User manual SONY OXFORD RESTORATION TOOLS

[. . . ] Oxford Restoration Tools Plug-In Manual For Digidesign ProTools RTAS 1 Introduction The Oxford Restoration Tools plug-in is a collection of three applications designed to improve the sound of digital audio recordings which are impaired by clicks, noise and buzz. Recordings with moderate damage can be restored with no trace of the disturbances left. Even severely damaged recordings which would otherwise be unlistenable can be salvaged. The aim of the Oxford Restoration Tools is to provide quick, high quality restoration with little intervention, but also to provide the more experienced operator with enough control for optimum results. 2 System Requirements Approved Digidesign CPU and hardware configuration Windows XP or Mac OS 10. 2. 6 or later 800x600 minimum display Pro Tools 6. 1 or later ILok USB Smart Key 3 Principles of Audio Restoration 3. 1 Science and Art Audio restoration is both a science and an art. [. . . ] In the multichannel (stereo) instantiation of DeNoise, the left-channel and right-channel automatic fingerprints are linked together by the algorithm. The graphical display will therefore only display a single fingerprint. The reason for this is to preserve stereo separation and stability, especially when de-noising dialogue with voices positioned at various locations around a stereo microphone. However, as soon as the Freeze button is pressed, the linking is removed, and two separate noise fingerprints are displayed, as for a conventional noise spectrum snapshot. The Level and Smooth controls operate exactly as described for fingerprint capture mode. Automatic mode works best with Smooth set at or near maximum; if set too low, the automatic fingerprint simply tends to follow the shape of the observed spectrum. The Correlation control is provided in Automatic mode as a means for fine-tuning the way the automatic noise fingerprint is calculated. It controls the amount of correlation between the observed audio and the automatic fingerprint, particularly in the presence of low-level signals. By decreasing the Correlation, the fingerprint will remain steadier in the presence of low-level signals, providing fewer artifacts with decaying notes and intakes of breath, for example. The risk is that there is an increased chance of noisepumping, in which the noise becomes audible during loud passages. Conversely, if you are experiencing noise-pumping, try increasing the Correlation. The effect of the Correlation control is somewhat signal-dependent, and it can have either a large or subtle effect. 5. 3. 2 Noise Removal The Attenuation fader is the primary means for controlling how much noise reduction occurs. At the top, no noise is removed, and as you bring it down, increasingly more noise reduction occurs. The greater the amount of noise reduction, the greater the risk of artifacts. The most common and unacceptable artifact is the presence of tinkly random tones, which we call musical noise. 5. 3. 2. 1 Musical Noise Elimination There are three ways to reduce or eliminate musical noise artifacts: · · The first method is to raise the Level fader, but this carries the risk of dulling the sound because there is less vertical separation between the noise fingerprint and the signal. The second method is the reduce the amount of noise reduction, by raising the Attenuation fader towards 0dB, but this simply masks the musical noise. · The third method is to increase the Soft control setting. The action of this control is multifaceted. Between the minimum (fully anticlockwise) and halfway positions it acts to eliminate musical noise with very little dulling of the signal. Between halfway and maximum (fully clockwise), it eliminates even more musical noise, but also causes the noise gates at each frequency to act in a softer, more forgiving manner, which is particularly suited to tackling loud noise and dialogue. . The figure below shows how musical noise occurs. The blue spectrum is obtained from some white noise. The red line shows the optimum noise fingerprint when the Soft control is at minimum, because it lies just above all the random peaks in the noise spectrum. Keeping the Soft control at minimum, suppose we now lower the Level fader so that the noise fingerprint is depicted by the green line. [. . . ] If you know the exact fundamental frequency of the buzz, you can alternatively enter Freeze mode and simply type the frequency into the numeric scribble area of the Frequency fader. The high-resolution spectrum display shows the presence of a detected buzz; the higher the peak, the stronger the buzz. The confidence with which a buzz has been detected is indicated by the length of the white bar at the top of the display. Beneath that bar, the frequency of the buzz removal filter currently being applied is displayed. [. . . ]