Topic Understanding and parameter tting of STA/LTA trigger
algorithm
Author Amadej Trnkoczy (formerly Kinemetrics SA);
E-mail: koczy@siol
Version September 1999
1 Introduction
By introducing digital ismic data acquisition, long-term continuous recording and archiving of ismic signals has become a demanding technical problem. A ismic network or even a single ismic station operating continuously at high sampling frequency produces an enormous amount of data, which is often difficult to store (and analyze) locally or even at the recording center of a network. This situation has forced ismologists to invent triggered ismic data acquisition. In a triggered mode, a ismic station or a ismic network still process all incoming ismic signals in real time (or in near-real-time) but incoming data is not stored continuously and permanently. Processing software - a trigger algorithm - rves for the detection of typical ismic signals (earthquakes, controlled source i
smic signals, underground nuclear explosion signals, etc.) in the constantly prent ismic noi signal. Once an assumed ismic event is detected, recording and storing of all incoming signals starts. It stops after trigger algorithm 'declares' the end of the ismic signal.
Automatic trigger algorithms are relatively ineffective when compared to a ismologist's pattern recognition ability during reading of ismograms, which is bad on years of experience and on the enormous capability of the human brain. There are few exceptions, where the most complex detectors, mostly dedicated to a given type of ismic signals, approach to human ability. In all practical cas, automatic trigger loo some data on one side and generate fally triggered records, which are not ismic signals, on the other. Small amplitude ismic signals are often not resolved from ismic noi and are therefore lost for ever, and, if the trigger algorithm is t nsitively, fal triggers are recorded due to irregularities and occasionally excessive amplitude of ismic noi. Fal triggers burden off-line data analysis later and unnecessarily occupy data memory of a ismic recording system. As a result, any triggered mode data acquisition impairs the completeness of the recorded ismic data and produces some additional work to delete fal records.
Several trigger algorithms are prently known and ud - from a very simple amplitude threshold ty
pe to the sophisticated pattern recognition, adaptive methods and neural network bad approaches. They are bad on the amplitude, the envelope, or the power of the signal(s) in time domain, or on the frequency or quency domain content of ismic signal. Among the more sophisticated ones, Allan's (1978; 1982) and Murdock and Hutt´s (1983) trigger algorithms are probably the most commonly known. Many of the algorithms function in association with the ismic pha time picking task. Seismic array detection algorithms fall into a special field of rearch, which will not be discusd here. For more advanced algorithms e, e.g., Joswig (1990; 1993; 1995). However, in practice, only relatively simple trigger algorithms have been really broadly accepted. and can be found in ismic data recorders in the market and in most network's real time processing packages. The simplest trigger algorithm is the amplitude threshold trigger. It simply detects any amplitude of ismic signal exceeding a pre-t threshold. The recording starts whenever this threshold is
reached. This algorithm is rarely ud in weak-motion ismology but it is a standard in strong motion ismic instruments, that is in systems where high nsitivity is mostly not an issue, and where conquently man-made and natural ismic noi amplitudes are much smaller than the signals which are suppod to trigger the instrument.
The root-mean-square (RMS) threshold trigger is similar to the amplitude threshold algorithm, except that the RMS values of the amplitude in a short time window are ud instead of 'instant' signal amplitude. It is less nsitive to spike-like man-made ismic noi, however it is rarely ud in practice.
Today, the ‘short-time-average through long-time-average trigger' (STA/LTA) is the most broadly ud algorithm in weak-motion ismology. It continuously calculates the average values of the absolute amplitude of a ismic signal in two concutive moving-time windows. The short time window (STA) is nsitive to ismic events while the long time window (LTA) provides information about the temporal amplitude of ismic noi at the site. When the ratio of both exceeds a pre-t value, an event is 'declared' and data starts being recorded in a file. Several more sophisticated trigger algorithms are known from literature (e.g., Joswig 1990; 1993; 1995) but they are rarely ud in the ismic data loggers currently in the market . Only some of them are employed in the network's real time software packages available. When in the hands of an expert, they can improve the events/fal-triggers ratio significantly, particularly for a given type of ismic events. However, the sophisticated adjustments of operational parameters to actual signals and ismic noi conditions at each ismic site that the triggers require, has proven unwieldy and subject to error i
n practice. This is probably the main reason why the STA/LTA trigger algorithm still remains the most popular.
Successful capturing of ismic events depends on proper ttings of the trigger parameters. To help with this task, this Information Sheet explains the STA/LTA trigger functioning and gives general instructions on lecting its parameters. Technical instructions on tting the trigger parameters depend on particular hardware and software and are not given here. Refer to the corresponding manuals for details.
2 Purpo
The short-time-average/long-time-average STA/LTA trigger is usually ud in weak-motion applications that try to record as many ismic events as possible. The are the applications where the STA/LTA algorithm is most uful. It is nearly a standard trigger algorithm in portable ismic recorders, as well as in many real time processing software packages of the weak-motion ismic networks. However, it may also be uful in many strong motion applications, except when interest is limited to the strongest earthquakes.
The (STA/LTA) trigger significantly improves the recording of weak earthquakes in comparison with a
ideo
mplitude threshold trigger algorithms. At the same time it decreas the number of fal records triggered by natural and man-made ismic noi. To some extent it also allows discrimination among different types of earthquakes.
The STA/LTA trigger parameter ttings are always a tradeoff among veral ismological and instrumental considerations. The goal of arching for optimal parameter ttings is the highest possible ismic station nsitivity for a given type of ismic signal (which may also includes the target 'all earthquakes') at a still tolerable number of fal triggers.
The STA/LTA trigger is most beneficial at ismically quiet sites where natural ismic noi (marine noi) is the dominant type of ismic noi. It is also effective in ca of changes of 'continuous' man-made ismic noi. Such changes, for example, occur due to day/night variation of human activity nearby or in urban areas. The STA/LTA algorithm is less effective in the prence of irregular, high amplitude man-made ismic noi which is often of burst and/or spike type.tell me a lie
3 How it works - basics
The STA/LTA algorithm continuously keeps track of the always-prent changes in the ismic noi amplitude at the station site and automatically adjusts the ismic station's nsitivity to the actual
ismic noi level. As a result, a significantly higher nsitivity of the system during ismically quiet periods is achieved and an excessive number of fally triggered records is prevented, or at least mitigated, during ismically noisy periods. Calculations are repeatedly performed in real time. This process is usually taking place independently in all ismic channels of a ismic recorder or of a ismic network.
The STA/LTA algorithm process filter ismic signals (e ction 5.1 'Selection of trigger filters' in this Information Sheet) in two moving time windows – a short-time average window (STA) and a long-time average window (LTA). The STA measures the 'instant' amplitude of the ismic signal and watches for earthquakes. The LTA takes care of the current average ismic noi amplitude.
First, the absolute amplitude of each data sample of an incoming signal is calculated. Next, the average of absolute amplitudes in both windows is calculated. In a further step, a ratio of both values — STA/LTA ratio—is calculated. This ratio is continuously compared to a ur lected threshold value - STA/LTA trigger threshold level. If the ratio exceeds this threshold, a channel trigger is declared. A channel trigger does not necessarily mean that a multi-channel data logger or a network actually starts to record ismic signals. All ismic networks and most ismic recorders have a 'trigger voting' mechanism built in that defines how many and which channels have to be in a trigger
兴趣爱好英文
ed state before the instrument or the network actually starts to record data (e ction 5.4 below - 'Selection of voting scheme parameters'). To simplify the explanation, we shall obrve only one signal channel. We will assume that a channel trigger is equivalent to a network or a recorder trigger.
After the ismic signal gradually terminates, the channel detriggers. This happens when the current STA/LTA ratio falls below another ur-lected parameter - STA/LTA detrigger threshold level. Obviously, the STA/LTA detrigger threshold level should be lower (or rarely equal) than the STA/LTA trigger threshold level.
In addition to the data acquired during the 'trigger active' time, ismic networks and ismic recorders add a certain amount of ismic data to the event file before triggering – pre-event-time (PEM) data. After the trigger active state terminates, they also add post-event-time (PET) data.
refer toFor better understanding, Figure 1 shows a typical local event and the trigger variables (simplified) during STA/LTA triggering. Graph a) shows an incoming continuous ismic signal (filtered); graph b) shows an averaged absolute signal in the STA and LTA windows, respectively, as they move in time toward the right side of the graph; and graph c) shows the ratio of both. In addition, the trigger active state (solid line rectangle), the post-event time (PET), and the pre-event time (PEM) (dotted line rectangles) are shown. In this example, the
trigger threshold level parameter was t to 10 and the detrigger threshold level to 2 (two short horizontal dotted lines). One can e that the trigger became active when the STA/LTA ratio value exceeded 10. It was deactivated when the STA/LTA ratio value fell below 2. On graph d) the actually recorded data file is shown. It includes all event phas of significance and a portion of the ismic noi at the beginning.
In reality, the STA/LTA triggers are usually slightly more complicated, however, the details are not esntial for the understanding and proper tting of trigger parameters.
犯罪心理第9季
Figure 1 Function and variables of STA/LTA trigger calculations (e text for explanations).
4 How to adjust STA/LTA trigger parameters
To t the basic STA/LTA trigger algorithm parameters one has to lect the following: •STA window duration
•LTA window duration
•STA/LTA trigger threshold level
•STA/LTA detrigger threshold level.
However, optimal triggering of a ismic recorder or a ismic network does not depend only on the parameters. There are usually four additional associated parameters which, only if well tuned with the trigger parameters, guarantee optimal data recording. The parameters are:
•trigger filters
•pre-event time (PEM)
怎么样画妆•post-event time (PET)
min是什么意思•trigger voting scheme.
Although not directly related to the STA/LTA trigger algorithm, the additional parameters are also be discusd below in order to provide a complete information.
The STA/LTA trigger parameter and associated parameters’ ttings depend on the goal of the application, on the ismic noi condition at the site, on the properties of ismic signals at a given location, and on the type of nsor ud. All the issues vary broadly among applications and among ismic sites. Obviously, there is no general, single rule on tting them. Each application and every ismic site requires some study, since only practical experience enables the determination of really optimal trigger ttings.
Note that ismic recorders and network software packages come with a t of default (factory t) trigger and trigger associated parameter values. They are rarely optimal and must therefore be adjusted to become efficient in a particular application. For best results, changing the parameters and gradually finding the best ttings is a process which requires a certain amount of effort and time.
柬埔寨语
4.1 Selection of short-time average window (STA) duration
公司管理模式Short-time average window measures the 'instant' value of a ismic signal or its envelope. Generally, STA duration must be longer than a few periods of a typically expected ismic signal. If the STA is too short, the averaging of the ismic signal will not function properly. The STA is no longer a measure of the average signal (signal envelope) but becomes influenced by individual periods of the ismic signal. On the other hand, STA duration must be shorter than the shortest events we expect to capture.
To some extent the STA functions as a signal filter. The shorter the duration lected, the higher the trigger’s nsitivity to short lasting local earthquakes compared to long lasting and lower frequency distant earthquakes. The longer the STA duration lected, the less nsitive it is for short local earthquakes. Therefore, by changing the STA duration one can, to some extent, prioritize capturing of distant or local events.
ihtcThe STA duration is also important with respect to fal triggers. By decreasing the duration of the STA window, triggering gets more nsitive to spike-type man-made ismic noi, and vice versa. Although such noi is usually of instrumental nature, it can also be ismic. At the sites highly pollut
ed with spike-type noi, one will be frequently forced to make the STA duration significantly longer than the spikes, if fal triggers are too numerous. Unfortunately, this will also decrea the nsitivity of the recording to very local events of short duration. Figure 2 explains the effect of STA duration on local events and spike-type noi.
