Trait spectrusty::audio::Blep

source · 
pub trait Blep {
    type SampleDelta: SampleDelta;

    fn ensure_frame_time(
        &mut self,
        sample_rate: u32,
        ts_rate: f64,
        frame_ts: i32,
        margin_ts: i32
    );
    fn add_step(
        &mut self,
        channel: usize,
        timestamp: i32,
        delta: Self::SampleDelta
    );
    fn end_frame(&mut self, timestamp: i32) -> usize;
}
Expand description

A trait for interfacing Bandwidth-Limited Pulse Buffer implementations by square-wave audio generators.

The perfect square wave can be represented as an infinite sum of sinusoidal waves. The problem is that the frequency of those waves tends to infinity. The digitalization of sound is limited by the finite sample frequency and the maximum frequency that can be sampled is called the Nyquist frequency.

When sampling of a square wave is naively implemented it produces a perceptible, unpleasant aliasing noise.

Square waves that sound “clear” should be constructed from a limited number of sinusoidal waves, but the computation of such a wave could be costly.

However thanks to the Hard Sync technique it is not necessary. Instead, a precomputed pattern is being applied to each “pulse step”. The implementation of this is called the Bandwidth-Limited Pulse Buffer in short Blimp or Blep.

The audio stream is being produced in frames by the Blep implementation. First, the pulse steps are being added, then the frame is being finalized and after that, the audio samples can be generated from the collected pulses.

The way audio samples are being generated is outside of the scope of the Blep implementation. This trait only defines an interface for adding pulse steps and finalizing frames.

Required Associated Types§

A type for sample ∆ amplitudes (pulse height).

Required Methods§

This method allows the Blep implementation to reserve enough memory for the audio frame with an additional margin and to set up a sample time rate and other internals.

This method should not be called again unless any of the provided parameters changes.

  • sample_rate is a number of output audio samples per second.
  • ts_rate is a number of time units per second which are being used as input time stamps.
  • frame_ts is a duration of a single frame measured in time units specified with ts_rate.
  • margin_ts specifies a largest required margin in time units for frame duration fluctuations.

Each frame’s duration may fluctuate randomly as long as it’s not constantly growing nor shrinking and on average equals to frame_ts.

Specifically frame_ts + margin_ts - frame start specifies the largest value of a time stamp that can be passed to Blep::add_step or Blep::end_frame. frame_ts - margin_ts - frame start specifies the smallest value of a time stamp that can be passed to Blep::end_frame.

The smallest possible time stamp value that can be passed to Blep::add_step is a frame start time stamp. It starts at 0 after calling this method and is modified by the timestamp value passed to the last Blep::end_frame to timestamp - frame_ts. In other words, the next frame starts when the previous ends minus frame duration.

This method is being used to add square-wave pulse steps within a boundary of a single frame.

  • timestamp specifies the time stamp of the pulse.
  • delta specifies the pulse height (∆ amplitude).

The implementation may panic if timestamp boundary limits are not uphold.

Finalizes audio frame.

Some frames can end little late or earlier and this method should allow for such a flexibility.

timestamp specifies when the frame should be finalized marking the timestamp of the next frame.

Returns the number of samples that will be produced, single channel wise.

The caller must ensure that no pulse step should be generated with a time stamp past timstamp given here.

The implementation may panic if this requirement is not uphold.

Implementors§