[BioCam]

The Problem

The v1 CLARIOstar driver can read a well only at one point, its nominal centre from the plate definition. It cannot tell where a well's signal actually is: whether the plate sits where the model says, whether a plate definition's well positions are right, or where the centre of a well of non-standard labware lies.

The reader itself can find this out. It has a scan (07) that reads the fluorescence signal over a 31 x 31 grid of points around a point on the plate holder and reports the centre of the signal it found. The scan takes one or two points; no driver of ours sends it.

PR Content

Two methods on clariostar.fluorescence:

• measure_grid: the grid of counts around one or two given points on the plate holder (x and y in mm from its left-front corner), at one focal height and gain, with the centre the reader found for each. The points need not be wells.
• measure_well_centers: the same for any wells of the plate on the plate holder, scanned two at a time. Without a focal height, a z-scan of the first well (measure_z_profile, which reads the signal at a range of focal heights) gives the focal height, and the gain too unless one is given.

from pylabrobot.lib.fluorescence import FluorescenceChannel
from pylabrobot.resources import Coordinate

channel = FluorescenceChannel(
  excitation_wavelength=488,
  excitation_bandwidth=14,
  emission_wavelength=535,
  emission_bandwidth=30,
  dichroic_split_wavelength=507.5,
)

# the plate's first and last well (wells=None); any wells with wells=[...]
grids = await clariostar.fluorescence.measure_well_centers(channel, focal_height=15.0, gain=2239)
for grid in grids:
  print(grid.well, grid.center, grid.spacing, len(grid.counts), "x", len(grid.counts[0]))
# A1 Coordinate(014.450, 074.130, 000.000) 0.33 31 x 31
# H12 Coordinate(113.330, 011.480, 000.000) 0.33 31 x 31

# any one or two points on the plate holder
grids = await clariostar.fluorescence.measure_grid(
  channel, positions=[Coordinate(14.3, 74.3, 0)], gain=2239, focal_height=15.0
)

A real scan of a plate's A1 and H12, replayed through measure_well_centers (the call above): each grid's counts, with the well's rim as the bright ring, and the centre the reader found against the plate definition's, 0.13-0.28 mm off. The panels are drawn in the order the reader sends the points; which way its rows and columns run on the plate is not verified yet.

Each result holds the grid's position, its well (None for a given point), the centre the reader found (None if it found none), the spacing between points in mm, the counts row by row, and the focal height and gain it was read at. A grid takes about 50-80 s a point; cancelling the call stops the scan on the reader.

The scan sends three commands that are new to the driver:

• 07, the scan itself;
• 05 0d, the grid's layout (side and point spacing), right after it;
• 05 04, the centres the reader found, once it is idle.

Each grid comes back through the existing data request (05 02) whenever status reports data ready. The scan reuses the plate block and the optics encodings of a fluorescence read.

The decoding is tested against a recorded scan of a plate's first and last well, the only pair verified on the reader. Any other point or well is accepted as the scan allows, but it is untested on hardware and warns so.

🤖 Generated with Claude Code