Short answer: A three-bit repetition code stores one bit as three copies and fixes a single flipped bit by majority vote. The quantum version can correct one bit-flip under stated assumptions but does not protect against phase-flips, so it is a teaching step and not a complete scheme for protecting quantum information.
The classical version
Encode one classical bit into three identical copies. If one copy flips in transmission, the other two still agree, and taking the value that appears twice recovers the original. Two or three flips defeat the vote and return the wrong bit.
Why it helps
Suppose each bit flips independently with probability p. The vote fails only if two or three bits flip, which happens with probability 3p² - 2p³. For small p that is smaller than p, so redundancy lowers the error rate. When p is below one half, the code reduces errors.
What changes with qubits
Qubits can be in superpositions, and they cannot simply be copied. Measuring them to compare copies would destroy the state. So the quantum version of the repetition code works differently: it uses extra qubits and parity checks that reveal whether an error happened without revealing the stored data.
It also has limits. A qubit can suffer bit-flips and phase-flips, and the simple repetition code covers only one of these. IBM's course on quantum error correction uses the repetition code to explain the ideas before moving on to the nine-qubit Shor code, which handles both error types.
The lesson
The repetition code is a useful teaching step, not a complete protection scheme. It shows that redundancy can suppress errors, and also that more physical qubits do not automatically give better protection. What matters is which errors the encoding catches and which it misses.
What a useful test report would say
For any eventual Quantum Ireland test, the report should state the error model, the assumptions, and the recovery results. A larger qubit count is not proof of protection. This is a proposed way to evaluate a test. It is not a report of an implementation.
Written from Clonmel, Tipperary for Quantum Ireland.
A worked way to think about it
Imagine sending the bit 1 as 111. If the middle bit flips, you receive 101, and the majority still says 1. If two bits flip you receive 001 and the vote says 0, which is wrong. The code never knows it failed. That is why error rates have to be low enough for the vote to win more often than it loses, and why real systems use stronger codes when errors are more likely.
The same reasoning shapes quantum hardware. Error correction only helps once the physical error rate is below a threshold. Above it, adding more qubits makes things worse, not better.
FAQ
What is a repetition code?
A code that stores one bit as several identical copies and recovers it by majority vote.
How many errors can a three-bit repetition code fix?
One. Two or three flips in the same block give the wrong result.
Does the repetition code work for qubits?
Only partly. A quantum repetition code can correct one type of error, bit-flips, but not phase-flips.
What protects against both bit-flips and phase-flips?
Codes such as the nine-qubit Shor code, which IBM's course introduces after the repetition code.
Do more physical qubits always mean better protection?
No. What matters is which errors the encoding detects and corrects, and the quality of the hardware.
All field notes