Summary
This metric cross-correlates two channels from the same network, station and location code and reports the zero-lag correlation coefficient. Data windows for comparison include events of magnitude 5.5 or greater; they begin 2 minutes before the event origin time and have a 30-minute duration.
Uses
When the zero-lag correlation coefficient is near one, it indicates common signal that could mean leakage in an instrument’s cabling from one channel to another. It would consistently be unity if a channel configuration error caused the same channel to be sent twice under different channel names. Triaxial sensor drift can appear as an episode of strong correlation between channel pairs. Thermal noise will show strong correlation between horizontal channels.
Data Analyzed
Events – magnitude >= 5.5
Traces – two different Channels from the same N.S.L (Network.Station.Location) per measurement
Window – 30 minutes beginning 2 minutes prior to the origin time of the event
Data Source – IRIS miniSEED archive or IRIS PH5 archive, and USGS event service
SEED Channel Types – ?H?, ?P? | High Gain, Geophone
Algorithm
- Request events for the current day with magnitude >= 5.5,
- For each N.S.L having multiple channel components,
- Request 30 minutes of time series data starting 2 minutes before the origin time for available Channels,
- For each channel combination,
- Cross-correlate the channel pair time series,
- Report the value of the correlation function at zero lag:
Rxy = (1/N) * sum( x(n)*y(n) ) where N = # of samples and n = 1…N
Metric Values Returned
value – correlation coefficient (cross-correlation function value at zero lag)
target – the trace analyzed, labeled as N.S.L.C.Q (Network.Station.Location.Channel.Quality)
start – the integer second (UTC) prior to the time series window described in Data Analyzed
end – the integer second (UTC) following the time series described in Data Analyzed
lddate – date/time the measurement was made and loaded into the MUSTANG database (UTC)
Notes
This metric assumes that the best opportunity for observing cross-talk is in the presence of signal. Nevertheless, magnitude 5.5 events are not visible globally, so the absence of a large correlation coefficient does not necessarily mean that a given channel produces no cross-talk – only that it wasn’t observed during that measurement.
Because metric start and end times are listed as integer seconds and the 30-minute time series usually begin between integer seconds, start and end times have been expanded by one second at either end to ensure bracketing of the time series and predictable query results.