NoNOISE II — User Manual

NoNOISE II — User Manual
High Resolution Audio Restoration & Enhancement Tool
for soundBlade
©2006-2012 Sonic Studio, LLC — All rights reserved
This manual, as well as the software described in it, is furnished under license and may only be used or copied in accordance with the terms
of such license. The information in this manual is furnished for informational use only, is subject to change without notice, and should not be
construed as a commitment by Sonic Studio, LLC. Sonic Studio, LLC assumes no responsibility or liability for any errors or inaccuracies that may
appear in this document.
Except as permitted by such license, no part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or
by any means, electronic, mechanical, recording, or otherwise, without the prior written permission of Sonic Studio, LLC.
SONIC STUDIO, LLC MAKES NO WARRANTIES, EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, REGARDING THE APPLE SOFTWARE. SONIC STUDIO, LLC DOES NOT WARRANT,
GUARANTEE, OR MAKE ANY REPRESENTATIONS REGARDING THE USE OR THE RESULTS OF THE USE OF THE SONIC STUDIO, LLC SOFTWARE
IN TERMS OF ITS CORRECTNESS, ACCURACY, RELIABILITY, CURRENTNESS, OR OTHERWISE. THE ENTIRE RISK AS TO THE RESULTS AND PERFORMANCE OF THE SONIC STUDIO SOFTWARE IS ASSUMED BY YOU. THE EXCLUSION OF IMPLIED WARRANTIES IS NOT PERMITTED BY
SOME STATES. THE ABOVE EXCLUSION MAY NOT APPLY TO YOU.
IN NO EVENT WILL SONIC STUDIO, LLC, ITS DIRECTORS, OFFICERS, EMPLOYEES, OR AGENTS BE LIABLE TO YOU FOR ANY CONSEQUENTIAL,
INCIDENTAL, OR INDIRECT DAMAGES (INCLUDING DAMAGES FOR LOSS OF BUSINESS PROFITS, BUSINESS INTERRUPTION, LOSS OF BUSINESS INFORMATION, AND THE LIKE) ARISING OUT OF THE USE OR INABILITY TO USE THE SONIC STUDIO SOFTWARE EVEN IF SONIC STUDIO
HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. BECAUSE SOME STATES DO NOT ALLOW THE EXCLUSION OR LIMITATION OF
LIABILITY FOR CONSEQUENTIAL OR INCIDENTAL DAMAGES, THE ABOVE LIMITATIONS MAY NOT APPLY TO YOU.
Sonic Studio, Sonic Clarity, SSE, soundBlade and the Sonic Studio logo are trademarks of Sonic Studio, LLC. All other company or product
names are either trademarks or registered trademarks of their respective owners.
reNOVAtor Legal Notice
This document may not, in whole or in part, be copied, photocopied, reproduced, translated, transmitted, or recorded to other media without
prior written consent of Algorithmix® GmbH. The information in this document is subject to change without notice.
http://www.algorithmix.com
Algorithmix® End User License Agreement
This End User License Agreement (“License”) is a legal agreement between you (either an individual person or a single legal entity) and Algorithmix GmbH (“Algorithmix®”) regarding Algorithmix® software and the accompanying documentation (collectively, the “Software”). The
Software also includes any software updates, add-on components, web services and/or supplements that Algorithmix may provide to you or
make available to you after the date you obtain your initial copy of the Software to the extent that such items are not accompanied by a separate
license agreement or terms of use.
PLEASE READ THIS LICENCE CAREFULLY BEFORE USING THE SOFTWARE. BY INSTALLING, COPYING, DOWNLOADING, ACCESSING OR OTHERWISE USING THE SOFTWARE, YOU AGREE TO BE BOUND BY THE TERMS OF THIS LICENSE. IF YOU DO NOT ACCEPT OR UNDERSTAND THE
TERMS OF THIS LICENSE, YOU SHOULD NOT INSTALL, ACCESS OR USE THE SOFTWARE.
1.
LICENCE GRANT: Subject to the restrictions set forth below, this License grants you a non-exclusive, perpetual license to (a) install the
Software on only one computer; (b) use or authorize use of the Software on only one computer; (c) make one copy of the Software, in machinereadable form, solely for backup purposes; provided you include all copyright and other proprietary rights notices on the copy and (d) transfer
the Software and this License to another party if the other party agrees to accept the terms and conditions of this License. If you transfer the
Software, you must at the same time either transfer all copies to the same party or destroy any copies not transferred. If you transfer possession
of any copy of the Software to another party, your License is automatically terminated.
2.
LICENCE RESTRICTIONS: You may not, or allow any third party to, (a) rent, lease, sell, loan or otherwise transfer the Software or any of
your rights and obligations under this License; (b) install the Software on a network for use by multiple users, unless each user has purchased
a license; c) reverse-engineer, de-compile or disassemble the Software or hardware in whole or in part; (d) remove or destroy any copyright
notices or other proprietary markings of the Software or any third-party software; (e) modify or adapt the Software, merge the Software into
another program or create derivative works based on the Software; (f) make copies of or distribute, for profit or otherwise, the Software, except
as expressly provided above; (g) make any alteration, modification, connection, disconnection, improvement or adjustment of any kind to or
use the Software except as explicitly contemplated in the enclosed documentation and in this License and (h) sublicense, transfer or assign this
License or any of the rights and obligations granted under this License, except as explicitly contemplated in this License. Any purported transfer
or assignment will be void.
3.
EXPORT RESTRICTIONS: The Algorithmix® Software may not be exported or re-exported to any country to which the European Union
embargoes goods. By downloading or using Algorithmix® Software you are certifying that you are not a national of any country to which the
European Union embargoes goods.
4.
OWNERSHIP: The license granted hereunder does not constitute a transfer or sale of ownership rights in or to the Software. Except for
the license rights granted above, Algorithmix® retains all right title and interest in and to the Software including all intellectual property rights
therein. The Software is protected by applicable intellectual property laws, including European Union copyright laws and international treaties.
Page 2
5.
TERM AND TERMINATION: This License shall be effective upon installation of the Software and shall terminate upon the earlier of (a)
your failure to comply with any term of this License; (b) return, destruction or deletion of all copies of the Software in your possession or (c)
transfer of the Software and this License to another party in accordance with Section 1(d). Algorithmix® rights and your obligations shall survive
termination of this License.
6.
LIMITED WARRANTY: Algorithmix® warrants to the original licensee that the Software, as delivered, will perform in accordance with
the accompanying documentation for a period of 30 days from the date of the original purchase (“Limited Warranty”). Algorithmix® entire liability and your exclusive remedy for breach of the preceding Limited Warranty shall be, at Algorithmix® sole option, repair or replacement of
the Software that does not meet the warranty set forth herein and that is returned to Algorithmix®. This limited warranty shall be void if failure
of the Software has resulted from any accident, abuse, misuse or misapplication by you. Any replacement Software will be warranted for the
remainder of the original warranty period or 30 days, whichever is longer.
7.
NO OTHER WARRANTIES: EXCEPT AS SET FORTH ABOVE, THE SOFTWARE IS PROVIDED ON AN “AS IS” BASIS. YOU ASSUME ALL
RESPONSIBILITIES FOR THE QUALITY AND PERFORMANCE OF THE SOFTWARE. TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW,
ALGORITHMIX® DISCLAIMS ALL WARRANTIES, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF
MERCHANTABILITY, NONINFRINGEMENT OF THIRD-PARTY INTELLECTUAL PROPERTY, AND FITNESS FOR A PARTICULAR PURPOSE. ALGORITHMIX® DOES NOT WARRANT THAT THE FUNCTIONS CONTAINED IN THE SOFTWARE WILL MEET YOUR REQUIREMENTS, BE UNINTERRUPTED OR ERROR FREE.
8.
LIMITATION OF LIABILITY: YOU AGREE THAT IN NO EVENT WILL ALGORITHMIX® BE LIABLE FOR ANY CONSEQUENTIAL, SPECIAL,
INDIRECT, INCIDENTAL OR PUNITIVE DAMAGES WHATSOEVER (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR LOSS OF PROFITS, BUSINESS INTERRUPTION, LOSS OF INFORMATION), EVEN IF ALGORITHMIX® HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. IN
NO EVENT SHALL ALGORITHMIX® LIABILITY EXCEED THE TOTAL AMOUNT OF THE FEES PAID FOR THE SOFTWARE. Because some jurisdictions do not allow the limitation or exclusion of incidental or consequential damages in certain circumstances, the above limitations may not
apply to you.
9.
GENERAL: Any claim arising under or relating to this License shall be governed by the laws of the Federal Republic of Germany, without
reference to conflict of laws principles. This License constitutes the entire agreement between you and Algorithmix® and supersedes any other
communication with respect to the Software. No modification of or amendment to this License will be effective unless in writing signed by both
parties. If any provision of this License is held invalid or unenforceable, the remainder of this License will continue in full force and effect.
reNOVAtor is made by Algorithmix® GmbH, Germany. Copyright ©1998-2012 by Algorithmix® GmbH. All rights reserved.
reNOVAtor for soundBlade copyright ©2006-2012 by Sonic Studio, LLC. All rights reserved.
Page 3
NoNOISE II SR User Manual Table of Contents
Chapter 1
NoNOISE II SR...................................................................................... 5
NoNOISE II Options Installation —..................................................................................................6
1.1
Manual DeClick.......................................................................................................7
1.1.1
Interpolation Algorithms............................................................................................................. 7
1.1.2
Using Manual DeClick................................................................................................................12
1.1.3
Obscenity Reduction..................................................................................................................12
1.2
Broadband DeNoise.............................................................................................13
1.2.1
Introduction................................................................................................................................13
1.2.2
The Noise Estimate....................................................................................................................13
1.2.3
Running the Broadband DeNoiser............................................................................................16
1.3
reNOVAtor............................................................................................................20
1.3.1
Main Features............................................................................................................................ 20
1.3.2
Typical Applications.................................................................................................................. 21
1.3.3
Getting Started.......................................................................................................................... 21
1.3.4
Quick Reference......................................................................................................................... 24
1.3.5
User Interface Details................................................................................................................ 27
1.3.6
ReSynthesis Parameters........................................................................................................... 29
1.3.7
Spectrogram Parameters.......................................................................................................... 30
1.3.8
Interpolation Alternatives......................................................................................................... 31
1.3.9
Application Tips......................................................................................................................... 33
1.3.10
reNOVAtor Key Equivalents..................................................................................................... 34
NoNOISE II SR User Manual Index......................................................................... 37
Page 4
Chapter 1......................................................... NoNOISE II SR
NoNOISE II is the world’s premier tool for restoring vintage and problematic audio recordings.
NoNOISE II SR includes the entire NoNOISE II toolkit, including Manual Declick, complete
Broadband DeNoise, Production DeCrackle FFT (Fast Fourier Transform) generation and full
Sonic EQ. Plus, NoNOISE II SR adds the full implementation of Algorithmix’s reNOVAtor.
The Manual DeClick toolset precisely removes impulse noise, analog or digital overloads,
harmonic and intermodulation distortion, and provides a unique solution to obscenity masking.
The Broadband DeNoise option is the highest fidelity single–ended noise reducer available today
and can rescue both old and new recordings alike.
Production DeCrackle is the background or faster than real time version of theType E interpolator.
The DeCrackler, both manual and background versions, provide unparalleled distortion reduction
for a segment or an entire sound file.
Sonic EQ is a minimal phase EQ toolkit, providing seventeen different topologies, and is
suitable for both repair, enhancement and correction, while Algorithmix’s reNOVAtor is the
high performance spectral repair tool. Unlike work–alike products, reNOVAtor is able to dig into
recordings with very low signal–to–noise ratios and pull out usable program. Plus, the ability to
cut and paste time and spectral patches and its harmonic noise reducer functions all add to its
amazing usefulness.
Page 5
Figure 1.1: The NoNOISE II menu
NoNOISE II Options Installation —
NoNOISE II requires the installation of both an iLok license and Sonic Studio Options Files for
NoNOISE II FX, NoNOISE II FR or NoNOISE II SR to be enabled for use. Running the appropriate Options Installer, located in your soundBlade app folder’s Sonic Options folder, installs the
necessary software.
This does not install the actual option software, it merely enables specific options you have licenses for on your iLok.
Page 6
NoNOISE II Options Notes:.
1) Options are installed on the Boot Volume at:
/Library/Application Support/Sonic Studio/Sonic Options 2.0/
2) Use the Remove NoNOISE II Options utility for removing options.
NOTE: Only install NoNOISE II options that you have licenses on your iLok for. Installing options that are not on your iLok will make soundBlade launch very slowly.
1.1
Manual DeClick
The Manual DeClick option provides tools for isolating and removing individual transient
impairments in a sound file. Manual DeClick assists in removing unwanted noises such as clicks,
pops and thumps. It offers five different interpolation algorithms that are capable of correcting
even difficult audio anomalies.
The algorithms analyze audio on either side of the anomaly and, based on this information,
synthesizes replacement samples. Manual DeClicking substitutes the repaired samples for the
original program material.
Note that Manual DeClick processes generate two new files for each sound file
that is repaired. These “.cd” and “.rl” files contain the samples removed during repair and a list of their locations, respectively. If you move or delete either of these
files, you will not be able to restore the original samples, undoing the repair.
With the exception of the Type E interplator, Manual DeClick is generally used on very short
duration regions of 14 msec. or less. Though the algorithms are capable of credible repairs
over longer durations, only experience using each algorithms will allow you to judge sensible
parameters.
1.1.1
Interpolation Algorithms
There are several interpolators that are available in soundBlade. Each is suited to a particular
type of audio problem and context.
1.1.1.1
The B Type General Interpolator
The Type B interpolator is the general purpose algorithm. The majority of declicking situations
can be handled by simply choosing this option.
Page 7
The default Type B interpolator examines the audio on either side of the selection to determine
the context for resynthesizing audio to fill the gap. For the B Type, there are two additional
variations of the command that bias the context in a particular “direction,” ignoring the material
before or after the impairment.
If, for example, a click occurs just after a percussive event, the default Type B interpolator would
include part of the percussive event in its resynthesis, producing a unconvincing repair. The
B - General R-> L option would ignore the audio to the left, using only the samples after the
selection is build the repair. Conversely, the B - General L-> R option uses only samples before
the selection to resynthesize a repair.
Here are three examples:
Figure 1.2: The original material
Figure 1.3: A “repair,” using the default B Type algorithm
Page 8
Figure 1.4: A better “repair,” using the B - General L-> R option
Though the above examples are extreme, a 60 msec. selection to clearly show the result, they
should illustrate the concept of using the left–oriented or right–oriented B Type option when
needed.
1.1.1.2
The A & C Type Pitched Interpolators
The Type A and Type C interpolators are designed for pitched or periodic material, such as solo
instruments or any time the fine structure of a waveform is visually repetitive. The difference
between these two options is that the C Type has validation built into it for certain cases in which
the interpolator may produce less than perfect results. The A Type lacks these “protections,” so
its results may occasionally be unusable. In addition, both the A and C Type may fail to find an
“acceptable” periodicity and will alert you to try a different algorithm. If this occurs, try the D
Type or, select a slightly different region of audio and try again.
The A and C Type interpolators take contextual information from six periods to the left and
right of the selection. After interpolation, the Restore Bar may extend for some distance outside
the selected region. This is because these interpolators are repairing based on wavelength in
addition to simple selection duration.
1.1.1.3
The D Type DeClicker
The D Type and E Type DeClickers, discussed below, are very high-order algorithms used to
correct problems that elude other repairs. Both interpolators use 64 bit precision to produce very
high quality interpolations.
The Type D DeClicker is tailored for use on the human voice, though it will provide excellent
results on most any semi–periodic material. It is only capable of replacing about 80 milliseconds
worth of samples before it bogs down. Even so, a repair half that duration will still take quite a
while, even on a fast computer.
Page 9
Figure 1.5: A fairly periodic section with click in Panel 1, sympathetic click in Panel 2
Figure 1.6: The repair using D Type
Page 10
Figure 1.7: A zoomed in view of the above repair
1.1.1.4
The E Type DeCrackler
The Type E Interpolator, though fundamentally similar to the D Type, is implemented so it can
be applied to passages of unlimited duration. As with the D Type, expect to wait a while for
your result but, it is worth it. The E Type DeCrackler is capable of reducing distortion, including
offensive, harsh sounding material. Its micro–repairs leave the audio sounding better without
resorting to low pass filtering.
Figure 1.8: E Type’s micro–repairs
If you own the DeCrackle native option for NoNOISE II, then an additional E Type command
appears under the default entry in the NoNOISE II menu. The second E - DeCracker… command,
with an ellipses in the command name, is an adjustable Type E interpolator, with a “Clip Fraction”
or threshold that ranges from 0.95 to 0.99. The 0.95 setting is highest sensitivity, with the heaviest
amount of repair, while the 0.99 setting is least sensitive, with the lightest amount of repair.
Page 11
Note that this command also forces or presets the threshold of the default E
Type command for subsequent repairs. The current threshold value is shown, in
parentheses, next to the command’s name as part of the menu entry.
1.1.2
Using Manual DeClick
Manual DeClick performs stereo repairs. Either channel of a stereo pair can be operated on, the
repair will be performed on both.
1.1.2.1
Removing Clicks
1.
Using the Waveform display and playback, identify the location of an impairment.
2.
Zoom in until you can clearly see the impairment.
3.
Click–drag in the Panel to create a time region selection that fully contains the
damaged samples.
4.
From the NoNOISE II menu, select one of the Manual DeClick types.
NoNOISE replaces the compromised audio with repaired samples. Don’t worry about selecting
on zero crossing boundaries, the software’s intelligence will provide a seamless transition.
1.1.2.2
Undoing — Restoring Repaired Clicks
1.
You can undo a Manual DeClick repair at any time, restoring the original samples back
into the sound file. To do this, zoom in on the waveform and locate the red “Restore
Bar.”
2.
Click–drag in the Panel to select a region that contains the Restore Bar.
3.
From the NoNOISE II menu, choose Restore Click.
NoNOISE replaces the interpolated audio with the original audio samples containing the
anomaly.
1.1.2.3
Hiding Interpolation Bars
The Show Interpolations command in the NoNOISE II menu is a toggle. When enabled, it displays
the interpolation bars that mark the site of a repair. When disabled, the bars are not visible,
which reduces visual clitter.
1.1.3
Obscenity Reduction
In addition to restoration duties, Manual DeClicking can be used to insure the public acceptability
of obscene material or to conceal any audio that may not “pass muster” with downstream listeners.
Simply select the obscenity as though it were an impairment, and choose your Type. The B Type,
Page 12
when given a one second region to “repair,” does a great job of removing the objectionable
material and inserting something that will often be preferable to editing in replacement audio.
An additional consideration is that, since no material is added or removed, the timing or tempo
is not affected.
1.2
1.2.1
Broadband DeNoise
Introduction
Broadband noise, whether white, pink or brown in spectrum, is one of the most common forms
of audio degradation. Noise can be introduced from any of a number of sources, including the
modulation and asperity noise inherent in analog tape recording and Johnson or thermal noise
from microphones, preamps, and other analog signal processing equipment. Broadband DeNoise
is a single ended broadband noise reducer that, unlike less refined examples, can suppress or
eliminate broadband noise with little or no audible artifacts, even at extreme settings.
To suppress such noises in your program, it is necessary to analyze the noise spectrum and
adapt the denoising algorithm to the characteristics of the material. Broadband DeNoise operates
by means of analysis and resynthesis, though it can be thought of as 2048 bandpass filters,
each followed by a below–threshold expander. This is somewhat analogous to the classic analog
Dolby and dbx multi–band, double–ended noise reduction systems of yore.
An FFT or Fast Fourier Transform analysis is performed on a user–specified sample of noise
from the material to be processed. The average amplitude of noise in each of 2048 individual
frequency bands or “bins” is determined. The result of this analysis is a “Noise Estimate,” which
is used to set the average threshold of the noise reduction. The Estimate sets aggregate threshold
for the expansion.
During processing, the source material is also subjected to a 2048 point FFT analysis. The
average amplitude of signal in each bin is compared against the threshold value determined by
the noise Estimate. Based on this comparison, the algorithm determines whether a given band
at that particular instant contains audio signal or only noise.
If a frequency bin is found to be at or above threshold, its gain remains at unity. If it is determined
that the signal amplitude in that band falls below threshold, then it is considered “noise” and
the amplitude of that band is reduced by an amount determined by the Attenuation parameter,
discussed later. The results of this comparison and amplitude compensation for all bands is a
modified version of the original FFT frequency analysis. A reverse FFT is then performed using
the new, adjusted version, reconstituting the audio signal with aggregate noise attenuated by
the specified amount. Because the Broadband DeNoiser operates with high frequency resolution
and at extended precision, the removal of noise is precise and artifact–free.
1.2.2
The Noise Estimate
The first step in denoising is to derive a Noise Estimate from the material to be processed. The
Noise Estimate, or simply Estimate, is an individual “fingerprint” of the noise and determines local
threshold values for each frequency bin. The NoNOISE II menu provides a group of commands
Page 13
for creating, processing, editing and storing the Estimate. The Estimate determines the result
of the entire denoising process, so it is important to ensure that the estimate taken is valid and
represents the true noise floor of the source sound file.
1.2.2.1
Noise Estimates
The procedure for taking a usable Estimate has several steps. First, open the source sound file
into a Project. Then, identify a short section of audio where there is only noise or predominantly
noise. About 0.3 to 0.5 seconds is sufficient. Click–drag to create a time region selection of the
noise. Section 1.2.2.2 below discusses where to take a Estimate in detail.
From the NoNOISE II menu, select the Take and Interpolate Estimate command to first create
the Estimate, then automatically interpolate the Estimate, applying “Bin Controls,” which are
individual threshold controls for separate regions of the frequency spectrum.
The separate Take Noise Estimate and Interpolate Noise Estimate commands are for
experimenting when creating an optimal Estimate and can usually be skipped in favor of the
combined NoNOISE II > Take and Interpolate Estimate command. See section 1.2.2.3 below for a
discussion of these individual commands and their function.
The Auto Noise Estimate… command combines the Take and Interpolate Estimate and Write
Estimate commands for those jobs where you are in a hurry or cost takes precedence over
quality. It is not recommended to use this shortcut if you desire the best result as it does not
include the interim step of adjusting Bin Controls.
Figure 1.9: An interpolated Estimate with Bin Controls
At this point, you now have a green “Interpolation,” with blue Bin Controls attached. Select the
Write Estimate... command to save the Estimate as a file that will be used by the Broadband
DeNoise processor.
1.2.2.2
Where to take an Estimate
Once the source sound file is opened into a Panel of a Project, the first step is to identify a
suitable location from which to take the Estimate. Since the denoising algorithm depends on a
constant amplitude and spectrum in the noise floor, try to listen and locate a time region with
uniform noise. As to duration, optimum results are obtained when the Estimate is taken from a
section of pure noise between about 0.3 and 0.5 seconds in length, with a worst case minimum
of 100 milliseconds.
If a region of pure or “clean” noise, noise uncontaminated with program, is unavailable as is
often the case with tightly edited material, then choose a region with minimal program. The
resulting Estimate will require manual adjustment as discussed below. If an Estimate must
Page 14
be taken in presence of signal, it is advisable to avoid sections of spectrally complex or non–
harmonic material as it makes manual adjustment more time consuming.
It is usually necessary to derive a separate Estimate for each cut or take. If these are contained in
a consolidated sound file and the spectrum of the noise varies for each cut, then the consolidated
file should be denoised in sections, so that the optimal combination of estimate and parameters
can be applied to each cut.
Unless there is strong reason to believe that each cut in a compilation or consolidated file was:
1.
recorded in the same session with the same equipment
2. at precisely the same levels onto the same media
3. stored in the same way
4. transferred to the same intermediate media in precisely the same way
5. and converted to digital samples in the same signal chain
then it is advisable to take separate estimates for each cut or take. When denoising stereo
material, it is also recommended to derive a separate Estimate for each channel. Likewise, if
the character or level of the noise floor can be heard to change at all during the recording, then
the best results are be obtained by dividing that piece into sections to be denoised individually.
After denoising, the individual sections can edited together and sequenced to create a seamless
whole.
1.2.2.3
Post–processing an Estimate
After taking an estimate, the Panel changes to show the Estimate, a smoothed FFT. Before the
Estimate can be used to provide threshold data for denoising, it must be additionally smoothed
and, perhaps, manually adjusted to account for the presence of desired signal within the noise.
After deriving the Estimate, select the Interpolate Estimate command. soundBlade performs
a best fit approximation from the composite FFT data, displaying it as a smooth green spline
connecting a series of blue boxes, the Bin Controls mentioned above, superimposed on the
composite FFT. You can toggle the Interpolation display by clicking on the checkbox located in the
lower right corner of the Estimate display or with the Show/Hide Interpolation command.
If the Estimate is derived from a clean sample of noise, then Interpolate Estimate command
produces an accurate measurement of the representative noise floor. In cases where a clean,
program–free sample of the noise floor is not available, a smooth Interpolation can be derived
from just a portion of the FFT display, provided that the noise in question is purely analog tape
hiss, by using the Fit Estimate From Selection command.
Analog tape hiss has a characteristic noise curve or shape. Fit Estimate From Selection takes
advantage of this property to produce a smooth Interpolation from any individual region of the
Estimate based on that region’s average amplitude. Make a time region selection of a portion of
the Estimate that appears to be outside the range of the contaminating audio program. Choose
Fit Estimate From Selection from the NoNOISE II menu. soundBlade displays an Interpolation with
a characteristic tape hiss curve using the average amplitude of the Estimate’s selected region.
Page 15
1.2.2.4
Editing an Estimate
The Fit Estimate From Selection technique works well for recordings containing typical tape
hiss, but many recordings have other noise sources or peculiarities. In such cases, it is often
necessary to manually adjust the Interpolation, differentiating between the desired program
content and the noise floor.
As with a waveform display in a Panel, the same Zoom In/Out and Zoom To Selection commands
work for an Estimate. This means you can zoom to a view that provides a comfortable level of
detail.
To edit an Interpolation, hold down the option key and drag a Bin Control up or down. If,
for example, a number of piano harmonics are contaminating the Estimate, the Bin Controls
can be moved down so that the Interpolation follows the average amplitude rather than the
local amplitude. Moving a Bin Control down will command a lower threshold and cause less
sensitivity in that Bin, resulting in less noise reduction while moving a Bin Control up sets a
higher threshold, causing more noise reduction.
You can move multiple Bin Controls at once. To do so, click–drag in the Panel to create a time
region selection that contains the Bin Controls. The numeric display at the left side of the Panel
shows the selection’s start frequency, end frequency and span in Hertz. To edit the boxes within
the selection, hold both the shift and option keys and drag one of the included Bin Controls. All
Bin Controls within the selection will move together.
1.2.2.5
Saving an Estimate
Once you are satisfied with the Interpolation, you need to save it. To do so, select the Write
Estimate... command from the NoNOISE II menu. soundBlade displays a Save File dialog. Choose
a location for the Estimate file, rename it if you wish, then click OK. soundBlade saves the file,
including the Interpolation and Bin Control state, to disk.
1.2.3
Running the Broadband DeNoiser
1.2.3.1
Overview
Once you have the Estimate file written, you can start the real–time process. From the NoNOISE
menu, select the BBDN… command, which brings up the BBDN window.
II
The window, with its six controls discussed below, also has buttons to open an Estimate file, to
open and save the controls and their parameters, to bypass the process, and to toggle between
the processed signal and the suppressed or removed portion of the signal.
Page 16
Figure 1.10: The Broadband DeNoise window
Note that BroadBand DeNoise is also available for use as a plug–in. This plug in
can be used as a Desk Event in an EDL, as a plug in in the Mixer window or in the
Meter window. This plug in is accessed from the contextual menu for plug ins.
1.2.3.2
Button Functions
The Open Est button spawns a standard Mac browser that allows you to specify and open an
Estimate file. This will set the threshold and sensitivity of each Bin.
The Save Param buttons allow you save all the current settings in the window, including the
name and location of the currently selected Estimate file. The Open Param button allows you
to open an existing Parameters file. These files are ASCII text and can be opened in TextEdit for
inspection.
The last two buttons are the Bypass, which patches the process in or out of the signal flow, and
the NoNOISE/Noise button, which allows you to hear either the processed signal or the portion
of the signal that is being suppressed or reduced. When in the “Noise” mode, you can adjust the
controls and hear the direct result of your actions.
The operation of the Broadband DeNoiser is a subjective process. In general though, if too much
program is present in the suppressed signal, then you are probably using setting that are too
Page 17
aggressive or your Estimate requires adjustment. Also, it’s fairly easy to misinterpret noise as
high frequency content so, critical listening with wideband reproducers are required for proper
operation.
1.2.4
Broadband DeNoise Parameters
The following parameters are available as horizontal sliders:
Parameter
Threshold
Attenuation
Sharpness
Bandwidth
High Cutoff
Low Cutoff
1.2.4.1
Range
-60 to 60
-60 to 0
0.5 to 2.5
0.5 to 3
0 to 22050
0 to 22050
Default Value
0
0
1.25
1.50
22050
500
Recommended
n/a
-10 to -24
0.80 to 1.25
1.8 to 2.2
22050
50 to 2000
Extreme
n/a
> 1.50
< 1.0
n/a
25
Threshold
The Estimate defines the aggregate threshold applied to each of the over 2,000 individual Bins
used by the denoise process. The Threshold parameter allows the entire curve, as a whole, to be
moved up or down. Together with the Attenuation parameter, the Threshold parameter provides
the basic control over how aggressively the process is applied.
As Threshold is raised, more of the signal at all frequencies is processed. At extremely high
settings, a distinctive watery aliasing may be heard in the resulting audio. If the Threshold is set
too low, little or no noise reduction is obtained. The Threshold can be thought of as the fine line
between noise and music, globally raising or lowering the entire Estimate curve relative to its
original position. Local adjustment of the threshold according to frequency bands is effected by
adjusting the Interpolation curve and saving a new Estimate file.
The default Threshold is arbitrary. Threshold and Attenuation settings should generally be
adjusted together for best results.The Noise/NoNOISE button can help to determine an acceptable
compromise setting.
1.2.4.2
Attenuation
This value, in decibels, sets the maximum attenuation to be applied to any Bin. A setting of
0 produces no noise reduction. The higher or more negative this value is set, the greater the
reduction in noise, but with increasing danger of producing audible artifacts in the audio signal.
Again, Threshold and Attenuation settings should be adjusted together for best results. Critical
listening and the Noise/NoNOISE button can help to determine an acceptable compromise
setting. If the maximum attenuation setting is too extreme, ambience and/or high frequency
content may be lost.
1.2.4.3
Sharpness
As mentioned earlier in the introduction, the denoising process works much like a multiband
downward expander. As signal level in a particular band drops, the process reduces the gain in
Page 18
that band even further, using an internal attenuation curve. The Sharpness parameter sets the
slope of this curve. Higher values cause quicker attenuation as instantaneous energy falls off
from the Threshold value, resulting in a response similar to that of a noise gate.
Figure 1.11: Sharpness Curves
If Sharpness is set too low, there may be no reasonable amount of noise reduction possible
despite what the other parameters are set to. Generally this parameter should be set as high as
possible without audible ill effects on the program. A value of 1.0 is recommended for common
tape hiss problems, while a value of about 1.2 has been found useful for standard 78 RPM
phonograph recordings. If the sharpness is too high, you may hear the noise floor become less
uniform, producing a rapid modulation.
1.2.4.4
Bandwidth
Denoising with each Bin adjusted separately produces an unnatural sounding result. In
Broadband DeNoise, individual Bins share information for more pleasing subjective result.
Bandwidth governs this process. Higher values produce more intra–Bin sharing of gain changes.
Subjectively, a higher value will, in general, create a more natural sounding result but, with a
risk of noise floor modulation. A low value for Bandwidth eliminates the possibility of noise
pumping, but may sound more muffled. Low values command little sharing while higher values
command more sharing. As with other Broadband DeNoise settings, experiment for the best
compromise setting. This depends entirely on the program material and should be high enough
to retain high frequency response but low enough to avoid pumping or distortion caused by the
noise floor being modulated by the high order harmonics of a signal.
Page 19
1.2.4.5
High Frequency Cutoff
The denoiser lets signal in frequency bins above the high frequency cutoff point pass through
the process untouched. That is, they are not processed. This limits processing to frequencies
below the high cutoff point. This can be used in situations where noise is not objectionable
above a certain frequency, but in most cases, this parameter is left set at 22,050 Hz.
In special cases, better results may be obtained by processing your program in two passes,
with a low frequency pass and a high frequency pass run separately, using different parameter
values. Run your program through Broadband DeNoise twice, using the low and high cutoff
frequencies to define the two bandpass ranges to be noise reduced.
1.2.4.6
Low Frequency Cutoff
This parameter, the complement to the High Frequency Cutoff parameter discussed in section
1.2.4.5 above, allows signal below the specified frequency to pass through unchanged. This
can be useful if noise is not objectionable below a certain frequency and you wish to leave it
unprocessed. Low Frequency Cutoff is usually left set around 50-100 Hz though setting it at 2.5
kHz or above will ensure that the critical midband will remain untouched. If this parameter is
lower than 25 Hz, there may be artifacts because program wavelengths will exceed the analysis
“window.”
1.3
reNOVAtor
The ReNOVAtor™ option for soundBlade allows localization, identification and very precise removal of unwanted audio events without affecting the audio material you want to keep. The removed sound is replaced by a signal re-synthesized from the surrounding material. reNOVAtor
does not make deep gaps in your sound track when eradicating a disturbing sound event. Rather,
it’s an exactly tailored hole in the spectral representation of the processed signal that can be removed and replaced. The interpolation may even be restricted to certain gain ranges within the
selected area, which is very useful if only a certain part of the signal needs to be treated (e.g. one
specific harmonic). The reNOVAtor window is fully resizable for increased accuracy and optimal
compatibility with all screen resolutions.
Working with reNOVAtor is easy and intuitive. reNOVAtor loads the requested part of audio
material you’ve chosen and analyzes it. The result is displayed as a 3D spectrogram with time
on the horizontal axis, frequency on the vertical axis and amplitude of the spectral components
color-coded. The color assignment follows the order of the rainbow: red and yellow for low energy; green and blue for middle energy; and finally purple and white for high energy. After getting some experience, this 3D spectrogram representation allows a good feeling for localization
and identification of sudden unwanted acoustical events. The spectral area of interest can be
precisely marked with a resizable rectangular window. A Play button allows you to hear selected
parts of the processed signal.
1.3.1
Main Features
• up to 192 kHz sampling rate
Page 20
• extraordinary results compared to any other cleaning method, due to selective treatment of spectral representation of the signal and not its waveform.
• enormous time savings when repairing critical live recordings
• easy-to-learn identification and localization of unwanted audio events
• efficient removal of unwanted audio events and their replacement by signals
• re-synthesized from the surrounding audio material
• resizable and zoomable spectrogram window for sound repairing with surgical precision
• multiple selections of harmonics and automatic identification of tones and clicks
• audition of any selected area before and after processing
• multiple undo functions
• gain selective signal treatment
• different types of interpolations
• replacing one spectral region by another (copy & paste)
• no audible changes in desired signal and ambience after removing typical discrete
audio disturbances
1.3.2
Typical Applications
• Removing unwanted noises like sneezing, chair squeaks, coughing, car horns, fallen
coins and keys, ringing of a mobile phone, etc.
• Correcting instrumental tracks by removing scratches from stringed instruments,
wrong notes, rustle of sheet music, keyboard pedal noise, vocalist’s breathing, lip
smacks and microphone pops
• Restoring old recordings by removing scratches and dropouts
• Cleaning up environmental noise on location recordings for film and television
1.3.3
Getting Started
Start reNOVAtor by click–dragging a time region selection in a Project’s Panel. Then, select
NoNOISE II > reNOVAtor… to start the processor. After you have selected a wav file to be processed, the reNOVAtor window pops up and shows the spectrogram of the left channel (left).
Page 21
Figure 1.12: The main reNOVAtor window
Typically, files up to 10 minutes can be loaded all at once. The maximum length of audio files
is dependent on your host’s computational power and RAM complement, the required spectrogram resolution and sample rate for the source sound file. If you experience any performance
problems with long files, cancel out of reNOVAtor and shorten the time region selection.
Though reNOVAtor processing is undo–able at any time, we recommend you always make a
backup of your original audio material to have the opportunity to start again if, for any reason,
the NoNOISE metadata, the “.cd” and “.rl” files, are lost.
Before you begin, remember that “tool tips” are available at any time to remind you of the function of a particular control. To see a tip for any control, simply locate your cursor over a control
and hover there for a moment. A contextually appropriate tool tip will then pop up with a brief
description of that control or feature.
To adjust the display color range to the material, click–drag on the Range field and move the
mouse. Mouse movement along the y-axis, up and down, adjusts the middle position of the
gain while movement along the x-axis, left and right, adjusts the overall gain range. As with all
numeric fields in reNOVAtor, a double click in the field resets the gain parameter to the default
value. In this case, -35 ± 35 dB.
Page 22
To the right of the Spectrogram label, there are four buttons. The second button toggles between
three color mapping schemes:
• “physical,” with red representing lowest energy, through yellow, green and blue to
white, representing the highest energy
• “standardized,” with blue representing the lowest energy, through green and red to
white, representing the highest energy
• monochrome, with black representing the lowest energy, through to white, representing the highest energy
• inverse monochrome, with white representing the lowest energy, through to black,
representing the highest energy
To zoom in on an area of interest, first select the portion of the spectrogram by holding the left
mouse button and drawing a selection or marquee. Then right-click to open the popup menu, as
shown below. Finally choose one of the zooming operations (see Quick Reference for details).
The selected area can be moved or resized as indicated by the mouse cursor. Its length in seconds and milliseconds is displayed inside. The shading grade of the area can be controlled by
the Bright parameter in the Spectrogram Parameter Group from darkening (as shown above) to
brightening. A double left-click on the Bright resets it to the neutral position (192).
Another way to select the visible spectrogram portion is by resizing and moving the scrollbars
beside the display. By left-clicking and moving the edges of the horizontal or vertical scrollbar,
the visible spectrogram region can be zoomed in or out and moved. When the SHIFT-key is also
held down, zooming is performed symmetrically. The length of the displayed signal is indicated
in the upper part of the spectrogram.
The displayed audio channel can be toggled by left-clicking on the channel field (left, right, l+r).
Block size applied to the spectrogram and interpolation is accessed by left-clicking on the Blocksize field, holding the button and selecting the value from the pop-up list (selectable values: 64,
128, 256, 512, 1024, 2048, 4096, 8192, 16384 and 32768). In general, interpolation of short disturbances (like clicks) requires smaller block sizes, while a frequency selective interpolation (like
removal of discrete tone) requires larger block sizes.
The whole loaded audio material or any portion can be replayed at any time by positioning the
white play cursor at the desired position and clicking the Play button. To place the play cursor,
left-click at the bottom (black) end of the spectrogram where the mouse cursor changes to the
play cursor , or --while holding ALT-key-- just click on any position in the spectrogram. A left-click
on the Stop button stops the audio playback. After stopping, the play cursor returns to its initial
position.
To remove an unwanted disturbance, select an area around it and set up the desired parameters
in the ReSynthesis Parameter Group. These are Accuracy, Channel, Type and Gain (see in Quick
Reference for all setup possibilities). To remove clicks as shown in the example above, you will
usually use the hor interpolation type to replace the selected area containing respective click
with a new signal re-synthesized from the material surrounding the click along the time lime.
Page 23
The interpolation types left or right can be used, if one side of the click is not suitable for proper
interpolation (e.g., it includes strong percussive beats). However, the results in case of one-side
interpolation are less accurate.
To perform the interpolation in a selected area, hit the Process button or select Process from the
drop-down menu. After processing has been done, you can immediately listen to the result by
resetting the play cursor and hitting Play button. The lower screen shot from the previous page
shows the audio piece from the upper screen shot after click removal.
If you are not satisfied with the result, you may undo the interpolation with the button or by
pressing ctrl-z. The maximal number of undo steps is limited only by the available computer
memory. If you reach the memory limit you can clear the undo buffer by hitting the button. Once
an undo step is performed, it can still be re-done by hitting the button or ctrl-y on the keyboard.
However, after the reNOVAtor has been closed with Accept or Cancel, all pending re-do steps are
no longer accessible.
To finish your reNOVAtor session, hit the Accept button to accept the changes and write them
back to the track or hit cancel to exit the process and discard all changes.
For precise spectrum analysis the numerical display in the middle below the spectrogram is
provided. It shows spectrum properties at the current cursor position: time, frequency and amplitude.
For precise spectrum analysis the numerical display in the middle below the spectrogram is
provided. It shows spectrum properties at the current cursor position: time, frequency and amplitude.
Note that as long as you stay within a reNOVAtor session all processing steps are stored in the
temporary memory and can be re-done. However, after closing the session with the Accept or
Cancel button, all intermediate steps get lost.
All parameters, their settings, and remaining buttons are precisely described later in the Quick
Reference.
1.3.4
Quick Reference
1.3.4.1
Spectrogram Setup
To adjust the spectrogram color range of the material, click–hold in the Range field and move
the mouse. Movement along the y-axis, up and down, adjusts the middle position of the gain
while the movement along the x-axis, left and right, adjusts the gain range.
The
1.
button toggles between four color mapping schemes:
physical with red representing lowest energy through yellow, green and blue to white
representing the highest energy
2. “standardized” with blue representing the lowest energy through green and red to
white representing the highest energy
Page 24
3. grey scale with black representing the lowest energy through to white representing
the highest energy
4. inverse grey scale with white representing the lowest energy through to black representing the highest energy
button activates smoothing of the spectrogram. The
The
linear and logarithmic frequency axis.
1.3.4.2
button switches between
Spectrogram Resizing
In general, the entire reNOVAtor window is resizable. The spectrogram itself can be zoomed
in or out. It can also be moved independently in both directions, horizontally and vertically, by
resizing and moving the associated scroll bars. With the shift key held down, zooming is symmetrical. The length of the program portion actually displayed is indicated in the upper part of
the spectrogram.
When clicking in the scroll bar area, but outside the bar, the zoom area moves 1/3 display toward
the direction clicked. This can be useful for tracking the play cursor during playback.
1.3.4.3
Marking an Area
An area of interest can be click–dragging on the spectrogram and drawing a marquee or selection around it. The selected area can be moved or resized as indicated by the mouse cursor. Its
length in seconds and milliseconds is displayed inside. The contrasting shading of the area can
be controlled by the Bright parameter in the Spectrogram Parameters.
To move a marquee to another location, move the cursor to the center of the selection. The cursor changes to the Move cursor, a four quadrant arrow shape.
Figure 1.13: The Move cursor
Once you see the Move cursor, you can click–drag the marquee around the spectrogram. By holding
down the shift key before dragging, you can constrain movement to only the horizontal or time axis.
Page 25
1.3.4.4
Zooming a Selected Area
Zooming in a selected area can be performed in one of two ways. First, from the pop-up menu
that opens after control–clicking in the spectrogram or by using shortcuts, see section 3.3.10
below. The second way is to place your curson in the black, horizontal time scale or the vertical,
frequency scale. The cursor changes to a magnifying glass, allowing you to zoom in on that axis.
Holding down the shift key allows you to zoom out.
Note that resizing the spectrogram with scroll bars respectively changes the size of the selected area in the spectrogram, but its absolute length, in seconds and milliseconds, remains
unchanged.
A option–clicking inside the spectrum display enters the zoom/move mode. Moving the mouse
up and down zooms out and in. Move left or right to scroll. With this feature, it is possible to click
into an area of interest and move the mouse down in order to zoom the spectrogram around the
mouse cursor.
Additional zooming possibilities are provided by key equivalent. See section 3.3.10 below for
more information
1.3.4.5
Playback
The whole loaded audio material or part of it can be played back at any time by clicking in the
spectrogram to position the white play cursor at the desired position. Clicking the Play button or
tapping the space bar start playback. After playback has stopped, the play cursor returns to its
initial start position.
Additional playback possibilities are provided by key equivalent. See section 3.3.10 below for
more information.
1.3.4.6
Copy & Paste
A selected area can be copied to another spectrogram position. After clicking the Copy button or
selecting Copy Area from the contextual menu, it can be moved around with the mouse to the
desired destination. As long as the copy mode is active, the field Type shows copy.
To paste the selected area into the desired location, click the Process button or, select Process
from the contextual menu. The marquee changes from white to blue.
To exit the Copy mode, simply click outside the marquee. The ReSynthesis Type will switch back
to its previous selection and the marquee changes back to white.
The copy and paste function can be restricted to the original position in frequency if the shift key
is held down while relocating the marquee. This is very useful, when the area to be copied contains material, such as harmonics, that needs to be placed in exactly the same frequency region.
Page 26
1.3.5
User Interface Details
1.3.5.1
Buttons
- Undoes the interpolation (or ctrl-z).
The maximum number of undo steps is limited only by the available computer memory.
- Redoes the last undo step (or ctrl-y); be aware that after the reNOVAtor has been
closed with Ok or Cancel all pending re-do steps are no longer accessible.
- Introduces the Automatic Selection of Harmonics mode, see section 1.3.8.2 below.
It allows selecting multiple harmonics belonging to a marked fundamental frequency. The fine
tuning of the fundamental frequency and enabling or disabling each individual harmonic selection can be done in the Harmonics Selection window that pop-ups after clicking the harmonic
selection icon.
- Toggles between four color mapping schemas: (1) physical with red representing
the lowest energy, over yellow, green, blue to white representing the highest energy; (2) “standardized” with blue representing the lowest energy, over green, red to white representing the
highest energy and (3) monochrome from black representing the lowest energy, over gray to
white representing the highest energy and, (4) an inverse monochrome mode.
- Activates smoothing in the spectrogram; it should normally be on; to switch it off
makes sense if we need to precisely see the borders of very sharp spikes.
1.3.5.2
- Toggles between linear and logarithmic frequency axis.
Cursor Callout
if the mouse cursor is located within the spectrogram, the cursor callout indicates the properties
of the spectral component at the current cursor position: time, frequency and amplitude.
Figure 1.14: The cursor callout
1.3.5.3
Menus
A contextual menu opens after control-clicking on the spectrogram.
Page 27
Figure 1.15: The spectrogram’s contextual menu
• Zoom +
– This choice zooms in the selected area to the maximum dimensions,
limited only by the display size and spectrogram resolution.
• Zoom –
– This choice zooms out from selected areas.
• Zoom Hor
– This choice zooms in on the selected area horizontally only.
• Zoom Ver
– This choice zooms in on the selected area vertically only.
• Copy Area –
This
choice
copies
the
contents
of
the
selected area, allowing you to move it to another location in the spectrogram for later processing or pasting. This is equivalent to the Copy button. Once copied, the marquee changes from black to blue, and the entire selected areas
can now be dragged around to relocate it in the spectrogram. To limit movement of
the copied area horizontally, to accurately paste harmonic content for instance, hold
down the shift key before dragging.
• Process
– This choice carries out the interpolation. This is equivalent to the Process button or hitting the P key on the keyboard.
1.3.5.4
Main Buttons
– Starts playback from the current cursor position
– Stops playback and returns the play cursor to initial position
– Carries out the interpolation in the selected area
– Discards all signal modifications and exits the option
Page 28
– Accepts all signal modifications and modifies the source sound file with a processed region
1.3.6
ReSynthesis Parameters
Accuracy, Channel and Type can be selected from their respective menu that appears after clicking in the field. Gain can be adjusted by click–dragging the mouse up or down in the field.
Accuracy – Accuracy [1, 2, 4, 8] sets up the time resolution of the interpolation process and
the spectrogram while maintaining the frequency resolution determined by the Blocksize field.
The higher the number, the greater the accuracy in time. This doesn’t mean that the highest
number is always best, since it depends on your audio material.
Normally, for vertical interpolation, 1 or 2 works optimally. For horizontal interpolation, higher
values might be more advantageous. Changing the Accuracy parameter results in recalculation of the whole spectrogram, which can take considerable time and memory. Therefore, make
changes only when a smaller piece of audio is loaded, no more then a few minutes.
Channel
– Channel [all, current, left, right] selects the channel on which the interpolation will
be performed.
Type
- Type [hor, vert, left, right, top, bottom, 2-dim, Gain, Copy] selects the kind of interpolation to be applied to your selection. All interpolation types, with the exception of Gain, perform a two step operation. First, they remove the signal from the marked area and subsequently
fill the spectral hole with a replacement re-synthesized from the surrounding audio material. The
name of the interpolation type explicitly describes the “context,’ what part of the surrounding
audio material is used for the re-synthesis.
The Horizontal (hor) interpolation type is most suitable for removing transient disturbances like
clicks while Vertical (vert) is recommended for removing longer tones or harmonics. The interpolation types Left or Right are a special form of Horizontal, and should be used if one side of the
impairment is not suitable as interpolation context. An example would be a strong percussive
beat just after an impairment. A left interpolation would use only earlier material as the re–synthesis context, avoiding a flam effect or partial duplication of the later percussive event from
inclusion in the resynthesis.
Similarly, the top and bottom are special forms of vertical interpolation and should be used if
one “side” adjacent to the impairment includes strong tones or harmonics. However, be aware
that the results of one-sided interpolations are not as accurate as in symmetrical Horizontal or
Vertical choices.
The 2-dim or Two Dimensional interpolation is a special combination of both horizontal and
vertical interpolation and is recommended for removing small “square” shaped disturbances.
The Gain interpolation type, see below, does not remove the original material from the selected
area like all the other interpolation types. It simply reduces or amplifies existing spectral components in the selection, according to the Gain parameter in the ReSynthesis Parameters.
Page 29
After selecting either Copy Area in the conxetual menu by control-clicking on the spectrogram,
Copy is enabled in the Type field and persists during the copy procedure. It can not be selected
directly from the Type field.
Gain
– The Gain field provides two functions. For the Gain interpolation type, it controls
the gain applied to the signal components in the selected area and, for all other interpolation
types except copy, the value acts as an intelligent threshold. The range is -60 +20 with a default
value of 0 dB. As with all numeric fields, a double click on the field resets the value.
The Gain interpolation type does not remove the original material from the selected area like all
the other interpolation types. It simply reduces or amplifies original signal components in the
selection, according to the Gain setting. In general, the higher the Gain value, the more of the
original signal is preserved. This means that a deep interpolation, of any type, can be carried out
only with a low Gain setting, such as -40 dB, while Gain or actually threshold above 0 dB preserves some parts inside the selected area which might be useful material. With Gain being too
high, however, even the disturbances may partially remain. In such cases, lower the Gain until
the expected result is achieved.
1.3.7
Spectrogram Parameters
As with the ReSynthesis Parameters, these parameters are either menus or numeric fields. As
with all numeric fields, a double click on the Brightness or Range fields resets the value to the
default.
Blocksize – Block Size [32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768], even though
it resides in the Spectrogram parameters section, selects the signal analysis resolution for both
the spectrogram and interpolation algorithm. In general, interpolation of short disturbances, like
clicks, require smaller block sizes, while a frequency selective interpolation, like removal of discrete tones or harmonics, requires larger block sizes.
Block size allows you to locate specific impairments prior to effecting a repair. Because larger
block sizes tend to visually isolate and repair areas at low frequencies, it allows you to avoid
touching areas that should be left intact.
Channel
– The Channel setting [left, right, l+r] shows the spectrogram of left, right, or a sum
of both. The sum [l+r] allows the identification of all disturbances regardless of the channel in
which they occurred.
Bright
– Brightness [0 – 384] controls the contrasting brightness of the selected area from
bright (192 – 384) to dark (191 to 0). It can be adjusted by click–dragging up or down on the field.
A double click in the field resets the brightness to neutral.
Range
– Range adjusts the spectrogram colors to the dynamic range of the audio material,
with a middle value of -100 dB to + 40 dB, and a range of ± 10 to ± 70. A double click in the field
resets the Range to the default values of -35 dB ± 35 dB.
Range can be adjusted by click–dragging vertically on the field, with up equating to higher sensitivity and down lower sensitivity for controlling the middle value. Click–dragging horizontally
on the field controls the mapping range.
Page 30
Be aware that the Range values are related to spectral components of the processed signal and
not to the overall signal level. Therefore, prior to processing, they may look unusual.
1.3.8
Interpolation Alternatives
1.3.8.1
Copy and Paste
While interpolation is the preferred method of removing isolated artifacts from complex audio
material, there are situations where this method suffers from adequate material around the artifact that could be used as context for the interpolation. An example of this would be where the
artifact itself is surrounded by strong transients. In such cases, the copy and paste method can
produce far superior results. With copy and paste, you can simply find another area that looks
like it could “fit into the space” you want to interpolate and insert it. reNOVAtor copies an spectrum from one region to another.
To perform the area copy operation, click–drag to select a source area and click the Copy button. The selection can now be moved around with the mouse to the desired destination. As long
as the Copy Mode is active, the Type parameter will show Copy. To paste the selected area into
the destination location, either click the Process button or select Process from the contextual
menu. To leave the copy mode, simply click somewhere outside the selection. The Type parameter switches back to its previous value.
Copy and paste can, in desperate situations, be used to pitch correct material. First, locate and
mark the fundamental. Then, copy and paste that region and test the result. If it acceptable, then
move up the “chain” of harmonics, pasting them in at the desired locations. Once you have
achieved the desired result, localized in time, then you may want to perform a more global horizontal or 4-dim repair to smooth the result.
1.3.8.2
Automatic Selection of Harmonics
To accelerate removal of heavy hum or buzz as well as removal or shift of complete instrumental tones including thier fundamental and harmonics, reNOVAtor provides an automatic selection of harmonics associated with a selected fundamental. Of course, it’s possible to remove
an instrument tone just by drawing one big selection around the whole region covering both
fundamental and harmonics. For longer tones, however, that approach will produce artifacts and
reduce the integrity of your ambience. Using selection of harmonics, you keep much more of the
spectrum surrounding your harmonics unprocessed. In case of continues and extremely heavy
buzz, this is the only method to preserve at least a part of wanted signal. It is extremely useful
in forensic applications, especially if the desired signal is so strongly masked by buzz that the
signal-to-noise ratio drops even below zero. Also in a case when you need to distinguish two or
more overlapping or mixed instruments playing different tones, the only chance to remove or
delete individual tones is by using the Automatic Selection of Harmonics mode.
To activate Automatic Selection of Harmonics mode, mark the fundamental tone drawing a
horizontal rectangular around it and press the the Automatic Selection of Harmonics button. The
Harmonics Selection window pops up, allowing switching on or off each individual harmonic.
Page 31
Figure 1.16: The Harmonics Selection window
This selection function automatically draws rectangles around each harmonic enabled in the
interface. There are up to 19 individual harmonics and one fundamental which can be considered
in the interpolation process. The Harmonics Selection window monitors levels of each individual
harmonic belonging to the marked fundamental. The levels of harmonics, related to their energy,
can be displayed in dB or in % relative to the level of fundamental. Observing these levels you
can get a good feeling about weight of each harmonic. This helps also to estimate the correct
position of the rectangles marking the harmonics. If only part of a harmonic is bordered, the
displayed level is smaller.
The fundamental frequency originally selected on the spectrogram is numerically displayed in
the Freq or Frequency field of the Harmonics Selection window. It can be fine tuned by click–dragging in this field up or down. This is a very important adjustment, because even small changes
Page 32
of the fundamental frequency determine the proper position of the resulting marking rectangles
for harmonics. Alternatively, the frequency can also be fine-tuned by holding down the shift key
while resizing the rectangle defining the fundamental. Each individual harmonic, and the fundamental, can be enabled or disabled with their respective button. Also, all odd or even harmonics
can be simultaneously switched on or off by using the buttons All EVEN or All ODD.
To the left and right of the Freq field, there are two additional fields. Amp or Amplitude shows
the amplitude of the fundamental frequency, while ∑rms shows the RMS energy of harmonics
plus fundamental covered by all active selected areas.
The amplitudes of spectral components are normalized in such way that ∑rms shows 0 dB when
processing white noise having a maximum line level of 0 dBFs, and the marking rectangles covers the whole frequency range from 0 to maximum.
The “Process between marked areas” check box, at the bottom of the Harmonics Selection window, allows an interpolation which is complementary to the harmonic removal process. In this
case harmonics are preserved and the space among them is cleaned. It is especially useful for
cleaning sound samples used for keyboards and sequencers.
Note that the Automatic Selection of Harmonics tool can also be used for the Copy & Paste
function. This is especially advantageous if you want to shift a tone, a fundamental with accompanying harmonics, while keeping the ambience integrity untouched. It also helps to correct the
beginning of a tone if it has been not precisely played.
1.3.9
Application Tips
reNOVAtor is an easy to use tool, even if, at first glance, you may feel uncertain because of the
unusual interface. However, after reading the Getting Started chapter, you will be able to carry
out repairs with speed and quality you have never experienced before.
This manual gives you just an overview of the reNOVAtor functionality, but there are still a lot
of tricks and “secrets,” allowing previously impossible sound repair operations. Indeed, it is very
difficult to formulate general statements on how to proceed. Every impairment, in connection
with specific audio material you want to retain, creates a unique situation. We recommend you
spend some time trying to localize and remove certain audible features, or copy audio events
to other destinations. Together with all zooming, resizing and precise playback cursor-placing
possibilities, you’ll get a good feel for controlling reNOVAtor in a short time. Below are some
additional remarks which may help to speed your learning process.
reNOVAtor repairs are always undo–able at any time by selecting NoNOISE II > Restore Click.
However, before loading a sound file to be processed, we recommend you always make a backup of your original audio material to have the opportunity to start again if, for any reason, the
NoNOISE metadata, the “.cd” and “.rl” files, are lost.
For removing transient disturbances like clicks, use the hor (horizontal) interpolation type and
for removing longer tones or harmonics, use vert (vertical) interpolation type. The interpolation
types left or right are a special form of hor and should be used if one side of the disturbing click
is not suitable for proper interpolation such as when the material includes a strong percussive
beat. Similarly, the top and bottom are a special form of vertical interpolation and should be
Page 33
used if one side of the impairment includes strong tones or harmonics. However, note that the
results of one sided interpolations are not as accurate as symmetrical repairs.
Be aware of the proper selection of the blksize parameter. In general, interpolation of short
disturbances, like clicks, requires smaller block sizes, while a frequency selective interpolation,
like removal of discrete tones or harmonics, requires larger block sizes. In difficult cases, such
as with loud, short duration sounds mixed with low-frequency reverberation, a step–by–step
removal process works better. Instead of a wide, vertical area selection, begin with a narrow one
to concentrate on the main energy and then a wider horizontal but vertically smaller selection to
remove the reverberation part. In addition, for the first step select a smaller block size, such as
256, and for the second step, a larger one like 2048.
If the disturbance is a wideband signal that strongly overlaps with the signal you want to retain,
it helps to mark and process carefully selected small regions, step–by–step, instead of drawing
one big selection around the critical area.
In extremely difficult situations, especially if you do not have enough clean material surrounding the unwanted noise, we recommend using the copy and paste technique. This method is also
ideal for correcting timing problems in instrumental or vocal performance. You can grab a whole
tone and shift it a little bit backwards or forwards.
reNOVAtor is also a perfect tool for improving distorted signals. It can remove distortions caused
by overdriving analog equipment and works as an excellent de-esser.
If you cannot locate the disturbance you hear, switch the spectrogram to a different Channel if
possible. Also, try a different Blocksize and adjust the Range parameter to get proper amplitudeto-color mapping. It may also help to increase the Accuracy, since this will lead to a higher
time resolution of the spectrogram while maintaining its frequency resolution, as determined by
Blocksize.
1.3.10
reNOVAtor Key Equivalents
General
space bar
shift-p
command-z
shift-command-z
shift-c
shift-v
shift-drag
` (accent grave)
play
process
undo
redo
copy selection
paste selection
constrain selection in ƒ (frequency)
toggles smoothing
Zoom
0
Page 34
zoom to entire
1
2
3
4
5
up arrow
down arrow
<
>
, (comma)
. (period)
; (semicolon)
/ (forward solidus or slash)
zoom 10x
zoom 5x
zoom 2x
zoom 1x
zoom 0.5x
zoom 0.5x
zoom 2x
zoom ƒ by 0.5x
zoom ƒ by 2x
zoom ƒ to top half
zoom ƒ to bottom half
zoom ƒ to middle half
zoom ƒ to entire
Presets
^ (accent circumflex)
& (ampersand)
* (asterisk)
(
6
7
8
9
set preset 1
set preset 2
set preset 3
set preset 4
load preset 1
load preset 2
load preset 3
load preset 4
Play Cursor Placement
y
u
i
o
set play cursor before selection, 6 sec.
set play cursor before selection, 3 sec.
set play cursor before selection, 1.5 sec.
set play cursor before selection, 0.5 sec.
Accuracy
}
{
Channel
Page 35
select next accuracy
select previous accuracy
shift-l
shift-r
c
a
select left channel
select right channel
select current channel
select all channels
Type
h
v
l
r
t
b
d
g
horz
vert
left
right
top
bottom
type 2-dim (2 dimensional)
gain
Block Size
=
-
Page 36
select next block size
select prev block size
Note that, for the PDF version of this Index, only the page numbers, not the preceding descrip-
tive subject text, are hyperlinked.
NoNOISE II SR User Manual Index
A
A Type 9
B
Bin Controls 14, 15, 16
Broadband DeNoise 13
Broadband DeNoiser 16
B Type 7
C
C Type 9
D
dbx 13
Dolby 13
D Type 9
E
E Type 11
F
FFT 5, 13, 15
Fit Estimate From Selection 15
Page 37
I
Interpolate Noise Estimate 14
Interpolation Algorithms 7
M
Manual DeClick 7, 12
N
Noise Estimate 13
NoNOISE II
A Type 9
Bin Control 16
Bin Controls 14, 15
Broadband DeNoise 13, 16
Attenuation 18
Bandwidth 19
High Frequency Cutoff 20
Low Frequency Cutoff 20
NoNoise/Noise button 17
Open Est button 17
Open Param button 17
parameters 18
Sharpness 18
Threshold 18
B Type 7
C Type 9
D Type 9
editing the noise Estimate 16
E Type 11
Fit Estimate From Selection 15
Page 38
Interpolate Noise Estimate 14
interpolation algorithms 7
Manual DeClick 7, 12
noise Estimate 13
obscenity reduction 12
post-processing the noise Estimate 15
saving a noise Estimate 16
Take Noise Estimate 14
Write Estimate... 16
O
Obscenity Reduction 12
P
Project
Panel 2–3
T
Take Noise Estimate 14
© 2012 Sonic Studio, LLC • 330 Sir Francis Drake Blvd. • Suite A • San Anselmo CA 94960-2552 • 1-415-944-7642
soundBlade, NoNOISE, sonicstudio.com, the Sonic Studio logo and type are trademarks of Sonic Studio, LLC in the United States and other
countries. All other trademarks, trade names, service marks, and logos referenced herein belong to their respective companies.
Comments, corrections and suggestions regarding this manual are always welcome. Please contact us at <[email protected]
sonicstudio.com>.
NoNOISE_II_SR_UM_V1r04
Page 39
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement