Repository navigation
Conversation
… empty cache Choosing a device by `serial_number` compared it with `dev._serial_number`, which PyUSB fills only once something has read `dev.serial_number`. Straight after enumeration it is None, so every device was refused as having no serial number. The serial number is now read through the descriptor (`dev.serial_number`), and a device without one (`iSerialNumber == 0`) is still refused.
The Biomek tips (T50F, T190F, T230, T1025F, T1070), their 96-tip boxes (BC50F, BC190F, BC230, BC1025F, BC1070), plates (Thermo96VB_PP, BCFlat96, BCDeep96Square, Thomson_24w_Square_RB, with the lids Biomek seats on them) and reservoirs (ThermoNUNC_350CB, Reservoir, ModularReservoir), transcribed from Biomek 5's tip and labware classes. A tip is a `BiomekTip`, a `Tip` that also carries its cone (the radius at its end and the length over which it widens); a well is a `BiomekWell`, a `Well` that also carries Biomek's outline of it by height (`WellSegment`s), from which its height <-> volume table and its wall come. Biomek uses both to place a tip near a well's wall, as `HamiltonTip` and `TecanTip` carry their instruments' data. A tip box is a `StandingTipRack` whose tip spots sit on its top face, where the head's mandrels meet the tips. Each resource carries Biomek's grip point (`preferred_pickup_location`). A resource library page lists them with their part numbers.
`TipMountingShaft.serialize_state` holds the tool the shaft carries (as it serializes; None for none), as a tip spot's state holds its tip, and `load_state` puts it back: the saved tip goes onto the shaft (a tool of the same name held elsewhere in the tree gives way first), and a saved None takes off the one it carries. State saved before shafts had any leaves the shaft as it is. With it, `save_state_to_file` / `load_state_from_file` keep the tips a head carries across sessions, which a device that cannot sense its tips needs. Tests: a carried tip saved and read back, its own state after it; None taking a tip off; the same tip staying; a tip of that name held elsewhere giving way.
…covery The first part of a Beckman Coulter Biomek i7 driver that talks to the instrument's ACE boards directly over USB (0AB5:ACE0), one board per module: the chassis and one per pod. - `driver/framing.py`, `board.py`: length-prefixed frames packed into interrupt transfers; a board's replies matched to their requests by header, two-stage replies (ACCEPTED, DONE), unsolicited messages to subscribers. - `driver/protocol.py`, `errors.py`: every request the Master Controller was logged sending, plus three reads, and nothing else; each field layout as Biomek's `ace_io` log (Log.db3) decodes it. - `driver/master.py`, `features/chassis.py`, `features/pod.py`, `module.py`: discovery places each board by what it reports (Get Module ID: type and side), not by USB order or serial number, then reads each module's serial number, its boards with their application images and versions, and the names of its inputs, in the order of the Master Controller's cold start. It writes nothing to the boards. Unknown types or sides, two boards in one place, and no chassis are refused. - `driver/simulator.py`, `defaults/`: boards replayed from a cold-start recording; a generic one ships, every serial number replaced by a placeholder. Tests: one exact frame per command (from Biomek's `ace_io` log; three reads it never logged from their layout), build and parse round trips, framing inverses, the replay's rules, and discovery on the generic recording. Discovery sends the same requests as on the full driver this series comes from.
What the boards cannot report about an instrument, declared per instrument in one JSON file given by path (`declared_configuration_json`, now required): - `driver/configuration.py`: `DeviceConfiguration`, the motor tables (the Master Controller's instrument record: each axis's Configure parameters, home position and measured hysteresis, and the modules' serial numbers) and per pod the conversions, travel limits, geometry and settings from Biomek's instrument file, with the default deck's calibrated positions and layout. Each source's values carry a fingerprint; a file changed after import is refused at load. `anonymized()` makes a generic one: no serial numbers, no measured hysteresis. - `driver/motors.py`: the axes of each arm type and the limits read from their Configure bytes. - `units.py`: the one place PyLabRobot's units (mm, uL, mm/s) meet Biomek's (cm, its frame, the speed field). - `defaults/`: the generic configuration, one instrument's physical values with every serial number removed. - Discovery refuses a pod type other than the declared one, and on the instrument a configuration whose serial numbers are not the modules', or the generic one. Each pod gets its type's axes and its declared values. Tests: conversions give the counts Biomek's logs show for the instrument file's targets, the bridges meet where Biomek's logged moves stopped, save / load and the changed-file refusal, anonymizing, declared against found, the serial-number check on the instrument's own path, every speed field round-tripping on every Z and pump, and the motor tables' limits.
… checked against them `DeviceConfiguration.from_biomek` makes an instrument's declared configuration from its two Biomek sources, read-only, standard library only (`biomek_files/`): - `variant_tree.py`: Biomek's variant-tree files (the instrument file, `.bif`); a file with any byte it cannot account for is refused, not misread. - `instrument_file.py`: the pods (conversions, travel, geometry, settings, gripper, obstacles), the default deck's calibrated positions and layout; Biomek's runtime state in the same file is left out. - `master_controller.py`: the Master Controller's instrument record (its database's `instrument` table, or a text file of it): motor tables, home positions, measured hysteresis, the modules' serial numbers. - `importer.py`: the two put together and checked against each other, each source's path, kind, read time and fingerprint recorded; `check_sources` reads them again and lists what changed. On the instrument, discovery now also refuses a configuration whose sources have changed since it was imported (a re-framed position, a re-measured hysteresis), and warns where they cannot be found. Reading the database closes its connection. Tests: the readers on files written in the test from the generic configuration (a small variant-tree writer and a record writer, `files_tests_support.py`): every value kind and class body, the refusals, the round trip from configuration to files and back, inconsistent sources, changes found by `check_sources`, and the discovery check on the instrument's own path.
`setup()` now brings the instrument up after discovery, moving nothing: the motor tables are loaded where a board has lost them, the interlock is armed and every axis enabled, in the order Biomek's Master Controller logs show. - `driver/master.py`: `initialize`, `configure`, `arm_interlock`; a fault at rest is only noted, one under power is reset with a warning, one that does not clear trips the interlock and shows ERROR; per-pod skips; `stop` stops each pod's module and trips the interlock when the driver armed it; a motor failing on its own, or the interlock tripping under the driver, trips it and shows ERROR, as the Master Controller's logs show; `pause` / `resume` (a deliberate trip with PAUSED shown, then re-arm and enable). - `driver/features/chassis.py`, `door.py`: the interlock's writes, the status display, the door (`front_cover`), which only reports. - `driver/features/pod.py`: configure, enable, axis states, positions, homed state. - `driver/simulator.py`: `SimulatedAceUSB`, each axis's state and the interlock, door and drive supply the boards of one instrument share. - `session_log.py`: a run's log lines in a file of its own, to compare with Biomek's logs. Tests: the simulated boards' rules; configure sends the Configure, interlock and I/O box requests the Master Controller sent at the recorded cold start, in its order; enabling follows its order; faults at rest, under power and stuck; the drive supply, an open door, a broken light curtain; door reports; pause / resume; a motor failing on its own; stop; the session log.
…sion guards Moves in counts, planned as Biomek's Master Controller plans them, with the guards from the first X move: - `driver/motion.py`: the `AM` plan (each axis's speeds so all arrive together, jerk-corrected S-curves, the Move All timeout), the position record (position, adjusted) and the Span-8 Z channels' hysteresis, the order of the per-axis requests; checked against Biomek's logged moves. - `driver/recipes/homing.py`: Home All per arm type, every request as Biomek's logs show it, homed axes kept or homed again (`force`). - `driver/features/pod.py`: `move_axes` (positions read where unknown, force limits, sensor stops), `home`, `park`, `move_aside`, travel at a traverse height (`move_over`, `lift_channels`). - `driver/x_rail.py`, `driver/pod_geometry.py`: the shared X rail; a move into the other pod's range is refused before anything is sent, one crossing only its running move's path waits. - `driver/pod_collision.py`: the gripper (and what it holds) against its own pod's pipettor. - `driver/master.py`: `home`, homing in `initialize` and `setup`, a pod left out whose X is not homed refuses moves beside it, `move_axes` for both pods together, `resume` moving the pods back; positions forgotten on a trip, a failed move, `stop` and `pause`. - `driver/simulator.py`: positions, moves, homes and linked homes, force and sensor stops, and moves taking the instrument's time (`time_scale`). Tests: a cold Home All sends, on each pod's board, the requests the Master Controller sent at the cold start the generic recording was made from, first sent in its order; a forced Home All of a homed instrument as Biomek's logs show it; logged `AM` moves inline; the rail and self-collision guards; moves, homing, park, the position record; faults and positions.
A PyLabRobot deck and the instrument as a resource, in both modes, as the STAR's: - `deck.py`: the deck in Biomek's pin grid and frame, ALPs placed by Biomek's own `column` and `row`, the declared configuration's layout (`BiomekI7Deck.from_configuration`). - `resource_model/`: Biomek's ALPs and their positions (the tip-loading position's adaptor plate, trash disposal offsets), the bridges and pods with their heads, probes and grippers from each pod's declared geometry, with 3D models; Biomek's labware classes by name (`labware.py`: the definitions in `pylabrobot.resources.beckman_coulter`) and Biomek's touch-off distance. A pod's body is of category `pod`, an ordinary child of its bridge. - `targeting.py`: from a resource on the deck to where a pod's first channel goes, with the pod's calibration of the deck position; `layouts.py`: a method's Instrument Setup on the deck. - `device.py`: `BiomekI7Device` and `BiomekI7(...)`, the instrument or `simulation=True` (the generic instrument when given no files); the pods' models drawn from the declared configuration before setup and following each pod's axes after; `setup` (home, then park), `park`, `pause`, `resume`, `home_all`, `stop`, the deck's traverse clearance; `state_json`, the tree's state (`save_state_to_file`: the tips on the heads' shafts), restored when the device is built and carried over to the models rebuilt at setup. - `Pod.resource`, `driver.deck`. Tests: on the generic configuration's own deck: positions where Biomek's instrument file has them, the models taking up the X the rail guard counts and following each axis, targets against Biomek's logged moves, travel at the traverse height and the deck's clearance, Biomek's labware by class name and its touch-off distance, the device in both modes.
Biomek i7: the deck, the device and the pods' modelsBiomek i7: the deck, the device and the pods' models
cmoscy
force-pushed
the
bi7/05-deck-device
branch
from
November 17, 2026 03:48
f973c68 to
ec9d1ad
Compare
cmoscy
marked this pull request as ready for review
November 17, 2026 03:49
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
BiomekI7/ deck / resource models (ALPs, pods, heads, grippers) with 3D meshes.Test plan