Paper-Based Storage

Part of Data Storage

The original digital medium — how punched cards and perforated tape made computing possible before magnetic media existed.

Why This Matters

Before magnetic storage existed, before electronic memory was affordable, and before anyone had manufactured a transistor, computing machines were fed data and programs on paper. Punched cards predate the digital computer by sixty years — Herman Hollerith used them for the 1890 US Census. Punched tape was used in telegraph machines in the 1850s. These technologies are not primitive compromises but mature, well-engineered systems with documented advantages.

For rebuilders, paper-based storage has unique properties that no other medium can match. Paper requires no power to maintain data. It can be read visually by holding it up to light, with no hardware whatsoever. It can be reproduced on any surface capable of being punched. It uses no exotic materials. And if the paper is reasonably dry and protected from insects and fire, it can outlast every electronic medium.

These are not trivial advantages. In a scenario where electrical power is intermittent, where electronic components are scarce, and where long-term archival security matters more than high-speed access, paper deserves serious consideration.

How Punched Media Works

The principle is elemental: information is encoded as the presence or absence of holes at defined positions on a paper medium. A hole represents a 1-bit; the absence of a hole (solid paper) represents a 0-bit. To read the data, you determine which positions have holes and which do not.

This can be done mechanically (metal pins fall through holes and actuate switches), optically (light passes through holes and strikes photosensors), or pneumatically (air passes through holes and actuates pressure sensors). Mechanical and optical reading are most common.

The density and organization of holes defines the format. Rows of holes represent individual characters or data words. The number of hole positions per row defines the character set or word width.

Punched Cards: The Hollerith Format

The most widely used punched card format is the IBM 80-column card (Hollerith card), standardized in the 1920s and used until the 1980s.

Physical dimensions: 187.3 mm × 82.6 mm × 0.18 mm. Stiff card stock, slightly stiffer than modern filing cards. The lower-right corner is notched to allow orientation detection.

Column structure: 80 columns across the width, 12 rows top to bottom. The top 3 rows are the “zone rows” (rows 12, 11, and 0), and the bottom 9 rows are “digit rows” (rows 1–9). Each column can represent one character by punching one or more of the 12 possible hole positions.

Character encoding: The Hollerith code maps alphanumeric characters to specific hole combinations. Letters A–I are encoded as zone row 12 + digit row 1–9. Letters J–R are zone row 11 + digit rows 1–9. Letters S–Z are zone row 0 + digit rows 2–9. Digits 0–9 are single punches in the digit rows 0–9. Special characters use various zone/digit combinations.

Storage capacity: 80 columns × 12 rows = 960 possible bit positions per card. In practice, 80 characters of 6-bit representation per card. A box of 2,000 cards holds 160,000 characters ≈ 160 KB. A tall filing cabinet can hold millions of cards — megabytes of data.

Card reading speed: Early mechanical card readers processed 50–100 cards per minute; later readers did 1,000 cards per minute. At 80 characters per card and 1,000 cards per minute: ~1,300 characters per second.

Punching: Card punches had a keyboard layout similar to a typewriter. The operator typed each row, and the machine advanced the card and punched the column. Later machines had program drums that automatically skipped and duplicated fields. Output from computers could be punched by a keypunch-compatible output device.

Punched Tape: The Continuous Roll

Punched tape (paper tape) is a continuous strip of paper with holes representing data, used from the 1870s telegraph era through computer peripherals of the 1970s.

Physical format: Typically 25.4 mm (1 inch) wide, with a row of smaller sprocket holes down the center for mechanical feed and a row of data holes on each side.

Hole patterns: The most common formats:

• 5-hole (Baudot/ITA-2): 5 data columns, 32 possible codes. Used in telegraph and early teletype machines. Insufficient for the full ASCII character set.
• 7-hole (Friden): 7 data columns + parity, 128 possible codes. Used in some early computers.
• 8-hole (ASCII/ISO 646): 8 data columns, 256 possible codes. The dominant format for computer use. One data byte per row, with the sprocket hole between rows 3 and 4.

Storage density: 8-hole tape at standard density holds 10 characters per inch. A 100-meter roll holds ~39,000 characters ≈ 39 KB. High-density tape formats (12 or more characters per inch) exist but require more precise punching and reading equipment.

Reading speed: Tape readers typically ran at 50–2,000 characters per second in computer-era equipment. Older telegraph-era readers were much slower.

Tape punching: A tape punch receives character codes from the computer and mechanically punches the appropriate hole pattern using a bank of punch dies. Typical punch speed: 50–300 characters per second. The mechanical impact of punching creates the characteristic clicking sound.

Rewinding and splicing: Unlike cards, tape must be rewound to the beginning to re-read data. Specific sections can be spliced out and replaced by cutting and gluing tape segments — crude but functional.

Building a Simple Tape Reader

A functional optical tape reader is within reach of a basic electronics workshop:

Mechanical transport: A motor drives a rubber-coated capstan roller that pulls the tape through the reading station. The capstan contacts the tape through a pinch roller that applies gentle pressure. Speed is controlled by the motor voltage. Sprocket holes can be used for indexing by adding a sprocket wheel, or ignored in continuous-feed systems.

Optical sensing: An LED (or incandescent lamp behind a narrow slit) illuminates the tape from below. A row of photodiodes (one per data channel plus one for the sprocket) detects light passing through holes. For 8-hole tape, 9 photodiodes (8 data + 1 sprocket) are required.

Electronics: Each photodiode output is amplified and compared to a threshold to produce a digital logic signal. The sprocket channel provides a clock pulse for each character row. A shift register or microcontroller latches the 8 data bits on each sprocket pulse.

Materials for a scratch-built reader:

• Aluminium channel or steel plate for the frame
• Small 12V DC motor (from a cassette deck or similar)
• Rubber-coated roller (o-ring on a shaft)
• 9 photodiodes or phototransistors
• 9 comparator circuits (LM393 or similar)
• 8-bit latch (74HC573 or microcontroller I/O port)
• 5V regulated supply

Testing: Feed a known tape pattern and verify each character on an oscilloscope or logic analyzer. Common errors: misaligned sensors (some holes missed), insufficient threshold (reads noise as holes), mechanical slippage (characters skipped).

Punching New Cards and Tape

Making a hand punch for tape: A strip of hardened steel with a 1.17 mm diameter pin (for standard tape holes) mounted in a guide frame can punch individual holes by hand. Time-consuming but functional for small programs. A row punch with 8 pins (one per data position) punches an entire character at once with a single press.

Making a card punch: Cards require more precision (the hole position within a column must be exact for reliable reading). A punch plate with 80 column guides and 12 row guides allows hand-punching, but speed is very low (~100 cards per day for a skilled operator).

Encoding data for paper: Before punching, data must be serialized into a format suitable for paper. For programs: source code characters directly (one punch per character). For binary data: encode each byte as two hex digits (0–9, A–F) per character, or use base64 encoding for higher density.

Archival Properties

Longevity: High-quality paper in dry, insect-free, moderate-temperature storage can remain physically intact for centuries. The ink or pencil marks that are commonly added for human-readable labels may fade, but the holes remain. Paper punched in the 1880s by Hollerith’s Census machines has been read successfully in the 21st century.

Vulnerability: Paper is combustible, destroyed by water (and subsequent mold), and eaten by insects (silverfish especially). Store in airtight containers (sealed metal boxes or mylar bags with silica gel desiccant) in a cool, dry location.

Priority content for paper archival: Programs that must not be lost regardless of electronic media failure, critical lookup tables (trigonometric functions, constants), character encoding tables, protocol specifications. The goal is to preserve knowledge that would be needed to reconstruct all other programs from scratch.