Operation Manual 1
Sonnox Restore Operation Manual 1. Introduction The Sonnox Restore suite is a collection of three plug-ins designed to clean the sound of digital audio recordings that are damaged: • by pops, clicks, or crackle; • by hum, buzzes and whines; • or by noise. The tools are not just for restoring old material, but also for removing the clicks, pops, buzzes and background noises that can sometimes accidentally occur in new recordings.
. The Oxford DeClicker 3.1 Basic Operation The Oxford DeClicker has three main sections: DePop, DeClick and DeCrackle. These three sections deal with events that are up to 10ms, 3ms and 0.4ms in duration, respectively. Each of the three sections calculates an Excitation Profile for a given block of samples. The excitation profile is the residual signal left after harmonic components have been removed, and is a representation of how much the actual signal is deviating from the predicted signal.
You can imagine a click as a triangular shaped peak in the excitation profile, wider at the bottom and tapering to a point at the top. Keeping this in mind, you have two potential strategies for detecting and repairing a click. The first is to lower the threshold in the excitation profile until it is just above the noise floor, near the base of the triangle. (You can do this either with the Threshold slider, or drag the threshold line on the graph itself.
The DePop section can detect events from 10ms down to 1ms. However, it is best to adjust the DePop Threshold and Sensitivity to ensure that the fewest repairs are done in this section. Use a low DePop Threshold and a minimum of Sensitivity to achieve this. If possible, use the DeClicker section to capture the longest and widest of the clicks. Again use a low DeClick Threshold and a minimum of Sensitivity to achieve this. 3.
Clicking in the Events Graph will remove any popup graphs such as the Excitation Profiles to expand your view of the Events. 3.4 The Exclude Box The Events Graph allows an important new feature to be implemented. The idea of the exclude box is to stop over-corrections. For example, you may be happy with a piece of music and the associated repairs, except for a section with brass instruments - at which point too many high energy events are detected and corrected.
3.5 Dialogue Mode The Dialogue Mode is another unique feature of the Oxford DeClicker. It is a complete solution to the problems associated with dialogue where you need one set of Threshold/Sensitivity settings for the background level, when the speaker is pausing, and a quite different set of Threshold/Sensitivity settings whilst speaking. When you enable Dialogue Mode the background GUI changes to display an extra panel.
3.6 Description of Controls Touch Pad Controls Several controls in these plug-ins use a Touch Pad associated with a knob or slider.
registered as an event to be repaired. If a peak breaches the threshold line, then it is registered as a detected event and can be repaired. “Sensitivity” – This control determines whether detected events are discarded, starting with the smallest. A low Sensitivity means only the largest of the detected events go forwards to the repair section. “Enable” –With the section disabled, the detectors are still monitoring the audio and any detected events are shown on the Events Graph as red bubbles.
“Exclude Box Show” – This control shows or hides the Exclude Box. The status of Exclude Box Enable is not affected.
3.7 DeClicker Signal / Control Flow Diagram 3.8 DeClicker Stereo Behaviour For correct stereo operation, when instantiated into a stereo track the DeClicker has two independent detect and repair circuits.
4. The Oxford DeBuzzer 4.1 Basic Operation The Oxford DeBuzzer is divided into two stages: Detection and Removal. Removal of buzz is usually accomplished by using a comb filter. The comb filter has many tight notches in the frequency spectrum that eliminate the fundamental frequency and associated harmonics of a buzz. For this process to work well, the frequency of the fundamental must be very accurately determined. There are several tools available to help isolate the fundamental frequency.
The second tool to help isolate the fundamental is the large FFT display. The FFT displayed by default in this graph is no ordinary FFT of the input signal, but can accumulate FFT data. It is designed to accentuate any frequency components that are consistently present, and average out everything else. Any buzzes, hums, whines or whistles that are consistently present tend to stand out in the plot. A third method of isolating the nominal frequency is a tone generator.
quickly wandering fundamental. This mode operates on MF and HF frequency selections only (if Fast is selected in the LF range, the mode is forced to Auto and the Auto button flashes). Auto and Fast modes are both applicable to Comb or Para-EQ settings. Fast mode is most appropriate for quickly drifting high frequency buzzes possibly generated by lighting interference. 4.3 Removal Filters The default removal filter is a Comb filter.
The CPU usage in Para-EQ mode is dependent on the number of filter sections operative; this is not the case in Comb mode. This is another reason why Para-EQ should be used with care with a low frequency fundamental. The Para-EQ filters offer several advantages over the Comb filter. The first is that you can reduce the amount of attenuation as you increase in frequency. You use the Slope control to do this.
This diagram shows a fairly broadband buzz, with elements up to 10kHz or more: The following sequence of figures shows the Audio output with the DeBuzzer in circuit but with steadily decreasing sensitivity.
Lowering the sensitivity further means the first harmonic appears in the Audio output: Lowering further allows more and more harmonics into the Audio output: When the Sensitivity control is very low the whole buzz will be in the Audio output, and the DeBuzzer is removing no buzz: 17
Clearly here the Sensitivity is far too low. The optimal setting is just before the fundamental appears in the Audio output. The Alternative Mode of Buzz Removal There is a second mode of Buzz removal that might be more appropriate under some circumstances. The sequence above is using the default mode of repair. If Alt Mode is selected, the characteristics of the sequence reverse, with the harmonics appearing in the Audio output first, and finally the Fundamental.
More harmonics appear: Finally the whole buzz spectrum including the fundamental is in the Audio output: The final settings will be a trade-off between removal of buzz elements and quality of repaired signal. This will depend on the characteristics of both the signal and the buzz. As both the signal and the buzz can be almost infinitely variable, there will be cases when each sensitivity mode will be appropriate. For broadband buzzes it is likely that the Default mode gives the best results.
The CPU load of the host computer is also dependent on the number of filters used. While newer high-performance computers can accommodate very many filters, there might be situations where a lower CPU overhead is desirable, either for lower performance computers or when many plug-ins need to be instantiated.
“Master Enable” – With Master Enable selected, all processing is active. If not selected, all repairs are disabled, although all the detection circuits remain active. The Master Enable, Audio and Diff buttons all provide smoothly interpolated and glitchless audio switching. It is advisable to use these for auditioning, rather than using the host application’s Bypass switch (which is commonly not interpolated). Frequency Detect Section “Auto” – Auto mode is the default tracking mode.
“LF”, “MF”, “HF” – These three buttons set which frequency band the big rotary control will operate over. LF = 20Hz to 160Hz. MF = 160Hz to 1280Hz. HF = 1280Hz to 20kHz. “Hz”. The Frequency control in the centre is used to set the nominal frequency about which in Auto mode the detection algorithms hunt for a peak in the peak profile. In Freeze mode it determines the frequency of the removal filters. The frequency can be selected using the scribble tab in units of 1Hz (normal), 0.01Hz (Fine adjust) and 0.
- Smart FFT – permanent accumulation - Smart FFT – with time constant - Minimum FFT “Zoom In” – Allows magnification of the display. “Zoom Out” – Allows de-magnification of the display. “Log / Lin Select” – Toggles the display between log mode and linear mode. Sonnox Menu Button “Clip Lights” – Can be set to hold for 2 seconds, 5 seconds, or indefinitely. “Show Preset Name Path” – Can show or hide the preset name path in the Preset Manager.
4.7 DeBuzzer Signal / Control Flow Diagram 4.8 DeBuzzer Stereo Behaviour For correct stereo operation, when instantiated into a stereo track the DeBuzzer has two independent repair circuits. There is a single mono detection circuit.
5. The Oxford DeNoiser 5.1 Basic Operation The Oxford DeNoiser removes wide-band noise from audio material using a variety of sophisticated techniques. The key to good de-noising lies in helping the user produce pleasing and accurate results, rather than encouraging over-aggressive use of the controls. The Oxford DeNoiser is based around the concept of a “noise profile”. White noise, for example, is flat all the way up the frequency spectrum. Pink or red/brown noise is louder at LF than HF.
well below valid signal components, but well above the noise floor. If Thresh is set too high, some of the signal will be attenuated; if it is set too low, the noise removal will be less effective. The Oxford DeNoiser provides different means to arrive at an appropriate noise profile. The default method is called Auto mode.
The 17 handles are distributed over the entire frequency range. With the display set to linear view (HF view), 10 handles are visible from 2kHz to 20kHz. When the view is set to Log (LF), 9 handles are visible from 20Hz to 4kHz. Two of the handles (at 2kHz and 4kHz) are visible in both Linear and Log views, hence there are 17 separate handles in total. The shape of the profile as modified by the bias handles is preserved while switching between views.
current bias display are not at their null position. A small red or yellow dot appears on the Reset Bias button to indicate that the non-selected bias curve is non-null. 5.4 DeHisser Section The Oxford DeNoiser also contains a dedicated DeHisser section. This DeHisser is not as musical as the de-hissing that can be achieved with the de-noiser algorithms, but it does allow more aggressive reduction of hiss levels.
Left and Right channels to a bus called Mid, and sending the Left and the phase inverted version of the Right to a bus called Side.
Input and Output Sections “Meters” – The input and output meters indicate 1dB per segment for the top 18dB of dynamics range, and 2dB per segment thereafter. “Trim” – The input trim has a range of +12dB to -12dB. The output trim has a range of 0.0dB to -12dB “Listen Audio” – Selects auditioning of the cleaned audio signal. “Listen Diff” – Selects auditioning of what is being removed from the audio signal. “Warmth” – Can quickly add warmth into the signal if desired.
“Tune” – In some situations can help to remove musical noise. “Auto” Mode - Enables automatic updating of noise profile and the noise profile level relative to the input signal. “Freeze” Mode – Once you have obtained a good profile, this button will freeze the profile to prevent it updating. “Manual” Mode – In this mode, the profile is forced flat. You can then use the Colour, Air or bias controls to modify the shape of the profile. “Colour” – Modifies the LF shape of the noise profile.
“Reset Biases” – When a bias handle is moved this button lights half-bright to indicate that the bias curve is non-null. A single press nulls the bias control handles for the currently active and visible bias curve; using shift-click will reset the bias handles for both the LF and HF sections of the currently active bias curve. A small red or yellow dot indicates that the non-selected bias curve is non-null. “Log /Lin Select” – Switches the display between logarithmic (LF) and linear (HF) modes.
5.8 DeNoiser Signal / Control Flow Diagram 5.9 DeNoiser Stereo Behaviour For correct stereo operation, when instantiated into a stereo track the DeNoiser has two independent repair circuits. In Manual mode, there is only one noise profile that is applied to both stereo channels. In Freeze mode, there are two separate noise profiles, and both are displayed in the graph area. In Auto mode, a single combined left and right noise profile is derived and displayed.
6. Preset Manager The Sonnox Restore plug-ins come equipped with their own onboard Preset Manager, which is displayed at the top of the plug-in window as if the host created it. The reasoning behind this is to allow increased portability of your presets across all the host applications, while also providing a consistent and versatile interface. While most host platforms allow creation and loading of presets, those host-created preset files are not portable between different host applications.
Internal Buffer Sizes: Additionally, all three plug-ins use internal buffer sizes of 1024 for 44.1kHz and 48kHz operation and 2048 for 88.2kHz and 96kHz operation. We advise that the buffer sizes of the sound card, and thus of the host application, are set to match these figures. If the buffer sizes are not matched there will be an in-efficiency in the data transfer, and in some applications this leads to uneven CPU behaviour.
Width of the Frequency Detect window: Narrow Normal Wide LF 3% 12% 20% MF 2% 8% 12% Resolution of Frequency Adjustment: Normal: 1.0 Hz Fine: 0.01 Hz Shift-fine: 0.0001 Hz Bandwidth Range: Sharp/Q: Attenuation: Tone: 75 Hz to 20kHz 16 to 160 0dB to -144dB -96dB to -6dB Input trim: Output trim: -12dB to +12dB -12dB to 0dB 7.
8. Copyright and acknowledgements The Sonnox Restore suite was developed by the design team at Sonnox Ltd., which includes Nicolas Haynes, Nick Wiggins, Eric Smith, Josep Sola, Marcello Gagliardi, Matt Morton, Luke Snape and Rod Densham. Caroline Dempster, Nathan Eames and Adam Gillicker also added invaluable help to bring the product to market. Trademarks and content copyright © 2009-Present Sonnox Ltd. All rights reserved.
Platform Specific Supplement S1. Supported Platforms Digidesign ProTools (LE, RTAS, M-Powered) VST Native Audio Units Native Mac OSX 10.4 or later (PowerPC and Intel) Windows XP S2. System Requirements For latest System requirements, please see www.sonnoxplugins.com Pro Tools • Windows XP or Apple Mac OSX 10.4 or higher • Approved Digidesign CPU and hardware configuration • Pro Tools LE, M-Powered or HD v7.2 or above. • iLok USB key, loaded with the appropriate authorisations.
