Psycho Dither readme

schwa · September 12, 2007, 1:52am

PsychoDither
(God’s Own Dither)
© 2007 Schwa

Quick Start.

Use Psycho Dither as the very last effect on the master channel when rendering to a lower bit depth. If your audio host does not provide post-fader effect slots, insert Psycho Dither as the last effect before the master fader, leave the master fader at zero, and do not move it at all, or else Psycho Dither’s work will be undone.

To open this readme from the plugin, click the Schwa logo.

Basics.

A digital audio signal consists of a series of numbers, or samples. At the most basic level, these samples represent output voltage to the speakers. When the output voltage is +1, the speaker cone is at its “maximum out” position; -1 represents the “maximum in” position. The samples vary at a fast rate, causing the speaker position to vary, which reproduces the original audio.

Typical CD audio contains a sample rate of 44,100 samples per second. Your digital audio project may have a different sample rate. Sample rate conversion at render time is usually handled by the audio host software, and won’t be discussed further here.

Typical CD audio contains samples with a depth of 16 bits, meaning that each sample must be in one of 2^16 = about 65,000 possible positions between maximum in and maximum out. However, audio recording software usually uses a much higher internal bit depth (32-bit float is most common). If you render your audio project from a higher to a lower bit depth without using a bit reducer such as Psycho Dither, the extra resolution will simply be thrown away, which introduces distinctive ugly-sounding low-level audio artifacts known as truncation noise.

Truncation noise by definition consists of information that has been taken away, so there is no way to eliminate it. But there are ways to mitigate it.

Psychoacoustic Noise Shaping.

Because of the physical structure of the human ear, we are more sensitive to some frequencies and less sensitive to others. By sending truncation noise into a carefully managed high order feedback loop, the noise can be filtered into exactly those frequencies that are less perceptible, causing the apparent loudness of the noise to be greatly reduced or even eliminated.

Dither.

Another strategy for dealing with truncation noise is simply to blanket it with white noise. Dithering is randomly changing the last bit or two of each truncated sample in order to create a layer of white noise, or hiss. Dithering adds noise power, but white noise catches the ear much less than the truncation noise it covers up.

Adaptive Adjustments.

The state of the art in noise shaping dither is to randomly dither the last two bits of the truncated audio to create a white noise mask, then use precalculated fixed coefficients to shape that white noise into predetermined parts of the frequency spectrum. Psycho Dither’s noise reduction strategy is different: two separate adaptive adjustments are made during audio playback, so that the truncation noise can be actively managed in relation to the source audio to preserve the maximum possible
dynamic range with the minimum added noise power.

First, the strength and distribution of the dither noise itself is adjusted as the source audio signal changes, decreasing the dither strength in passages where the source audio contains noise-like characteristics and increasing the dither strength in passages where the truncation noise would otherwise be most audible.

Second, using a completely different approach from the standard precalculated coefficients, the noise shaping filter is constantly recalculated and adjusted in relation to the actual frequency profile of the source audio. This allows Psycho Dither to actively manage the noise shaping filter to avoid pushing noise into empty and audible parts of the spectrum.

Controls.

Dither Color: Select red or green. Red dither adds the minimum noise to the signal which will sound “hard” or “empty.” When used with high psycho power, red dither is the best choice for most types of audio material. Green dither adds more dither noise, which will sound “softer,” and is useful when it’s appropriate for the material to sound more analog. Green dither is similar to faint tape hiss when amplified. For a typical render to 16-bit CD audio, dither noise is below -60 dB, which is faint indeed.

Psycho Power: Select the degree, or order, of the noise shaping filter. You will see the black noise graph change shape as the psycho power changes. Higher order noise shaping filters move input noise more effectively into inaudible parts of the spectrum, but also add more noise energy to the output.

Target Bit Depth: Select the desired bit depth of the output audio.

Tips.

To choose the best settings for your particular audio material, try dialing the target bit depth way down below the actual render bit depth, so that the bit reduction noise, dither noise, and psychoacoustic noise shaping are all clearly audible. Once you dial in the settings that sound most appropriate to the material, dial the target bit depth back up to the desired number.

Do not use any bit reducer in front of a 32-bit plugin in a 64-bit signal path. This is a floating point conversion, and using a bit reducer will only add noise with no benefit.

If you have any support needs, or questions on usage or anything else regarding this plugin, please visit the forums at stillwellaudio.com. Woo hoo!