Results | Experimental methods

Low-FER results and precision

The tables below count unique measurements. AWGN curves reused at other symbol rates are counted once.

Points confirmed by the upper confidence bound

FER targetOriginal pointsCombined points
1e-0324902750
1e-040777
1e-0500

These counts do not describe the number of independent modes. Exact tested thresholds by mode and channel are in tested_thresholds.csv. No confidence guarantee is inferred between measured points.

Allocated precision targets

AllocationPointsTarget metTarget not metZero-error upper bounds
Both sides within 20% of the estimate14747737010
Both sides within 50% of the estimate12816386430
262,144 total trials; interval only1704--262

For a positive estimate p, a relative target r requires lower >= (1-r)p and upper <= (1+r)p. A target not met means that the interval is wider than planned; it does not lower its confidence level.

Among 1,704 fixed-tail points, 1,442 have observed errors and 262 have zero errors. For 0/262,144 errors the upper bound is 6.205515e-05. This supports FER <= 10^-4, but not FER <= 10^-5.

All added-point counts and precision status (CSV) | Precision totals (CSV)