Explanation of Seismograph Traces
In Progress
About the seismometer
Signal processing
The raw, unprocessed signal is logged at 100 Hz (100 points per second) and stored in data files, which can be processed at a later date. All data is archived on www.geol.pdx.edu/seismos, and on CD. Data (and a data access subroutine, in C++) since August 2000 are available from www.geol.pdx.edu/seismos or by request.
Ideally, we should be able to take the signal from the seismometer and display it. However, because of building noise, electrical interference, etc, we need to filter the signal. This can be done many ways, but for display purposes we often use a "leaky integrator" type of filer (e.g., Claerbout, J.F., 1976. Fundamentals of Geophysical data processing , McGraw-Hill, NJ, 24p.). This is a simple way of removing much of the high-frequency noise. The basic algorithm is:
yt = (1-e) yt-1 + xt
e is picked to give a reasonable looking display for our setup. We use a value of 10-3.
A more sophisticated analysis of the data may be performed, the above algorithm is used for real-time display of data since it requires very little computational effort. For plotting purposes a scale of +/- 2 units is normally used to show the traces. For large events in the NW this can cause the plot to become clipped. This is easily remedied by plotting at a different scale.
| Unfiltered Data (daily plot) | Filtered Data (daily plot) |
|
Unfiltered Data (2 hours of data) The vertical scale here is 25 times larger than the filtered data. If the same scale as the filtered plot was used the entire plot would be overlapping strips of solid color! |
Filtered Data (2 hours of data) The vertical scale for each trace is 25 times smaller than in the unfiltered example to the left. |
Type of events recorded in the seismographs
Disturbing the Seismometer
As the "pendulum" of the seismometer swings back and forward, an electrical current is produced. The current is proportional to the speed the pendulum is swinging. The mechanism has a "dampening" mechanism, so the pendulum will not oscillate wildly during an earthquake. If the seismograph is disturbed by an impulse (e.g., blowing gently on the pendulum weight, or someone slamming the door of the room where the seismometer is housed) the recorder will show a rapid increase in voltage, and then a slow decline back to the zero voltage level. The pendulum will swing past this point before returning to zero.
This pattern can often be seen on some of the daily records. Some examples are shown below (click on image for a larger version)
| Impulse -- Daily plot scale |
Filtered Impulse (using Leaky Integrator) - 1 minute between vertical dashed lines. The seismometer was deflected to produce the abrupt down motion on the recorder (some data clipping took place during the plotting - the raw data file contains the entire range). The pendulum then slowly swung back past the center mark before returning to the zero level. This represents the dampening in the system |
Electrical noise or Mechanical vibration
The system, being housed in an active building is prone to electrical noise beyond the background noise. Often at the end of a day (4 - 7 pm local time, 00 - 03 UTC) there appears to be some electrical noise that is picked up by the system (we are still trying to trace the source ...). This shows up in the following way:
| Electrical noise -- Daily plot scale | Electrical noise- 2 hour plot scale |
Building and city noise
Comparing several seismographs, and the same seismographs for different time periods gives you an idea of activity in the building and downtown Portland. The quietest day in any week is usually Sunday. The quietest part of any given day is usually between 10 pm and 6 am. This is when much of the mechanical infrastructure in the building (mostly air conditioning) is off (is is usually also off Saturday afternoon through Monday AM). The following two plots illustrate some of the variations depending on time:
| Building noise -- weekday | Building noise- Sunday |
Earthquakes
Last, but not least, are earthquakes. Earthquakes can look very different depending on the scale of the plots (see examples below). The seismometer we have is designed to look for long-period oscillations, so some of high-frequency portions of an earthquake will not be recorded.
