Measurement Errors and Repeated Measurement
Because syndrome measurements are themselves noisy, codes repeat them and decode over space and time to distinguish real errors from false readings.
The problem with a single round
Syndrome extraction relies on measuring stabilizers, but measurement and the ancilla circuits that perform it can fail. A faulty measurement reports a stabilizer flip that did not happen, or misses one that did. If a decoder trusted a single round of syndromes, a measurement error would be indistinguishable from a real data error and could trigger a wrong, possibly logical, correction.
The fix: repeat in time
The standard remedy is to measure every stabilizer many times in succession. A genuine data error changes a syndrome bit and it stays changed until corrected; a measurement error changes a syndrome bit for one round only. So decoders work not with raw syndromes but with differences between consecutive rounds, called detection events. A real error appears as a persistent change, while a measurement fault appears as a transient blip, and the two have different signatures in the space-time record.
- Measurements are noisy: false positives and false negatives occur.
- Repeat syndrome extraction over many rounds.
- Decode detection events (changes between rounds), not raw syndromes.
- Data errors persist; measurement errors are single-round blips.
Space-time decoding
Stacking the syndrome rounds turns decoding into a three-dimensional problem: two spatial dimensions of the lattice plus one time dimension of rounds. In the surface code a data error links two defects in space while a measurement error links two defects in time, and a matching decoder pairs both on the same space-time graph. To protect or read out a distance-d logical qubit reliably, roughly d rounds are performed so that even a chain of measurement errors cannot masquerade as a logical operation.
This time dimension is why fault-tolerant computation is measured in space-time volume, and why decoder latency, keeping up with the stream of rounds, is a hard engineering constraint. Repeated measurement is what elevates the phenomenological and circuit-level models above the idealized code-capacity picture.