Information Theory of the Pipeline

Information theory sets the fabric's limits: how fast to sample, how much to compress, and how much a diagnostic actually tells you.

The pipeline as an information channel

The fabric can be read as an information channel from plasma state to feature. Information theory bounds what it can do: the sampling theorem sets the minimum rate, channel capacity bounds the bits a noisy diagnostic can convey, and entropy sets the floor for lossless compression. Designing the fabric is, in part, budgeting information.

Capacity and noise

A diagnostic channel with bandwidth B and signal-to-noise ratio S/N has a capacity C = B*log2(1+S/N) bits per second. This is why the fabric fights noise at the front end and sizes the ADC so quantization is not the limit: every decibel of S/N is capacity that would otherwise be lost, and no downstream cleverness can recover information the channel never carried.

import math
def capacity_bps(bandwidth_hz, snr_linear):
    # Shannon-Hartley capacity of a diagnostic channel
    return bandwidth_hz * math.log2(1 + snr_linear)

# a 500 kHz magnetics channel at 40 dB SNR
capacity_bps(500e3, 10**(40/10))   # ~6.6 Mbit/s

Compression floor

The entropy of a channel's samples sets the smallest lossless representation. The fabric's lossless coding approaches this floor; bounded-loss compression is only applied where the discarded content is below the noise floor and therefore carries no information. This makes compression decisions principled rather than arbitrary.

Choosing what to measure

Information theory also guides which diagnostics add value: a channel that is highly redundant with others adds little information for its cost (see mutual-information selection). Applied across the breeder and burner constellations, this keeps the fabric information-rich and waste-poor. It is an analysis discipline for machines not yet built.