An amplified spontaneous emission (ASE) light source using an Er-doped fiber is an ideal random-number source in principle because ASE originates from vacuum fluctuations. Interferometrically measured ASE light directly reflects vacuum fluctuations in phase space; the interferometer does not need to be stabilized because the phase is completely random, and measurable random numbers are a continuous variable because vacuum fluctuations are continuous. These characteristics make the random-number source practical and ideal. Evaluated randomness was sufficiently ideal at the accuracy level of measurements and evaluations.
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
The unit of random numbers is ${{10}^6}$. Uniformity and proportion are evaluated for 1000 units. ${f_{\rm r}} = {0}.{625}\;{\rm GSps}$. When a test consists of multiple terms, the worst case is shown. Typical criteria are ${P} \mbox{-} {{\rm value}_T} \ge {0}.{0001}$ and proportion $ \ge \;{980/1000}$.
Failure counts for the 188 terms are shown. U denotes the uniformity test, and P denotes the proportion test. When both tests result in a zero count, each test is a “pass.” The tests use 1000 or 2000 units of random-number sequences. Each test was performed 4 times for the 1000-unit case and twice for the 2000-unit case at each sampling rate. The 2000-unit case used the same data as the 1000-unit case. The numbers indicated in parentheses in (c) express the counts of the values larger than the pass range under the ${2}\sigma $ condition. The failure counts in the differential data are reasonable, judging from their statistical fluctuations.
Table 3.
Failure Counts and Rates in the NIST SP800-22 Test
The rows labelled “${3}\sigma $” and “${2}\sigma $” indicate the results of the two-sided test, and the row labelled “${2}\sigma $ (lower side)” indicates the results of a one-sided test.
The unit of random numbers is ${{10}^6}$. Uniformity and proportion are evaluated for 1000 units. ${f_{\rm r}} = {0}.{625}\;{\rm GSps}$. When a test consists of multiple terms, the worst case is shown. Typical criteria are ${P} \mbox{-} {{\rm value}_T} \ge {0}.{0001}$ and proportion $ \ge \;{980/1000}$.
Failure counts for the 188 terms are shown. U denotes the uniformity test, and P denotes the proportion test. When both tests result in a zero count, each test is a “pass.” The tests use 1000 or 2000 units of random-number sequences. Each test was performed 4 times for the 1000-unit case and twice for the 2000-unit case at each sampling rate. The 2000-unit case used the same data as the 1000-unit case. The numbers indicated in parentheses in (c) express the counts of the values larger than the pass range under the ${2}\sigma $ condition. The failure counts in the differential data are reasonable, judging from their statistical fluctuations.
Table 3.
Failure Counts and Rates in the NIST SP800-22 Test
The rows labelled “${3}\sigma $” and “${2}\sigma $” indicate the results of the two-sided test, and the row labelled “${2}\sigma $ (lower side)” indicates the results of a one-sided test.