Module traits 

One verb trait per role in the chain.

Methods taking &self are pure; &mut self signals internal state (memory, adaptation). Every trait has an _into method writing into a caller-owned buffer, which it resizes and fills in place (never replaces), so that a loop passing the same buffer allocates no output once its shape is stable; scratch buffers may still allocate (docs/adr/0011-into-reuses-the-output.md). All but Precode and Detect also have an allocating method that calls it. Randomness is always explicit: a random block receives &mut SimRng.

The waveform traits are generic over the symbols they carry, Symbols by default: a waveform of complex symbols (OFDM) implements Modulate, one that carries the index of a symbol (chirp spread spectrum, continuous-phase modulation) implements Modulate<Labels> (docs/adr/0012-index-symbol-waveforms.md). Demap takes the a priori log-probabilities of the points and the Jacobian logarithm through Demapping, uniform and exact by default.

Trait	Role	Input → output
Encode / Decode	channel coding	Bits → Bits, Llr → Bits
Map / Demap	constellation mapping	Bits → Symbols → Llr
Modulate / Demodulate	waveform	Symbols or Labels → Signal → Symbols or Labels
Precode / Detect	MIMO	Symbols → Symbols
Propagate	propagation medium	Signal → Signal
Impair	component non-ideality	Signal → Signal

Examples

use comnum_core::prelude::*;

/// A noiseless gain, the smallest `Impair`.
struct Gain(f64);

impl Impair for Gain {
    fn impair_into(&mut self, x: &Signal, _: &mut SimRng, out: &mut Signal) -> Result<()> {
        out.resize(x.n_streams(), x.len(), x.sample_rate());
        out.as_array_mut().assign(&(x.as_array() * self.0));
        Ok(())
    }
}

let x = Signal::from_stream(vec![C64::new(1.0, 0.0); 4], Hertz(1.0))?;
let y = Gain(2.0).impair(&x, &mut batch_rng(1, 0))?;
assert_eq!(y.stream(0)[3], C64::new(2.0, 0.0));

Structs

Demapping

The a priori information and the approximation of a soft demapper (Demap::demap_with_into). With the log-priors $\lambda_{x} = \ln P(x)$ of the points $x$ of the alphabet and the variance $\sigma^{2}$ of the complex noise, the LLR of a bit of the received symbol $y$ is

Enums

Jacobian

The Jacobian logarithm of the log-MAP algorithms and of the soft demappers, exact or approximated by its maximum:

Traits

Decode

Channel decoding of log-likelihood ratios into information bits.

Demap

Soft demapping of received symbols into log-likelihood ratios.

Demodulate

Demodulation of a waveform back to symbols: complex Symbols by default, or the decided Labels of a waveform that carries the index of a symbol (Demodulate<Labels>).

Detect

Detection of transmitted symbols from received streams.

Encode

Channel encoding of information bits into codewords.

Impair

A component non-ideality (amplifier, oscillator, converter…).

Map

Mapping of bits onto constellation symbols.

Modulate

Modulation of symbols onto a waveform: complex Symbols by default, or the Labels of a waveform that carries the index of a symbol (Modulate<Labels>).

Precode

MIMO precoding of streams of symbols.

Propagate

Propagation through a medium. Randomness is always explicit.