The high bitrate quantum random number generator (QRNG) previously reported1 has been in continuous operation for 20 months to deliver random numbers available online.2 Over 600 TB of data have been delivered since then. It has been argued that continuous operation might lead to a degradation of the randomness due to wear and thermal aging of the involved components, foremost of the photomultiplier tube (PMT) in use. Apart from small scale routine testing throughout the device's runtime, it was therefore decided recently to perform a full scale re-testing as initially performed in development prior to deployment. The testing was performed in the same extensive way as described in our original letter, using the “big crush” battery from version 1.2.3 of the TESTU01 suite.3 The test battery consists of 31 tests and consumes files of 1.3 TB. 30 such files have been tested, and the results are shown in Table I. The results indicate no significant deviation from the initial testing and from the expected random fails of individual tests.
Results of the 31 different tests of the “big crush” battery, tested on 30 files of 1.3 TB each.
| Test . | Passed . |
|---|---|
| SerialOver | 30/30 |
| CollisionOver | 28/30 |
| BirthdaySpacings | 29/30 |
| ClosePairs | 30/30 |
| SimpPoker | 30/30 |
| CouponCollector | 30/30 |
| Gap | 30/30 |
| Run of U01 | 30/30 |
| Permutation | 30/30 |
| CollisionPermut | 30/30 |
| MaxOft | 30/30 |
| SampleProd | 30/30 |
| SampleMean | 30/30 |
| SampleCorr | 29/30 |
| AppearanceSpacings | 29/30 |
| WeightDistrib | 30/30 |
| SumCollector | 30/30 |
| MatrixRank | 29/30 |
| Savir2 | 30/30 |
| GCD | 30/30 |
| RandomWalk1 | 29/30 |
| LinearComp | 30/30 |
| Fourier3 | 30/30 |
| LongestHeadRun | 30/30 |
| PeriodsInStrings | 30/30 |
| HammingWeight2 | 30/30 |
| HammingCorr | 30/30 |
| HammingIndep | 29/30 |
| Run of Bits | 30/30 |
| AutoCorr | 30/30 |
| Test . | Passed . |
|---|---|
| SerialOver | 30/30 |
| CollisionOver | 28/30 |
| BirthdaySpacings | 29/30 |
| ClosePairs | 30/30 |
| SimpPoker | 30/30 |
| CouponCollector | 30/30 |
| Gap | 30/30 |
| Run of U01 | 30/30 |
| Permutation | 30/30 |
| CollisionPermut | 30/30 |
| MaxOft | 30/30 |
| SampleProd | 30/30 |
| SampleMean | 30/30 |
| SampleCorr | 29/30 |
| AppearanceSpacings | 29/30 |
| WeightDistrib | 30/30 |
| SumCollector | 30/30 |
| MatrixRank | 29/30 |
| Savir2 | 30/30 |
| GCD | 30/30 |
| RandomWalk1 | 29/30 |
| LinearComp | 30/30 |
| Fourier3 | 30/30 |
| LongestHeadRun | 30/30 |
| PeriodsInStrings | 30/30 |
| HammingWeight2 | 30/30 |
| HammingCorr | 30/30 |
| HammingIndep | 29/30 |
| Run of Bits | 30/30 |
| AutoCorr | 30/30 |
The device has been physically unchanged since deployment in October 2010. It may therefore be assumed that possible effects of component aging or detuning are not an issue after 20 months of continuous operation and that the randomness of output data is not compromised by the permanent use of the device. It may also be worth noting that there was no report of test failures from any of the over 700 registered users of the QRNG server.
Funding by BMBF (KEPHOSI) and DFG (SFB787) was acknowledged.