Barcode

Woodland and Silver's code is a bullseye of concentric circles, not parallel bars. Work out why before reading on — the answer is in step 3.

Write down three unknowns: distance, orientation, and speed. Every feature of the encoding exists to remove one of them.

Draw concentric rings and scan a line through the centre from any angle. It reads identically. A bullseye is orientation-independent — that is exactly what it buys.

Smudge one edge of your printed bullseye. Ink spread ruins every ring at once, whereas a linear code smears along the bars and can still be read.

Draw a two-width code — narrow and wide bars, narrow and wide gaps. Use Code 39, which encodes each character in nine elements, three of them wide.

Make wide elements about 2.5 times the narrow ones. Because the meaning is in the RATIO, the same label reads at any scanning distance or print scale.

Bracket the data with a reserved pattern. The reader uses it to find the code, to time itself, and — if the pattern arrives backwards — to know the label is upside down.

Leave blank margins about ten narrow-bars wide. Most real-world scan failures are a missing quiet zone, not a damaged code.

Derive a digit from the others by a fixed rule. A single misread bar then produces an arithmetic contradiction rather than a wrong but plausible product.

Contrast is what the sensor measures. Glossy stock reflects the illumination straight back and blinds the reader regardless of how good the code is.

Mount an LED and a photodiode side by side facing the label. Dark bars reflect little light and the photodiode output falls — that is the entire sensing principle.

Draw the head across the code and log the signal. You get a square-ish wave whose transitions are the bar edges.

Measure the interval between transitions and compare each against the others. Sweep faster or slower and the ratios are unchanged — which is why hand-scanning works at all.

Ink over part of a bar and rescan. A good reader reports a failure rather than a wrong number — refusing to answer is the correct behaviour.

The patent. US 2,612,994, "Classifying Apparatus and Method", filed 20 October 1949 and granted 7 October 1952 to Norman J. Woodland and Bernard Silver. The origin is well documented: Silver overheard a food-chain executive asking Drexel Institute for a way to capture product data at checkout. Woodland, sitting on a Miami beach, drew four lines in the sand by extending Morse code's dots and dashes downward into bars — and then wrapped them into a circle so the code could be read from any direction.

Why the bullseye failed anyway. Rotational invariance is genuinely valuable, but circular symbols are hard to print well. Presses smear ink in the direction of travel, and a smear across a bullseye corrupts every ring simultaneously; the same smear along a linear code merely lengthens the bars, which a ratio-based decoder tolerates. When IBM's George Laurer designed the Universal Product Code in the early 1970s he went back to the linear form for exactly this printability reason. The first UPC scan was a packet of Wrigley's chewing gum in Troy, Ohio, on 26 June 1974 — twenty-two years after the patent, and after it had already expired.

Ratio encoding is the deep idea. Absolute measurement requires a known distance and a known speed; a ratio requires neither. By carrying meaning in the relationship between adjacent elements, the code becomes invariant to scale and to scan velocity, and a hand-swiped wand works as well as a fixed slot scanner. Start and stop patterns supply framing and direction; the quiet zone supplies a reference for what white means; the check digit converts a plausible misread into a detected error. Each mechanism removes one specific unknown, and that decomposition is worth studying on its own.

Neither inventor profited. Silver died in 1963, before the technology was adopted. Woodland and Silver sold the patent to Philco in 1962 for a reported 15,000 dollars, and it expired long before supermarket scanning became universal. The most economically consequential identification system of the twentieth century made almost nothing for the people who invented it — a recurring pattern in this programme, and a reason to read patent dates against adoption dates rather than assuming they coincide.