How QR codes work: error correction and choosing a version
A QR code looks like a grid of black and white dots, but the internals include error-correction codes, encoding-mode switching, alignment patterns, and format information. This article walks through the structure and the choices you make when generating one.
Anatomy of a QR code
Every QR code has fixed structural marks:
- Finder patterns — the three large squares in the corners (top-left, top-right, bottom-left).
- Alignment patterns — smaller squares appearing in larger versions.
- Timing patterns — alternating black/white lines between finders.
- Format information — encodes the error-correction level and mask pattern.
- Data + error correction — fills the rest.
Reed-Solomon codes provide the redundancy.
Versions: size in steps
Versions go from 1 to 40, each larger than the last:
| Version | Modules | Max numeric data (level L) |
|---|---|---|
| 1 | 21×21 | 41 chars |
| 5 | 37×37 | 154 chars |
| 10 | 57×57 | 395 chars |
| 20 | 97×97 | 1,259 chars |
| 40 | 177×177 | 7,089 chars |
The maximum is version 40 — 4,296 alphanumeric or 7,089 numeric characters.
Error correction levels (ECL)
QR codes carry redundancy so they remain readable when partially obscured:
| Level | Recoverable | Notes |
|---|---|---|
| L | ~7% | Most data, least correction |
| M | ~15% | Standard |
| Q | ~25% | More correction |
| H | ~30% | Highest, least data |
“30% recoverable” means up to 30% of modules can be missing or wrong and the original data is still readable.
Encoding modes
QR codes switch encoding modes for efficiency:
| Mode | Characters | Bits per char | Use |
|---|---|---|---|
| Numeric | 0–9 | ~3.33 | Pure numbers |
| Alphanumeric | 0–9, A–Z, space, $%*+-./: | 5.5 | Caps + digits |
| Byte | Any 8-bit data | 8 | URLs, UTF-8 text |
| Kanji | Shift JIS kanji | 13 | Japanese (special mode) |
Numeric is most efficient for digit-only data; Byte handles URLs and UTF-8 text.
Picking a level by use case
Print (posters, business cards)
- Error correction: Q or H
- Print suffers wear, fingerprints, and dirt.
- Build in margin.
Screen display (web, apps)
- Error correction: L or M
- Less degradation risk.
- Pack more data.
QR codes with logos
To embed a logo in the center, use error-correction redundancy to compensate for the obscured area:
- Logo covers ~10% → level M is enough.
- Logo covers ~25% → use level H.
H allows 30% recovery, so up to ~25% obscured remains readable.
Mask patterns
Data alone can produce visual imbalances (“too much black”). QR codes XOR data with one of eight mask patterns chosen at encode time to maximize readability. The chosen mask is recorded in the format information.
Sizing a QR code in practice
“50-character URL on a print piece”:
- Data: 50 bytes (ASCII URL).
- Byte mode: 50 × 8 = 400 bits.
- Level H gives ~70% of data capacity.
- Versions 3–4 fit (33×33 modules).
For a business card with a product URL, version 4 at level H is plenty.
Minimum print size
Rule of thumb:
min size (mm) ≈ modules × 0.4mm × 1.5Version 4 (33×33) → about 20 mm. Standard business-card QR codes land at 20–25 mm.
For long-distance reading (station posters), aim for “diagonal ≥ reading distance / 10”.
Micro QR and Rectangular Micro QR
Beyond the standard square QR:
- Micro QR — as small as 11×11, for tiny tags.
- rMQR (Rectangular Micro QR) — long-rectangle shape for label strips.
ISO-standardized but with limited reader support, so mostly industrial.
Implementation notes
- Render as SVG when scaling is needed.
- 300 DPI is enough for print.
- Quiet zone — leave at least 4 modules of margin around the code.
- Contrast — the dark/light contrast must be high (don’t put a black code on a dark background).
Summary
- A QR code is defined by version (1–40), error-correction level (L/M/Q/H), and encoding mode.
- Print: Q or H. Screen: L or M.
- Logos in the center work with level H absorbing the loss.
- Always include a quiet zone.
To turn URLs or text into a QR code, the QR code generator on this site lets you set size and error-correction level.