[Purple10101]

What

The RSPduo's DAB and RF notch filters (capture.device.dabNotch, rfNotch) are now derived by config-merger from the final capture.fc and capture.fs on every merge, instead of being user settings.

Why

default.yml shipped both notches on, and the first-boot copy put that true/true into every node's user.yml. A node tuned inside a notch's stopband has been filtering out its own illuminator. The notches are a function of the frequency, so they should follow it.

Change

• derive_notches() runs on the merged config after forced.yml. A notch is on unless [fc - fs/2, fc + fs/2] overlaps its stopband: RF (FM) 77-115 MHz, DAB 155-235 MHz. The result always overwrites the merged value, so the persisted true/true needs no LEGACY_* migration.
• A missing or non-numeric fc/fs leaves the merged values alone.
• default.yml keeps both keys, commented as derived. Standalone (cp default.yml config.yml) and staging/merge-config.sh don't run the merger, and blah2 aborts on a missing key.
• At the shipped 503 MHz both notches stay on, so UHF nodes see no change.

Rollout note

A node whose fc is inside a stopband (for example VHF-high DTV inside the DAB band) has that notch turned off, so its illuminator arrives 20-30 dB stronger. Gain set with the notch on may overload the front end. That node's gain needs lowering, or Auto-Calibrate run again, after this lands.

Known limitation: blah2 applies the notches only at sdrplay_api_Init(). A live retune (POST /capture/retune, used by Auto-Calibrate to probe candidate towers) changes fc but not the notches. The applied result is correct because /calibrate/apply re-runs the merger and restarts blah2.

Testing

• 8 new tests in config-merger/test/test_merge_config.py: UHF, FM band, DAB band, band edges using fs, user.yml overwritten, forced.yml fc, missing and non-numeric fc, no capture section. pytest config-merger/test/ scripts/test/: 85 passed. ruff check . and the vulture pre-commit hook pass.
• On a test node tuned inside the DAB band: a merger image with this script, run through retina-gui's Apply, derived rfNotch=true, dabNotch=false. The value reached config.yml and blah2 (its startup log printed dab_notch_fg: false). tar1090.env and user.yml were unchanged. That node's gain was set for the notched signal, and it overloaded as described above. It was restored afterwards.
• Not verified: SNR on a node tuned inside a notch band after recalibrating with the notch off.

Docs

README.md (new "RF notch filters" section, including the overload note and the live-retune limitation), STANDALONE.md, staging/README.md, the staging/merge-config.sh header, default.yml comments and the merge_config.py docstrings.

After this

The retina-gui PR removes the two checkboxes from the Config page. The blah2-arm PR corrects the rspduo README.

🤖 Generated with Claude Code