Halftone, lpi and mesh count

#Overview

Halftoning turns your image's tones into dots of varying size — that's what lets a screen printing screen, which only knows "ink passes" or "ink doesn't pass," render a gradient. But a halftone dot only exists if the screen's mesh lets it exist. A dot smaller than the gap between two threads falls literally between the mesh: it doesn't expose cleanly, or it prints ragged.

Which leaves the only question that matters: what halftone fineness for what screen?

#The shop rule

Halftone fineness is measured in lpi (lines per inch — lines of dots per inch). Screen mesh is given in threads/cm in Europe (a "90T" has 90 threads per centimeter), in threads per inch elsewhere (multiply by 2.54: 90 threads/cm ≈ 230 threads/inch).

Rule: lpi ≤ threads per inch ÷ 4 to 5.

Each halftone dot needs at least 4 to 5 threads to anchor onto. Below that, the smallest dots (the print's highlights) disappear and the largest ones (shadows) merge together — the image loses both ends of its range at once.

These figures are shop-floor orders of magnitude, not laws: emulsion, screen tension and ink all shift the limit. A test strip always settles it.

#What Calibration Flow sets

In Calibration Flow, you don't choose lpi directly: halftone fineness follows export resolution, with a fixed-size halftone cell. The relationship is simple:

The halftone is an AM halftone (variable-size dots, rotated screen to avoid alignment artifacts) with automatic fallback to error diffusion (Floyd-Steinberg) when the setup doesn't suit it well — see Dithering your negative.

Cross-reading with the table above:

• Paper screens / fine mesh (90T, 120T): export at 300 dpi (≈ 50 lpi) — the comfortable pairing. Only go to 400-500 dpi if your exposure chain is very clean.
• Textile 77T: 300 dpi is at the upper limit. A test strip is mandatory.
• Textile 43-55T: the 50 lpi AM halftone is too fine for the mesh. Two options: work in flats (no halftone — the natural regime of Sanguine, where each ink carries its own tonal slice), or test the error-diffusion fallback's look, whose irregular grain sometimes forgives better than a regular halftone that moirés against the mesh.

#Mesh/halftone moiré

Two regular patterns overlapping — the film positive's halftone and the thread grid — can produce a visible interference pattern: moiré. Calibration Flow's screen rotation avoids it in most cases, but no single angle is safe for every mesh. If moiré shows up on your test strip: change the export resolution (and so the lpi), or rotate the film positive slightly on the screen at exposure time.

#Key takeaways