Comparison of ultra-stable BVA OSCillators

International audienceThis paper reports on precision short-term frequency-stability measurement of ultra-stable 5 MHz BVA oscillators carried out at the Institute of Photonics and Electronics (IPE) in June and October 2007. We measured two groups of Oscilloquartz (OSA) 8600/8607 series oscillators all of them with the flicker frequency modulation (FFM) floor on the order of 10-14 in terms of Allan deviation, σy(τ). The measurement concerned four oscillators in June (serial numbers 199, 291, 315 and 543) and five in October (the same four plus sn567). Thus by comparing the results obtained in these two campaigns along with the ones obtained in February 2006, we could assess long-term measurement repeatability. Of prime interest was the comparison of two best oscillators, sn543 and sn567, made to us available by courtesy of OSA. In October 2007 we performed sixteen measurements on this pair in different periods of day and different days of week. From the ensemble statistics of sixteen pair stabilities [σy(τ)543,567]i we have obtained the mean FFM floor σy(τ)543,567 = 6.3x10-14, with rms of 2.5x10‑15, minimum of 5.7x10-14 and maximum of 6.7x10-14 (taken at τ = 5 s where FFM predominates). Considering the same contribution from each oscillator, we obtain the mean σy(τ) = 4.5x10-14 for each oscillator. To estimate the inherent or “true” FFM floor of the oscillators discussed we rather use the minimum pair value than the mean, for which we obtain the individual FFM floor as low as 4.0x10‑14. A preliminary three-cornered hat analysis against sn199 (which was the third best out of the five oscillators) shows that the sn543 has a slightly better FFM performance than sn567. We will elaborate on this in the full paper. Thus it is most likely that the OSA's sn543 is currently world's best 5 MHz quartz oscillator regarding the FFM floor. The measurement was performed in a laboratory housed in an underground shielded vault specially assigned for precision stability measurements. We used a laboratory comparator based on the classical dual-mixer time-difference multiplication (DMTDM) with 5 Hz beat note frequency. Each measurement provided 10,000 samples at basic sampling interval τ0 = 200 ms. When tested with two signals power-split from one BVA oscillator (i.e. with correlated signals), the comparator shows a background flicker phase modulation of 7x10-15 at 1 s which can be neglected in the FFM region. We cannot verify, however, that the DMTDM background noise manifests itself the same way with two uncorrelated signals. The measured oscillators were put in additional shielding boxes with an arrangement for fine tuning to achieve synchronization needed for the DMTDM comparison. During the measurement the oscillators were operating with a battery power supply. The measurement system allowed to measure only one oscillator pair at a time