Part 1: Data stream parameters (Pre-mix vs Final)
These metrics compare the raw pre-mix signal with the final stream after cryptographic mixing.
Entropy (bits/byte)
Definition: Shannon entropy estimates the average amount of unpredictable information carried by one byte. For byte streams it is computed from the probabilities of values 0-255.
Target: 8.000000 bits/byte. This means full practical unpredictability: all 256 byte values are equally likely.
Anomaly: Low pre-mix entropy means the antenna signal contains structure: dominant ADC ranges, interference, a carrier, clipping, or insufficient sample dynamics. Such material must be conditioned before security use.
Chi-square test
Definition: Compares the histogram of all 256 byte values with a uniform distribution and detects values that occur too often or too rarely.
Target: For large samples the statistic should oscillate around the degrees of freedom, roughly 250-256. Zero is not the goal; normal random fluctuation is expected.
Anomaly: Huge values indicate strong hardware or environmental bias: selected byte values dominate, the signal has DC offset, the band is dominated by a transmitter, or the ADC path is outside its useful range.
Serial correlation
Definition: Measures dependence between adjacent samples: value n and value n+1. It shows whether the next byte is partially predictable from the previous one.
Target: 0.000000. Adjacent bytes in an ideal stream have no linear dependency.
Anomaly: Noticeable positive or negative values indicate temporal patterns: slow sampling, RF-path filtering, periodic interference, repeated buffered frames, or a stalled source.
Mean byte value
Definition: Arithmetic mean of byte values in the sample. For a uniform 0-255 distribution the center of mass is between 127 and 128.
Target: 127.5.
Anomaly: A lower or higher mean means the hardware favors the lower or upper half of the byte range. Common causes include voltage offset, automatic gain behavior, clipping, ADC level shift, or a strong signal component.
Bit balance (0/1)
Definition: Fraction of ones in the whole bit stream. It shows whether zeros and ones appear at similar rates independently of byte grouping.
Target: 0.500000, meaning 50% ones and 50% zeros.
Anomaly: ADCs and analog paths may naturally drift toward zeros or ones because of voltage movement, temperature, gain selection, or local interference. Pre-mix drift does not automatically disqualify a source, but it requires whitening and monitoring.
Part 2: Radio node diagnostics (SDR Health)
These metrics evaluate physical SDR receiver health before data is mixed.
Spectral flatness
Definition: Measures how similar the spectrum is to white noise by checking whether energy is spread evenly across frequencies.
Target: Values closer to 1.0 indicate a more white-noise-like signal with less dependence on one frequency.
Anomaly: Low flatness means one frequency or narrow band dominates: transmitter carrier, local interference, antenna resonance, intermodulation, or a poorly selected receive center.
Repeat margin
Definition: Distance between current source behavior and the rejection threshold for repetition. It shows how close the node is to being rejected by the sampler.
Target: A high margin relative to the threshold, for example far from 0.90. More headroom means lower risk of a stuck source.
Anomaly: A small margin means the raw stream is repeating values too often. Causes include a frozen receiver, stalled buffer, strong carrier, lost ADC dynamics, or a sampler process fault.
Dominant value ratio
Definition: Share of the most frequent value in the raw sample. It detects whether one byte value starts dominating the distribution.
Target: Very low values. In healthy noise, no single byte value should own a large part of the histogram.
Anomaly: A high value means the antenna may be receiving a stable carrier, the ADC path is shifted or saturated, the device is stuck, or the system is reading repeated data instead of dynamic noise.
Adjacent repeats
Definition: Percentage of cases where the same value appears two or more times in a row in the raw stream.
Target: Low value. Healthy physical noise should change dynamically between neighboring samples.
Anomaly: High adjacent repeats suggest stalled sampling, slow signal variation, strong filtering, duplicated buffer samples, or a physical source dominated by a stable signal instead of noise.
Node classes: high-throughput and reference nodes
BIG_BANG_ENTROPY can be read as a hybrid system. Some nodes provide high-volume raw data, while others act as low-throughput physical reference sources with stronger interpretability. Both classes are useful, but they address different risks.
1. High-throughput nodes
Role: Operational workhorses. They provide a broad continuous stream required for API traffic and fast pool replenishment.
Physical source: RF background noise, receiver thermal chaos, environmental variation, solar emissions, and broadband components.
Hardware: Broadband omnidirectional antennas such as discone or dipole systems listening to relatively empty VHF/UHF regions.
Risk: Sensitive to local interference: transmitters, industry, traction noise, ADC overload, or carrier waves. Spectral flatness, dominant value ratio, and adjacent repeats are therefore critical.
2. Reference / astrophysical nodes
Role: Low-throughput reference nodes that improve physical interpretability and help audit sources with well-described origin.
Physical source: Narrowband observations such as the neutral hydrogen 21 cm / 1420 MHz line, associated with the spin transition of atomic hydrogen in the interstellar medium.
Hardware: Directional antennas, dishes, low-noise LNAs, SAW bandpass filters, and stable receiver timing.
Risk: Low throughput, sky direction dependence, Earth rotation, antenna conditions, and calibration. Such a node complements volume; it does not replace it.
| Feature | High-throughput nodes | Reference nodes |
|---|---|---|
| Primary goal | Scale, SLA, fast pool input | Physical provenance and source auditability |
| Pre-mix throughput | High, continuous stream | Low, point or periodic observations |
| Bandwidth | MHz, broadband | kHz, narrow and precise |
| Typical risk | Local interference and hardware bias | Low volume, sky tracking, calibration |
| System contribution | Most of the data volume | Reference admixture with high audit value |