Module dither

The dither module provides methods to apply a dither to a sample.

Dithering is the process of adding noise to the least significant digits of a sample before down-converting (quantizing) it to a smaller sample type. The purpose of dithering is to decorrelate the quantization error of the down-conversion from the source signal.

Dithering is only applied on lossy conversions. Therefore the dither module will only apply a dither to the following down-conversions:

• { i32, u32 } to { i24, u24, i16, u16, i8, u8 }
• { i24, u24 } to { i16, u16, i8, u8 }
• { i16, u16 } to { i8, u8 }

Multiple dithering algorithms are provided, each drawing noise from a different probability distribution. In addition to different distributions, a dithering algorithm may also shape the noise such that the bulk of the noise is placed in an inaudible frequency range.

Structs

Identity

The Identity dithering algorithm performs no dithering and returns the original sample.

RandomNoise

RandomNoise represents a sample of noise of a specified length in bits.

Rectangular

Rectangular implements a dither using uniformly distributed (white) noise without shaping.

Triangular

Triangular implements a dither using a triangular distribution of noise without shaping.

Enums

DitherType

Enumeration of dither algorithms.

Traits

AddNoise

AddNoise is a trait for converting random noise into a Sample.

Dither

Dither is a trait for implementing dithering algorithms.

MaybeDither

MaybeDither conditionally applies a dither to a sample depending on the source and destination sample types.