HTML5 Webook

99/258

sig nal generated by the clock laser and the nearest comb line with index . Two chan nels of a zero-dead-time counter (K+K Messtechnik FXE) are dedicated to monitoring and Δ . e mea sure ments show fractional uncertainty contributions from and to be close to 1×10 at 1 s averaging time. Since+=++Δ+δν ,(2)the comb repetition rate then reproduces the optical clock’s stability. It is detected by a photo-detector (Discovery Semiconductors DSC40S) with a 3 dB band width of 15.6 GHz, using the comb output spectrum >1000 nm extracted by a dichroic mirror. e resulting radio-frequen-cy signal consists of a series of harmonics of the repetition rate, with more than 60 dB signal-to-noise ratio (at 300 kHz RBW) up to 10 GHz. Adjusted for mea sure ment bandwidth and considering only white phase noise, this corresponds to a power spectral density (PSD) of ℒ=−110 dBc/Hz . e repetition rate signal is down-mixed with the 9.2 GHz signal derived from the hydrogen maser, and the resulting com po nent at =50 MHz is amplied to 13 dBm for de tection. e output signal shows a phase noise level equivalent to ℒ=−90 dBc/Hz . e corresponding single sided fractional frequency PSD()=⋅2⋅10ℒ()=2×10 ,(3)allows the resulting stability limit for an investigated signal frequency to be calculated as an Allan variance()=2()()() .(4)For a detection bandwidth of =2.5 MHz realized by a tunable cavity lter, this yields ()=0.019⋅() . For detection at the fundamental of the repetition rate ==250 MHz , this would give (1)=7.8×10 . However, by detecting the repetition rate at a har mo nic =37 , the increased =37⋅250 MHz=9 250 MHz reduces this to (1)=2.1×10 . e frequency is measured on the third channel of the dead-time free fre-quency counter. e counter samples the phase every ms and averages the results over a 1 s sampling window to further reduce the phase-noise in duced instability.e synchronous measurements of Δ and , both of which show instabilities of < 50 mHz at 1 s averaging time, clearly indicate any cycle slips in the stabilization of the frequency comb, which typically result from disturbances of the clock laser frequency. By referencing the frequency counter to a 10 MHz signal provided by the maser, the mea sure ment of the clock laser frequency then includes the frequency deviation HM of the hydrogen maser as+(9.2 GHz+) +Δ(1+) .(5)Combined with =429 228 004 229 873.00(17) Hz as recommended by the International Committee for Weights and Measures (CIPM) for 87Sr as a secondary representation of the second in 2017 [25], this directly yields .Measurement stabilitye instability of and relevant contributions are shown in Fig.2 in terms of the fractional overlapping Allan deviation. For τ<10 s we observe an initial slope ()≈1×10(1⁄) , representing a dominant phase noise contribution. We nd this noise to be largely common-mode between the described frequency mea sure-ment system and a separate system detecting the fourth harmonic of the repetition rate, and expect that it re presents noise either present on the maser signal itself, or introduced in the transfer to the frequency comb laboratory.Over the interval from 10 s to 10 s , the Allan devia-tion is best described by ()=3.4×10 (10⁄ s). , with a slope close to the expectation for white frequency noise (WFN). We extrapolate this contribution according to (⁄ s) to nd a statistical uncertainty =3.9×10 for representing the total avail able data obtained during a 10 d measurement cam-paign.At the observed instability, a more convenient averag-ing time of 16 h ( ) is sucient to characterize the maser frequency to a statistical uncertainty of =1.4×10 , well below the limit of 2.0×10 set by the GPS link even for a 35 d evaluation period.In the following, we will show that such limited mea-sure ment times suce to compare NICT-Sr1 to TAI with an un certainty of 3×10 , on par with the uncertainties reported by the best cesium fountains, and sucient to test the recommended frequency of 87Sr as a secondary repre-sentation of the second beyond its stated uncertainty of =4×10 .4934-4　光 – マイクロ波リンクとTAI校正