bapsf_interferometer

Acquisition, analysis, storage, and plotting for the BaPSF microwave interferometers (ports 20, 29, and 40).

Branches: main is the legacy Raspberry Pi (Linux) version, unused since 2024-07-16. Use diagnostic-pc — it runs on both PC and RP and is current.

How it works

1. Raw interferometer signals feed three scopes:
   • LeCroy — ports 20 (282 GHz) and 29 (288 GHz). Runs in "Save Waveform → Wrap" mode, continuously writing traces to a folder shared on DAQ NET.
   • Rigol DHO (192.168.7.63) — port 40 (288 GHz). Runs in AUTO trigger; traces are read in-memory over telnet.
2. The diagnostic PC (interf_main.py) detects each new LeCroy .trc, pulls the matching Rigol shot, computes phase shifts (interf_raw.py), and appends to a daily HDF5 file (interf_file.py).
3. Old scope traces are deleted to keep disks clean (interf_cleanup.py).
4. A Raspberry Pi in the main lab runs a live plotting GUI (interf_GUI.py) by reading the shared HDF5.
5. After a run, interferometer data is merged into the datarun HDF5 (interf_merge_datarun.py).

Files

Core

File	Purpose
interf_main.py	Diagnostic-PC entry point: monitor LeCroy, read Rigol, analyze, save
interf_raw.py	Phase extraction (cross-correlation in use; Hilbert available but slower) and density calibration
interf_file.py	HDF5 schema, writes, and metadata
interf_read.py	Read phase/time arrays by date and timestamp; example plotting

Support

File	Purpose
interf_GUI.py	Live density plots on the lab Raspberry Pi
interf_plot.py	Shared plotting helpers and styles
interf_cleanup.py	Delete processed scope traces; log activity
interf_merge_datarun.py	Merge interferometer data into datarun HDF5 (timestamp matching, auto-detection of channels, per-channel and per-shot attrs)

Utilities

File	Purpose
read_scope_data.py	Read LeCroy .trc (binary) and .txt (ASCII) files
LeCroy_Scope_Header.py	Decode LeCroy binary headers; build time arrays
read_hdf5.py	Read LAPD datarun HDF5 via bapsflib (probe motion, digitizer signals, run sequence)
cpu_temp.py	Raspberry Pi CPU temperature monitor

Data structure

Standalone interferometer file: interferometer_data_YYYY-MM-DD.hdf5

Root attributes: created, description.

Groups (each contains datasets keyed by timestamp):

Group	Description	unit	Per-group attrs
phase_p20/	Phase, port 20	rad	Microwave frequency (Hz) = 288e9, calibration factor (m^-3/rad)
phase_p29/	Phase, port 29	rad	Microwave frequency (Hz) = 282e9, calibration factor (m^-3/rad)
phase_p40/	Phase, port 40 (Rigol; resampled onto LeCroy time grid)	rad	Microwave frequency (Hz) = 288e9, calibration factor (m^-3/rad)
time_array/	Time base for LeCroy channels	ms	—
time_array_p40/	Time base for Rigol (currently identical to time_array, kept separate so they can diverge in the future)	ms	—

phase_p40 datasets carry per-shot rigol_missing (bool) and rigol_missing_reason (str) when the Rigol was unreachable; the array is then a zero-filled placeholder.

Timestamps mark when the scope saved the trace's last channel (C4), so there can be a small delay vs. the actual shot.

Datarun file after merge

Merged data lives under diagnostics/interferometer/:

diagnostics/interferometer/
├── phase_p20/{shot_number}     # rad
├── phase_p29/{shot_number}     # rad
├── phase_p40/{shot_number}     # rad   (new-format files only)
├── time_array/{shot_number}    # ms    (LeCroy)
└── time_array_p40/{shot_number}# ms    (Rigol; new-format files only)

Each subgroup carries the same per-channel attrs as in the standalone file; calibration factors assume a 40 cm plasma path. phase_p40 datasets preserve the rigol_missing / rigol_missing_reason per-shot attrs.

Example — read shot 5:

• diagnostics/interferometer/phase_p20/5
• diagnostics/interferometer/phase_p29/5
• diagnostics/interferometer/phase_p40/5 (if present)
• diagnostics/interferometer/time_array/5
• diagnostics/interferometer/time_array_p40/5 (if present)

Naming: datasets are keyed by shot number starting at 1. A missing shot number means no interferometer data for that shot. For new-format files, a shot may have phase_p20/phase_p29/time_array but no phase_p40/time_array_p40 if the Rigol was down.

Backward compatibility: legacy interferometer files (pre-port-40) only have phase_p20, phase_p29, time_array. The merger auto-detects which subgroups exist and only creates those, so merging an old file produces the original three-group layout.

Merge process

interf_merge_datarun.py:

1. Pulls timestamps for completed shots from the datarun file.

2. Matches them to interferometer timestamps (±1 s tolerance).

3. Copies matched data into the structure above, preserving per-channel group attrs and per-shot dataset attrs (e.g., rigol_missing).

4. Defaults to merging only the first and last completed shot as a fast sanity check that keeps the file small. Pass all_shots=True to merge every shot:

   merge_interferometer_data(datarun_path, interf_path, all_shots=True)

5. The datarun file is opened and closed once per shot, so an interruption leaves all already-merged shots safely flushed to disk.

Update 2026-04-21 — port 40 added

• Third interferometer at port 40 (288 GHz, 40 cm plasma path) on a Rigol DHO at 192.168.7.63 (CH1 = ref, CH2 = plasma).
• Per-shot flow in interf_main.py: detect LeCroy .trc → send :STOP to Rigol → read both Rigol channels over telnet → send :RUN → read all four LeCroy channels via the multiprocessing pool → run all three phase_from_raw analyses in parallel.
• Schema added phase_p40 and time_array_p40. The Rigol phase is bounds/sanity-checked, then resampled onto the LeCroy grid via np.interp, so all phase traces and both time arrays share the same length per shot.
• If the Rigol is unreachable or errors out, the LeCroy pipeline keeps writing; phase_p40 becomes a zero-filled placeholder with rigol_missing=True and a reason string. Reconnect retries every 100 shots.
• interf_GUI.py and interf_plot.py show a third (P40) trace; older HDF5 files without phase_p40 still load.
• Added _worker_init() as the multiprocessing pool initializer to suppress SIGINT in workers. On Windows, Ctrl-C is broadcast to every console process; without this, workers raised KeyboardInterrupt mid-task and stalled pool.join(). The main process keeps full Ctrl-C handling.