This paper presents a new, broadband (from 0.2 mHz to 18 Hz) estimate of the self-noise of the Streckeisen STS-2 sensor based on the application of the three-channel correlation technique (Sleeman et al., 2006) to one year of continuous seismic recordings. A novel presentation of instrumental self-noise is shown using a probability density function. The measurements were obtained from three collocated sensors in the Conrad Observatory (Austria); the sensors are covered with a new type of thermal insulation consisting of thin layers of neoprene, which significantly reduces noncoherent
signals between the sensors. We observe a noise reduction ranging from about 10 dB at 1 mHz to 2 dB at 1 Hz when compared to noise estimates from noninsulated sensors. The new self-noise estimate for the STS-2 falls below the new low-noise model (NLNM; Peterson, 1993) at frequencies above 4 mHz and even reaches the NLNM at frequencies between 0.2 mHz and 4 mHz during optimal conditions.
From our observations, we conclude that the broadband STS-2 sensor can achieve
a noise power level of 1.56 × 10e-18 m^2/s^4/Hz, or -178 dB, at 1000 s. We also show results from synthetic experiments to quantify the effect of sensor misalignment on the self-noise estimate and to quantify the misalignment between the sensors in our Conrad experiment. The synthetic experiment shows that the three-channel correlation technique potentially can extract noise from signals with signal-to-noise ratio (SNR) up to 140 dB but is in practice limited to signals with 60–80 dB SNR when the alignment errors are on the order of 0.2° and 0.02°, respectively. From the observations, we conclude that the misalignment between the sensors in the Conrad experiment is on the order of 0.2°.