A Single Bit Matters: Coherent Noise of Seismic Data Loggers

Cross‐correlating seismic random signals, such as coda waves or ambient noise, at two sites can extract seismic wavefields as if a source is at one site and a receiver is at the other if certain conditions are met (e.g., Aki, 1957; Campillo and Paul, 2003). Given the recent development of dense seismic network, the last decade has witnessed the rapid emergence of imaging Earth’s local (Nagaoka et al. , 2012), regional (Shapiro et al. , 2005; Nishida et al. , 2008; Lin et al. , 2011), and global (Nishida et al. , 2009) seismic structures and their temporal changes (Brenguier, Campillo, et al. , 2008; Brenguier, Shapiro, et al. , 2008; Nagaoka et al. , 2010; Brenguier et al. , 2014) from these random signals. With this trend of research, we fortuitously found peculiar pulses in the cross correlations of the Japanese high‐sensitivity seismograph network (Hi‐net; Obara et al. , 2005) seismic records (Fig. 1). Here, we used one‐month records for January 2012 at 142 Hi‐net stations in southwest Japan. Applying a band‐pass filter of 10–20 s to correlograms reveals sharp pulses at lag times of every 60 s (Fig. 1a). Unfiltered correlograms nicely demonstrate a propagation of Rayleigh waves, but a closer look at locations where no wavetrain is extracted reveals sharp pulses at lag times of every second (Fig. 1b). Figure 1. Correlograms of ambient seismic noise aligned along the interstation distance. (a) Correlograms with a band‐pass filter between 10 and 20 s. (b) Unfiltered correlograms of a portion where no extracted wave propagation is expected. (c,d) Same as (a) and (b) but the quantization errors are subtracted from the raw data before cross correlation. (e,f) Same as (a) and (b) but the difference of ZR and RZ correlations is taken without subtraction of the quantization error. The amplitude scales of (c,e) and (d,f) are the same as those of (a) and (b), …