Last modified 16 months ago
Last modified on 05/22/12 12:47:17
carlstatrig Triggering Parameters:
Trigger ON when eta > 0.0
Trigger OFF when eta <= 0.0
eta is calculated by:
eta = STAR - (Ratio * LTAR) - ABS(STA - LTA) - Quiet
eta = Term1 - Term2 - Term3 - Term4
where:
STA = the short-term (1 second) average of the trace.
Represents the current DC offset of the trace.
LTA = an 8 second average of the STA.
Represents the long-term DC offset of the trace.
STAR = short-term rectified average; the absolute value of the
difference between the trace and the LTA, averaged for one second.
Represents how much signal there is with respect to DC offset.
LTAR = long-term rectified average; the 8 second average of STAR.
Quiet stations have low LTAR.
Noisy stations have higher LTAR.
Ratio = configurable parameter; as Ratio gets smaller, carlstatrig gets more
sensitive (more and longer triggers).
Ratio affects noisy stations more than quiet ones.
Quiet = configurable parameter; as Quiet gets smaller, carlstatrig gets more
sensitive (more and longer triggers).
Quiet affects all stations equally.
"eta" is calculated once a second for every station independently.
"eta" is calculated using the current second's updated STAs with the
previous-second's LTAs. After eta is calculated, LTAs are updated.
In a sense, the LTAs are lagged by 1 second.
Term1: STAR
------------
Term1, STAR, increases as the amplitude of the signal increases, thus
making a trigger more likely. This term has no configurable elements, but
is totally data-determined.
Term2: Ratio * LTAR
--------------------
Term2 affects how high above background noise a signal has to be in
order to declare a trigger. The constant "Ratio" in Term2 is configurable.
For a perfectly flat trace, Term2 and Term3 are zero and triggering is
controlled by Term4 alone. Normal quiet traces have a low LTAR, noisy traces
have a higher LTAR. Changes to "Ratio" therefore have a larger impact on
noisy traces than on quiet traces. As "Ratio" gets larger, larger amplitude
signals are required to turn the trigger ON.
Since the LTAR is an 8-second average, lagged by only 1 second from the
STAR, it "catches up" with the STAR rather quickly during an extended event.
Therefore, triggers will be ON for at most 8 seconds (more likely 4-5 seconds)
at an individual station, regardless of the total duration of the signal.
The total duration of an event (in a network sense) is therefore controlled
more by the distribution of triggered subnets than by the total duration of
a trigger at individual stations.
Term3: ABS( STA - LTA )
------------------------
Term3 works to eliminate triggers from one-sided signals, which
are usually noise signals. This term has no configurable elements, but is
totally data-determined.
Balanced (well-centered) signals should have similar STA & LTA values,
regardless of the signal's amplitude. Therefore, Term3 should be close
to zero for centered signals.
One-sided signals will have different values for STA & LTA. Depending
on the sense of one-sidedness, STA could be larger or smaller than LTA;
however, Term3 is the absolute value of the difference. Therefore, Term3
will have a positive value (reducing chance of trigger) when a one-sided
signal (usually noise) occurs. If a station is unhealthy and always produces
one-sided signals, Term3 will always inhibit triggers during real seismic
events.
Term4: QUIET
-------------
Term4, QUIET, is a configurable constant representing how large a signal
is required to initiate a trigger. For a trace that is perfectly quiet
(totally flat so that Term2=Term3=0) except during an event, Term4 is the
maximum 1-second average absolute amplitude (STAR) allowed for an untriggered
trace. That is, as soon as STAR exceeds QUIET, the trace is considered
triggered.
Changes to Term4 affect all traces equally. If QUIET is increased,
triggers require larger signals and visa versa.
