Nomicon Fonts

Capture

Capture

Every card is scanned twice on a flatbed scanner, and a printed target keeps the scanner honest.

Each card goes on the glass at 1200 ppi in greyscale, saved without compression. The scanner’s own straightening is switched off, because it resamples the image and changes the weight of the strokes.

Each card is scanned once upright and once turned a quarter turn. That puts the same printed letter on two different pixel grids. Averaging the two copies cuts the noise of the scan by about a third, and the difference between them measures that noise directly. The Noise Lab below shows it.

A calibration target printed on paper is scanned both ways too. It reads 47.32 pixels per millimetre, which is 1202 ppi, the same in both directions to within 0.03 %. The two things below are the ones that limit what the project can claim.

Calibration Bench

116 µm to go from paper to ink

Each dot is one edge of the slanted squares on the target, measured as the width over which it changes from 10 % to 90 % ink. They range from 93 to 131 µm. One scanner pixel is 21 µm, so an edge is blurred over about 5 pixels. Stacking reduces noise, but it cannot recover detail finer than this.

Not a straight line from white to black

The target has eleven tone steps from no ink to solid black. The scanner compresses the darks: the steps from half black upward all read nearly the same. Turning grey level into ink amount needs this curve. Hover or focus the chart and use the arrow keys.

Tone steps, grey level against percent black Grey level falls steeply for the first steps and flattens in the darks, for two scans. A straight line is drawn for comparison. Use the left and right arrow keys to read each step. 064128192255 020406080100 percent black printed grey level read
Printed black
50 %
Scan A, solid
57
Scan B, dashed
61

The dotted line is what a linear scanner would read between the white and the black ends. The two scans differ slightly in the mid-greys, so the curve is read per scan session.

The scale from each ruler, every edge width and the tone values are in the tables on the data page.

Noise Lab

Drag the line across one letter to compare its two copies, or switch to the difference. It is the same ink, so where they disagree is scan noise.

The letter e, scanned at 0°
The letter e, scanned at 90°

Two copies of one printed e from Reef Worm, scanned at 0° and 90°.

In the pilot this noise was about the size of the real differences between two printed copies of a letter, which is why each card is scanned twice and many copies are stacked. One scan is off by about 7.5 µm at the edge and the mean of a pair by about 5.3 µm. The distribution over all pairs is on the data page.