Documentation

MitraSETI

Radio technosignature detection pipeline — algorithms, architecture, and practical reference.

Saman Tabatabaeian — Deep Field Labs v0.2.0 · March 2026

INGEST RFI FILTER DE-DOPPLER CLUSTER ML + SCORE OUT Cosmic source Radio telescope Processing pipeline 10⁶ ch ~1000 hits ~10 candidates Signal flow: cosmic radio emission is collected by a dish, digitised, and passed through MitraSETI's six-stage pipeline — from millions of channels to a short ranked candidate list.

About This Reference

This documentation covers the full MitraSETI stack: the radio astronomy concepts that motivate the search, the algorithms that power it, the machine learning that prioritises candidates, and the practical tooling for hardware validation. Each chapter is self-contained and cross-referenced where relevant.

★

New? Start Here

If the chapters feel too technical, read the Simple Explanation first — it covers every major concept using everyday analogies like stadiums, coin flips, and tournament brackets. No prior knowledge required.

📰

Release Article

Read the full MitraSETI v0.2.0 & AstroLens v1.2.0 Release Article — algorithmic breakthroughs, benchmark results, and comparisons with existing SETI tools.

⬡

Suggested Reading Paths

Chapters

ARTICLE v0.2.0 Release Article Full technical article with benchmarks, algorithm comparisons, and streaming results on Breakthrough Listen data. Publication 01 Foundations of Radio Astronomy Electromagnetic spectrum, spectrograms, noise, signal-to-noise ratio, and key frequencies in radio SETI. Beginner 02 Doppler Effect and De-Doppler Search Why signals drift in frequency, brute-force integration along diagonals, and normalisation. Beginner → Intermediate 03 Taylor Tree Algorithm The O(N log N) de-Doppler engine: divide-and-conquer construction, butterfly pattern, and complexity analysis. Intermediate → Advanced 04 RFI and Spectral Kurtosis Filtering Radio interference taxonomy, statistical kurtosis, adaptive thresholds, and layered defence. Intermediate 05 Clustering and Anomaly Detection HDBSCAN density-based clustering, candidate deduplication, and out-of-distribution scoring. Intermediate 06 Machine Learning Pipeline CNN and Transformer classifiers, SimCLR pre-training, attention maps, and interestingness scoring. Intermediate → Advanced 07 Computer Science Foundations Complexity theory, divide-and-conquer, FFT, parallel computing, memory hierarchy, and Rust/Python architecture. Intermediate → Advanced 08 Pipeline Walkthrough End-to-end data flow: raw telescope file through six processing stages to ranked candidate output. Intermediate 09 Recommendations and Future Work GPU acceleration, Radio-Frequency Transformers, Bayesian detection, federated learning, and more. Advanced 10 Signal Transmission Research METI history, link budgets, modulation techniques, encoding strategies, and new transmission algorithms. Advanced / Research 11 Testing with Real Devices RTL-SDR, HackRF, walkie-talkies, IQ-to-filterbank conversion, and hardware validation experiments. Practical

Deep Field Labs · 2026