Pressure-curve calibration is usually obtained by sampling a few strokes at different force

What it is

Individual fitting need not mean dragging an abstract gamma curve. The usual method is to write a few strokes at named levels—the lightest still-visible line, ordinary writing force, the heaviest that can be held—and fit dead zone and slope from those force anchors. Calibration is a sampling procedure, not the settings sliders shown again.

Why it happens

The ends and centre of a habit band are enough to pin a monotone curve: the lightest anchor sets the dead zone (lighter is noise), the heaviest sets saturation (harder does not widen), the middle sets working slope. Sampling while writing is dynamic force, closer to real use than a static press on a scale. A couple of strokes per level, reduced with a robust statistic, keep one impact spike from flying the saturation point. Instructions must name a repeatable act (“as when taking notes”), not “use 50% force”. The UI should map those strokes to width on the spot: if the lightest stroke is already thick, the person can see they are being treated as heavy-handed and change it. Fitting needs a monotone constraint so jitter among three points cannot produce a curve that thins in the middle. Too few samples (one “draw anything” stroke) cannot estimate band width; too many (a dozen levels) fatigue people, and the last “heaviest” is no longer heaviest.

Studying it

Compare no calibration, three-point dynamic writing, static press calibration, and free-form curve dragging. Draw target thicknesses immediately after calibration and a day later.

Independent variables: number and type of anchors (dynamic write / static press), wording, live preview of the mapping. Dependent measures: time to calibrate, hit rate after, hit rate a day later, whether people understand what they are doing.

Separate “finished the flow” from “the band was actually aligned”: some people write all three prompts at the same force.

Where it stops holding

People who cannot grade force on command (children, some motor impairments) will fail calibration; fall back to a conservative factory-soft curve plus a simple soft/hard pair, rather than blocking on the wizard. The first stroke before the nib and sensor have settled can bias force; do not use the first second of a session. A mandatory calibration on every launch invites scribbling through, which yields garbage anchors.

Applying it

• Offer a three-stroke “lightest / usual / heaviest” calibration, show the widths those strokes map to, and write the curve on confirm.
• Take a robust value per level and drop obvious impact spikes.
• Keep calibration in a findable setting, not as a launch gate; on failure, allow skip and keep last time or factory.
• How to check: after calibration, new strokes under the three prompts should land thin, medium, bold, and monotone. If all three are written at nearly the same force, the UI should say so rather than fit a near-vertical curve. A day later, without recalibrating, hits should still land.

Related

• Same group: C5.09.1 Natural grip force varies; the same curve is too soft or too hard · C5.09.2 Fitted curves should persist as user settings · C5.09.4 Unshared parameters force retuning in every app
• Adjacent: C5.01 Pressure Sensitivity · C5.13 Paper Feel, Friction, and Writing Haptics
• Search: pressure calibration · force anchors · light medium heavy