Update Corning 3603 with empirical cLLD height_volume_data

[BioCam]

Migrate Cor_96_wellplate_360ul_Fb (Corning 3603) from the old geometry-based compute_volume_from_height / compute_height_from_volume closure functions to a first-class height_volume_data dict (introduced in #938), populated with empirical cLLD measurements taken on a Hamilton STAR.

The Problem

The previous model used calculate_liquid_volume_container_1segment_round_fbottom — a pure cylinder approximation with constant diameter d=6.86 mm. According to the Corning tech drawing (DWG00121 Rev D), the well actually tapers from ⌀6.35 mm (bottom) to ⌀6.86 mm (top) over a 10.67 mm depth — a frustum, not a cylinder.

Even the geometrically correct frustum model overestimates volume by +2 to +16 µL across the working range compared to empirical measurements. This is expected: injection-molded polystyrene wells don't follow perfect mathematical geometry — wall draft angles, fillet radii at the bottom-wall junction, and material shrinkage all contribute to deviations that no closed-form model can capture.

The Corning 3603 is actually a good case study because the well geometry is theoretically perfectly understood (frustum of a cone with known top/bottom diameters), yet still shows systematic deviation from the physical reality. The cylinder model (old) overestimates by +12 to +45 µL (+12–25%), while the frustum cross-reference overestimates by +2 to +16 µL (+4–8%). Only empirical calibration data can close this gap.

There may be small measurement errors in the empirical data (cLLD detection edge, meniscus effects at low volumes), but piecewise-linear interpolation over 9 calibration points at pipettable volumes provides far better accuracy than any geometric model.

PR Content/Solution

• Removed the two _compute_*_Cor_96_wellplate_360ul_Fb closure functions and unused calculate_liquid_*_1segment_round_fbottom imports.
• Added _cor_96_wellplate_360ul_Fb_height_volume_data dict with 10 empirical calibration points (height mm → volume µL) measured via cLLD on a Hamilton STAR:
  • Dead volume at 0.45 mm / 20 µL (covers full flat bottom + LLD-detectable)
  • 8 working-range points at pipettable volumes (50–350 µL)
  • Max fill at 10.95 mm / 360 µL
• Passed height_volume_data= to the Well constructor, leveraging the auto-generation from Add height_volume_data attribute to Container with piecewise-linear interpolation #938.

Cylinder vs Frustum vs Empirical comparison

OLD: Cylinder model (d=6.86 mm) — replaced
REF: Frustum model (⌀6.35→6.86 mm) — cross-reference from tech drawing
NEW: Empirical (cLLD-measured) — replacement

🤖 Generated with Claude Code

[rickwierenga]

back link to original PR: #722