Morty0311/battery-tests-cfo-analysis

Battery-test CFO analysis

This folder contains the all-packet CFO fingerprints, statistics, and plots for all ten Morty battery experiments (10% through 100%). Source H5 recordings were read without modification. The estimator reuses CSE237D_weyl\pipeline\parallel_h5_cfo_all_packets.py.

Start with summary\all_packets\ALL_PACKETS_RESULTS.md for the consolidated index. Each experiment has:

• —all_packets_cfo\packet_cfo_all_fingerprints.csv — every assigned packet.
• —all_packets_cfo\statistics — phone median, variance, MAD, percentiles, and statistical plots.
• —all_packets_cfo\plots_segmented_mac — reference-style PNG and zoomable SVG.

The segmented plots use at most 2,000 evenly selected points per phone for readability; all CSV summaries and statistics still use every packet. PNG and SVG use the same sampled rows and similar small marker sizes. Every continuous x-axis is fixed at 0--40 minutes, so shorter recordings leave blank space. --shared-y means every phone panel has the same median-centered +/- X Hz range, while preserving each phone's distinct CFO center.

To rerun levels 20–100 with the satellite Conda environment:

C:\Users\15255\anaconda3\envs\satellite\python.exe .\run_remaining_battery_all_packets.py

To rebuild and validate the consolidated index without recomputing CFO:

C:\Users\15255\anaconda3\envs\satellite\python.exe .\build_all_packet_battery_summary.py

The preferred fingerprint is relative_cfo_hz: raw packet CFO minus the simultaneous cross-device median in a 120-second window. This reduces common receiver-oscillator drift.

Raw and EMA-calibrated plots

summary\raw_ema\RAW_EMA_RESULTS.md indexes raw and EMA-calibrated PNG/SVG plots for every level. The EMA workflow uses 10-second robust cross-device center observations and a causal 120-second EMA span, avoiding the hard two-minute boundaries in the original block calibration.

To regenerate raw and EMA outputs for all levels:

C:\Users\15255\anaconda3\envs\satellite\python.exe .\run_all_levels_raw_ema_plots.py

To validate and rebuild their consolidated index:

C:\Users\15255\anaconda3\envs\satellite\python.exe .\build_raw_ema_index.py

Smoother offline calibrations

summary\smoothers\SMOOTHER_RESULTS.md indexes centered 120-second moving average and Gaussian (sigma=60 s, truncated at +/-3 sigma) calibration plots for every level. These offline filters reduce the influence of a single new observation compared with the causal EMA.

C:\Users\15255\anaconda3\envs\satellite\python.exe .\run_all_levels_smoother_plots.py
C:\Users\15255\anaconda3\envs\satellite\python.exe .\build_smoother_index.py