dspark::Dither< T > Class Template Reference

TPDF dithering processor with optional 1st-order noise shaping. More...

#include <Dither.h>

Public Member Functions
	Dither (int targetBits=16, bool noiseShaping=false) noexcept
	Constructs a dithering processor.
void	setTargetBitDepth (int bits) noexcept
	Sets the target bit depth and recalculates scaling factors.
void	setNoiseShaping (bool enabled) noexcept
	Enables or disables 1st-order noise shaping.
void	reset () noexcept
	Resets the noise shaping state. Call this when playback stops or flushes.
T	processSample (T input, int channel=0) noexcept
	Applies TPDF dither and quantises a single sample.
void	processBlock (T *data, int numSamples, int channel=0) noexcept
	Applies dithering to an entire audio buffer in-place.
int	getTargetBitDepth () const noexcept
T	getQuantisationStep () const noexcept

Detailed Description

template<typename T>
class dspark::Dither< T >

TPDF dithering processor with optional 1st-order noise shaping.

Adds Triangular Probability Density Function (TPDF) noise of 2 LSB peak-to-peak before quantisation. This replaces correlated quantisation distortion (truncation harmonics, noise modulation) with a constant, signal-independent noise floor.

The optional noise shaper feeds back the total requantisation error (dither + quantiser) through a one-sample delay, first-order highpass shaping the entire noise floor: about 15 dB less noise at 1 kHz (44.1 kHz rate) in exchange for a gentle rise towards Nyquist where hearing is least sensitive.

Quantised output lies exactly on the integer level grid of the target depth: levels in [-2^(bits-1), 2^(bits-1) - 1], so positive full scale is one step below 1.0, exactly like the int16/int24 file writers. Feeding the output to WavFile round-trips bit-exactly.

Threading is owner-managed: call setters and processing from the thread that owns the stream (the PRNG and error state are plain members by design). Use one instance per processing thread.

Template Parameters

T	Sample type (float or double).

Definition at line 55 of file Dither.h.

Constructor & Destructor Documentation

Dither()

template<typename T >

dspark::Dither< T >::Dither ( int  targetBits = 16, bool  noiseShaping = false  )

inlineexplicitnoexcept

Constructs a dithering processor.

Automatically seeds the internal PRNG with a globally unique seed to ensure uncorrelated noise between multiple instances (e.g., Left and Right channels).

Parameters

targetBits	Target bit depth (min 8). Max is 24 for float, 32 for double.
noiseShaping	Enable 1st-order noise shaping.

Definition at line 69 of file Dither.h.

Member Function Documentation

getQuantisationStep()

template<typename T >

T dspark::Dither< T >::getQuantisationStep ( ) const

inlinenoexcept

Definition at line 155 of file Dither.h.

getTargetBitDepth()

template<typename T >

int dspark::Dither< T >::getTargetBitDepth ( ) const

inlinenoexcept

Definition at line 154 of file Dither.h.

processBlock()

template<typename T >

void dspark::Dither< T >::processBlock ( T *  data, int  numSamples, int  channel = 0  )

inlinenoexcept

Applies dithering to an entire audio buffer in-place.

The noise shaping branch is resolved once outside the sample loop. (The PRNG recurrence is serial, so this loop does not vectorise.)

Parameters

data	Pointer to audio data.
numSamples	Number of samples to process.
channel	Channel index.

Definition at line 140 of file Dither.h.

processSample()

template<typename T >

T dspark::Dither< T >::processSample ( T  input, int  channel = 0  )

inlinenoexcept

Applies TPDF dither and quantises a single sample.

Parameters

input	Input sample in [-1.0, 1.0].
channel	Channel index for independent noise shaping history (0 to 15; out-of-range indices dither without shaping).

Returns

Quantised sample on the target grid, in [-1, 1 - step].

Definition at line 122 of file Dither.h.

reset()

template<typename T >

void dspark::Dither< T >::reset ( )

inlinenoexcept

Resets the noise shaping state. Call this when playback stops or flushes.

Definition at line 110 of file Dither.h.

setNoiseShaping()

template<typename T >

void dspark::Dither< T >::setNoiseShaping ( bool  enabled )

inlinenoexcept

Enables or disables 1st-order noise shaping.

Definition at line 107 of file Dither.h.

setTargetBitDepth()

template<typename T >

void dspark::Dither< T >::setTargetBitDepth ( int  bits )

inlinenoexcept

Sets the target bit depth and recalculates scaling factors.

Parameters

bits	Target bit depth. Automatically clamped based on the precision of T.

Definition at line 86 of file Dither.h.

---

The documentation for this class was generated from the following file:

• Core/Dither.h