Fig. 4: RF clock performance.

a, The fractional Allan deviation of the clock without (orange) and with (green and purple) noise suppression, the extracted interferometric noise (yellow) and the optical reference (blue). b, Simplified dual-comb setup portraying the interferometric noise of this configuration. The blurring of the different comb lines represents frequency jitter due to interferometric phase noise. c, The noise suppression scheme exploiting dual clock signals (the dashed yellow boxes indicate the two optical filters to select them) generated using distinct photodetectors driven by main-Vernier comb line pairs at frequencies higher and lower than the pump frequency. The dashed comb lines illustrate a hypothetical case when the pump to the main ring (green line) and the resulting main ring comb lines (blue) are shifted up in frequency by time-varying phase ϕ₁(t), while the pump to Vernier ring and resulting Vernier comb lines are unaffected (ϕ₂(t) = 0). The frequency noise of the RF beats generated from the higher and lower comb line pairs is strongly correlated but of opposite signs. Therefore, by summing the two clock signals with an electronic mixer, the interferometric frequency noise is largely suppressed. d,e, Frequency counter traces of the 871 nm laser (d) and the RF clock (e) with noise suppression. The 871 nm laser is initially free-running, then is locked at around 40 s. The f_xCEO and f_{V ernier} locks are on throughout the measurement. The y-axis spans of the optical reference and the RF clock differ by their scaling factor of 17,292 × 84. The stabilized clock output exhibits a mean frequency of 235,070,310.72 Hz with a standard deviation of 0.18 Hz, limited by the counter measurement.