Error Correction for Free

11 = 1 + 2 + 3 + 5

The fifth prime doesn't compute. It checks. Every element in the ring carries its own error-detecting code -- for free. No extra bits. No overhead. 11 = 1 + 2 + 3 + 5. Add mod-13 for dual parity: 100% detection and correction.

How It Works

Every number in Z/214,414,200 decomposes into 7 channels via CRT: (mod 8, mod 9, mod 25, mod 49, mod 11, mod 13, mod 17). The first four channels carry data. The last three (mod 11, mod 13, mod 17) provide triple-parity ECC -- rate 4/7, 100% correction. Error correction is structural, automatic, free.

Encode: take any element n. CRT gives 7 residues. The first four (mod 8, 9, 25, 49) are data -- they determine n within Z/88,200. The last three (mod 11, 13, 17) are parity. Start with the simplest check: n mod 11.

Detect: reconstruct n from the first 4 channels. Compute n mod 11. If it matches channel 5: intact. If not: corruption detected. 100% for prime channels (mod < 11). Prime-power channels (mod 25, 49): ~92% -- errors at multiples of 11 are invisible to mod-11 alone. Dual-check with mod-13 closes the gap.

Correct: if you know which channel was hit, try all values in that channel (at most 49 trials for mod-49). For prime channels: exactly one value passes the mod-11 check. For prime-power channels: multiple candidates may pass. Dual-check with mod-11 + mod-13 narrows to exactly one (lcm(11,13) = 143 > 49).

Detection
100% prime, ~92% prime-power
Prime channels (mod < 11): perfect. Prime-power: delta=11k invisible. Dual mod 11+13: 100%.
CRT Detection Theorem
Overhead
0 bits
No extra storage needed. Protection is the number system.
Structural
Correction
100% with dual
Mod-11 + mod-13 dual-check: unique fix for ALL channels.
Brute-force search
Cost
9.1%
Adding 11 costs 1/11 of unit density. Full protection for 9.1%.
phi analysis

The Detection Theorem

CRT Detection Theorem
Let R = Z/q1 x ... x Z/qk x Z/p where gcd(qi, p) = 1 for all i. If exactly one channel ci is corrupted by delta != 0 (mod qi), detection via mod-p succeeds iff delta is not divisible by p. For qi < p (prime channels): 100% -- no such delta exists. For qi > p (prime-power channels): ~92% -- deltas at multiples of p are invisible.

Proof sketch: CRT lifting idempotent ei satisfies ei mod p != 0 (from gcd condition). Corruption delta in channel i shifts the mod-p residue by delta * (ei mod p). Since ei mod p != 0, the shift vanishes iff delta = 0 mod p. With p = 11: prime channels have max delta < 11, so no blind spot. Prime-power channels (mod 25, 49) have some deltas divisible by 11. Dual-check with mod 13 eliminates these: lcm(11,13) = 143 > 49.

11 = 1 + 2 + 3 + 5
The parity prime equals the sum of the identity plus the first three primes. Forced by the Cunningham chain c(n) = 2n+1: c(5) = 11. Not arbitrary -- the chain builds the check digit from the data digits.

Three Rings, Three Levels of Protection

Three rings, three ECC stories. Raising prime exponents increases data capacity but changes detection rates:

RingChannelsDetectionNotes
Z/2102, 3, 5, 7 (no 11)0%Unprotected. Fast but fragile.
Z/2,3102, 3, 5, 7, 11100% + 100% corrAll channels < 11. Perfect ECC.
Z/970,2008, 9, 25, 49, 11100% prime, ~92% ppPrime-power channels leak 1/12.

Z/2,310 has perfect protection: every data channel (mod 2, 3, 5, 7) is smaller than 11, so the mod-11 check catches everything. Raising exponents to (mod 8, 9, 25, 49) means mod 25 and mod 49 are larger than 11. Some corruptions in those channels produce the same mod-11 residue.

Exponent-ECC Tension
Raising 5 -> 5^2 = 25 and 7 -> 7^2 = 49 increases data capacity but weakens mod-11-only detection. Mod-25 channel: 22/24 = 11/12 detection. Mod-49 channel: 44/48 = 11/12 detection. Mod-8 and mod-9 channels: still 100% (both < 11).

The Dual Check: mod 11 + mod 13

The solution to prime-power channel leakage: add 13 as a second check channel. A corruption is invisible only if delta mod lcm(11, 13) = delta mod 143 = 0. The maximum delta in any channel is 48 (from mod 49). Since 48 < 143: no corruption can be invisible to both checks.

mod-11 only
~92% prime-power
Mod-25 and mod-49 leak 1/12 of corruptions (delta divisible by 11).
mod-11 + mod-13
100% detection
lcm(11,13) = 143 > max delta = 48. Complete.
Dual-Check Theorem
Correction
100% known-loc
With both syndromes: unique fix for ALL channels.
Cost
7.2 bits
log2(143) overhead. Full protection of 17+ bits of data.

The 490 split goes further: {2, 5, 7} = data (8 * 25 * 49 = 9800 values), {3, 11, 13} = parity (9 * 11 * 13 = 1287 values). 9800 * 1287 = 12,612,600. Three parity channels for three data channels: a (6,3,3) code that corrects any single-channel error. See the live demo on the ECC Live page.

Why 11 Stays at Exponent 1

When the ring raises exponents from Z/2,310 to Z/12,612,600, every data prime's exponent increases: 2 -> 2^3, 3 -> 3^2, 5 -> 5^2, 7 -> 7^2. But 11 stays at exponent 1. It is the only channel unchanged by raising exponents -- the fixed reference against which data channels grow.

11's exponent
always 1
11 is the only prime whose exponent doesn't increase. Parity stays at first power.
Byte recovery
100% unique
For [0,255] in Z/970,200: CRT spacing >> byte range.
AI weights
free ECC
Quantize to Z/970,200. Corruption detectable without overhead.

Explore: ECC Syndrome Check

Enter any element n (0 to 970,199). See its CRT decomposition and mod-11 syndrome. The mod-11 channel acts as a free checksum -- if it mismatches after corruption, the error is detected.

Enter n:

Try: 137, 42, 210. Corrupt one channel mentally and see if mod-11 catches it.

Paradigm Contrast

AspectStandard (INT8)Ring (CRT + mod 11)
Error detection0% -- bit flips are silent100% with dual parity mod 11 + mod 13 (proved)
Error correction0% -- no mechanism exists100% known-location with dual parity
Levels256 (8 bits, arbitrary)210 or 970200 (structured)
Cost0 overhead, 0 protection0.3 bits SQNR -- free protection
On corruptionWrong value, no warningDetected -> located -> corrected
ParadigmPray nothing goes wrongRing self-heals

Reed-Solomon, LDPC, Turbo codes all ADD redundancy to protect data. The CRT ring has redundancy built in. 11 is not bolted on -- it is the fifth prime in the number system. Protection is a structural property of the ring's CRT decomposition.

Proved on Hardware

Three ESP32 boards running .ax WASM firmware on the TRANS ring Z/214,414,200. Seven channels. Triple-parity ECC (rate 4/7). Packets broadcast via ESP-NOW radio every 3 seconds. Error injection via serial command corrupts one CRT channel per packet. The receiver corrects it.

Full sweep
123/123
7 channels x 3 error sizes (delta +1, +2, max-1) = 21 configurations. 123 corrupted packets corrected, 0 failures. Every data channel (mod 8, 9, 25, 49) and every parity channel (mod 11, 13, 17) tested at multiple corruption magnitudes.
Sensor round-trip
6/6 exact
Real BME280 temperature, humidity, and pressure from two boards. Each reading decomposed into 7 CRT residues. Reconstructed value matches original exactly. ECC passes on all 6. The ring is not abstract -- it carries physical measurements.
Mirror on data
4/4 perfect
For every sensor reading n: CRT(n) + CRT(N-n) = 0 in all 7 channels. The mirror theorem holds on physical data. Even a temperature reading obeys the ring's additive symmetry.
Uniqueness
Proved
Any 6 of 7 channels uniquely determine the element. lcm of any 6 moduli exceeds the data ring (88,200). Single-channel correction always has exactly one solution.

Source code · Public domain (CC0)

Report issue

.ax source compiled to WASM via self-hosting compiler. Zero HTML authored.