A measurement-device-independent link behaves differently on deployed fibre than on a spool. This note sets out the measurements worth taking before installation, and the records worth keeping during a trial.
Characterising the route
Begin with an optical time domain reflectometer trace of every span, taken at the wavelengths the quantum channel will use rather than at the standard maintenance wavelength. Record splice losses individually. A route with the same total loss can behave very differently depending on whether that loss is distributed or concentrated in a handful of bad joints.
Polarisation and timing
Log polarisation drift over at least one full diurnal cycle, and preferably across a weather front. Deployed fibre in ducts follows ground temperature with a lag of hours. Timing stability should be measured against the same reference the deployed system will use, not against a laboratory clock.
Coexistence with classical traffic
Measure with the classical channels running at their operational power, not turned down. Raman scattering from classical traffic is the dominant noise source on a shared strand, and a trial that quietly reduces classical power is not a trial of the deployed condition.
What to record during operation
Secret key rate, quantum bit error rate, and channel loss, sampled continuously and stored with timestamps. Interruptions should be logged with cause where known. The value of a field trial lies almost entirely in the interruptions: anyone can report a good week.
Judging the result
A credible result states the measurement duration, the loss budget, the classical load, and the number and cause of interruptions. A result that states only a peak key rate describes an experiment, not a network.