[hongquanli]

Adds Squid support for the Cephla laser engine v2: a multi-line engine (carrier board rev 1, Teensy 4.1 firmware, text protocol over USB) in two variants, the 400 µm fiber standard engine and the 50 µm Dragonfly engine, which adds a 560 nm fiber laser behind an AOM. The legacy v1 engine (control/squid_laser_engine.py) keeps working unchanged.

Configuration

The machine .ini selects the engine with one key and gives its USB serial number with another, for every version:

[GENERAL]
laser_engine = v2                   # v1 | v2 | blank = none
laser_engine_sn = <Teensy USB serial>

use_squid_laser_engine / squid_laser_engine_sn are still read (as v1, with one warning to switch); use_squid_laser_engine = True together with laser_engine = v2 is an error. Full setup: software/docs/laser-engine.md.

How Squid drives the engine

• One LightSource (LaserEngineV2LightSource): intensity over USB (per-line set-points, linear in the line's drive; calibration to % of power comes in a follow-up PR), exposure on the Squid controller TTL lines D1–D5 cabled to the engine's TTL1–5. Software shutter mode is refused: a software gate would leave a line emitting between exposures.
• Readiness: lines are armed and enabled on use, and an acquisition waits until its lines are ready (TEC in window, set-point reached). A background poll is the engine's host heartbeat and reports faults, interlocks and a lost link.
• MultiPointWorker tells the engine which channels an acquisition uses (note_use; v1 has none), so the 560 idle-off cannot trip mid-run.
• Small changes outside the new files: lighting.py maps 560 nm to D3 by default and adds LightSourceType.CephlaLaserEngineV2; gui_hcs.py and microscope.py select on LASER_ENGINE; the v1 engine gains an on_startup() hook that does what prepare_for_use did before (wake every channel).

Dragonfly 560 nm line

• The operator sets the 560 laser power (mW) in the Laser Engine tab; the power and the idle-off time are kept in cache/laser_engine_v2.yaml, not in the machine .ini. A simulated session never saves them.
• Squid's 560 intensity drives the AOM amplitude (engine line 3 analog, 0–5 V): linear in AOM volts. 0 % is always 0 V.
• The engine's fail-safe shutter is used for safety only: it opens when line 3 is ready, stays open through the line-3 ramp of an intensity change, and is closed in every other state and before any source restart. It never moves per exposure; the AOM's digital input does exposure timing.
• The 560 source driver is an optional module, control.laser_engine_v2_560_driver, supplied separately and not part of this repository. Without it the 560 line reports "not configured" and the other lines work normally.

UI and tools

• Laser Engine tab: engine state, arm / disarm / fault reset / wake / sleep, recent notices, a per-line table (wavelength, state, set-point, max, reason) and the 560 box. It is compact and scrolls instead of squashing when the side panel is short.
• Simulator for --simulation: a fake engine and a fake 560 source.
• Bench app: python3 tools/laser_engine_v2_bench.py (connection, service panel, the same tab), with a simulate option.

Testing

• 9 new test files, 179 tests (tests/control/test_laser_engine_v2*.py), each file in its own process: all pass. test_squid_laser_engine.py (v1) passes unchanged.
• Full suite against master 37d2e00, file by file: no test changes outcome. The failures already on master (test_MultiPointController 4, test_laser_af_settings 4, widgets_fluidics/test_protocol_tab 2, squid/test_stage 1) are unchanged.
• test_HighContentScreeningGui.py with a display: 7/7, as on master.
• Squid started in --simulation with laser_engine = v2: the engine connects and every line comes up ready.
• Not yet: the first full run on the Dragonfly hardware (560 power setting, shutter on key-off / sleep / disarm, AOM volts on GATED3). The engine itself has been brought up on the bench with its firmware tools.

38 commits including review rounds; squash-merging is suggested.

🤖 Generated with Claude Code