carried out only at the final
stages
of mastering, immediately prior to release. If your mix is not already
a final master, it is technically preferable to send a TPDF dithered
24bit file to mastering rather than a 16bit noise shaped file.
Another important factor is that the effectiveness of psycho-acoustic
noise shaping relies heavily on our sensitivity to noise spectra at the
threshold
of hearing.
Therefore if noise shaped dither actually gets to be heard directly it
will often sound quite strange and intrusive and may detract from the
listener's experience of the programme. Therefore the most effective
and safest approach is to apply noise shaping at the minimum amount
necessary to render the noise inaudible within the conditions the
programme is destined to be auditioned.
|
|
|
 |
|
 |
|
|
|
