XPHASE MANUAL

Table of Contents

• Name
• Version
• Synopsis
• Description
• Arguments
• Options
• Some Basic Usage Notes
• An Overview of the Procedures Used in Xphase and Xest
• Preferences and Defaults
• The Buttons on the Top Panel
• Canvas activity
• Keyboard Shorcuts
• Selection Boxes
• Canvas Activity Help
• Screen Scale
• Option Windows
• Overlay
• X Axis
• Characteristic (K) functions
• Absolute and Relative time
• Notes on Determing Polarities
• Advance Usage Notes
• On the TO DO List
• See Also
• Author
• Acknowledgments
• Bug Reports

NAME

Xphase - interactive phase picking GUI software

VERSION

These notes pertain to version 2001.1111

SYNOPSIS

There are two general invocations of Xphase. The first reads in segy, sac, ah, or ascii files:

Xphase ( SEGY_files or -s SAC_files or -a AH_files or -c ASCII_files ) [ -p pick_file ] [ -t timemark_file ] [-g] [-r] [-h]

The second uses conventions similar to a DataScope style database:

Xphase ( -db DB_name -sta station -tstr start_time -tend end_time -mode imode) [ -p pick_file ] [ -t timemark_file ] [-g] [-r] [-h]

DESCRIPTION

Xphase is a program used to interactively run FORTRAN subroutines developed by A. Kushnir and others for autoregressive analysis of time series. Useful functions include phase detection, estimation of onset times, and reconstitution filtering. Xphase also provides a useful tool for testing arguments to the batch version of this software, called Xest (which see).

Xphase was developed on a SUN SPARCstation LX running Solaris 2.3, and compatibility with other environments should not be assumed. It is a fairly straightforward program, however, so modifications should not be extensive. Xphase requires a color monitor to work properly.

ARGUMENTS

In the first invocation, the only argument is the file names. Wild cards (? and *) can be used.

SEGY_files	Specifies the SEGY trace files containing seismic data.
-s SAC_files	Read in SAC formatted data rather than SEGY.
-a AH_files	Read in AH formatted data rather than SEGY. Note the the routine that reads in the AH data is not the standard LDGO AHIO routines but a cludge based on an interpretation of some AH seismic data that I had available. Therefore it may not work on all datasets.
-c ASCII_files	Read in ASCII data rather than SEGY. Format is based the ASCII format of Klaus Stammler's SHC. SHC ASCII files have a few header variables followed by a sequence of amplitudes. At minimum the header contains the time interval (DELTA) and the number of samples (LENGTH) and may optionally include the time of the first sample (START), the station code (STATION) and the component (COMP). A suitable file can be produced in SHC by issuing the command: writea/npl=1 1 start station comp recall that the outfile name will appear as all caps (this is the SHC convention). Here, we expect the header to have 5 lines in the order of DELTA, LENGTH, START, STATION, and COMP. Here's an example: DELTA: 5.000000e-02 LENGTH: 46520 START: 5-OCT-1997_18:21:33.000 STATION: ANA COMP: E 0.23 0.45 -0.33 ... which says the sample interval is 0.05 s with 46520 samples in the file. The time of the first sample is 5-OCT-1997_18:21:33.000, the station is ANA and the component is E (east).

In the second invocation, several arguments must be sepecified:

DB_name	Specifies the databas name, a la DataScope.
-sta station	The code of the station for which data is to be uploaded. Note that only one station can be used.
-tstr start_time	The start time of the data. Note that any of the DataScope conventions for specifying time (e.g., epoch or date) can be used.
-tend end_time	The end time of the data. Note that any of the DataScope conventions for specifying time (e.g., epoch or date) can be used.
-mode imode	Specifies the number of channels to be used: imode = 1 (Z) imode = 2 (N,E) imode = 3 (ZNE)

OPTIONS

-g	For SEGY data, do not convert from digital counts to volts. The SEGY data is presumed to be in digital counts, and the default is the convert this number to volts using the scaleing values in the header. Specifying the -g option skips this conversion.
-r	Reset the origin time so the the first sample of all files is at time 0:0:0.0. This is a useful option if files with very different origin times are viewed simultaneously.
-h	Prints out the usage summary.
-p pick_file	Causes a pick formatted file (see discussion below) by the name of pick_file to be read in for display.
-t timemark_file	Causes a timemark formatted file (see discussion below) by the name of timemak_file to be read in for display.

SOME BASIC USAGE NOTES

READING IN TIME SERIES (SEISMOGRAMS)

Xphase will read in DataScope, SEGY, SAC, AH, and SHC-style (i.e., Seismic Handler) ASCII files from the command line. Simple ASCII files may also be read in by activating the "READ" button and selecting "Ascii Time Series". All other files read in that have have references to particular times (pick files, for example) are converted to the nearest sample.

Note on filename conventions: In some routines the program follows PASSCAL file naming conventions in its operation. In particular. the program expects that the last field is related to the channel and the next to last field is the station code. If the channels are numbered, then because the vertical component is usually channel one, the vertical will be the first component read in. For example, in the pick file line shown below, the file name is "22.17.7360.1" so when records are read in with a "*". The program takes the 7360 to be the station code, and the "1" will make this the first file read in. Note that later conventions use, for example, BHZ, BHE, and BHN to label the vertical, E-W, and N-S components, and in this case the wild card read will read the vertical in last. The users needs to be aware of when this is important (see below). The channel order is easy to circumvent (just read them in a different order or reorder them using the selection boxes - see below), but the station name convention is hard wired. Proper station identification can be a useful tool when correleating records with pick file entrees.

READING IN PICK FILES

1999 206  6 17   0.0994 0  -4.0398   0.6600  0 -1 1999206055828.47.BBU.BHZ Whitened Wed Aug  2 15:01:38 2000

Where the numbers in the colums correspond to year, julian day, hour, minute, second, phase type (0 = P, 1 = S, 2 = Other), uncertainties to the left (-) and right (+) of the pick, a mode identifier used in auto picking (see discussion below), a polarity for the P phase (1 up, -1 down, 0 neither or not a P phase) an identifier showing which file was used in making the pick, and finally several columns composing a time stamp. The time stamp is not read in and is used only to separate "new" picks from "old" reads.

READING IN TIME MARK FILES

Time marks are like picks but are static (i.e., they are not edited or written). They are used primarily to show the user where certain times, for example predicted arrival times, occur. Files containing time marks are read in at the time Xphase is called if the -t option is used. The format of a time mak is exactly the same as that of a pick (see above).

READING IN FIR FILTER FILES

FIR filter files are created for specific recording devices and sample rates. Certain combinations (e.g., RefTek A08 at 100 sps) are included in the program, but the user can read in a user-defined filter and apply it. The file is just a sequence of numbers (one per line) that describes the filter.

DRAWING ON THE SCREEN

QUITTING

AN OVERVIEW OF THE PROCEDURES USED IN XPHASE AND XEST

The main functions of Xphase and Xest are (1) Data whitening, (2) Phase Detection, (3) Onset Estimation, and (4) Automatic Detection and Onset Estimation. Starting with version 2001.1030, Xphase also adjusts picks using least-squares waveform correlation. In addition, Xphase can calculate characteristic functions of particle motion to determine useful indicators of polarity, rectilinearity, and other phase-related properties.

The following provides some details about how these work. At the heart of much of the math behind Whitening, Phase Detection, and Onset Estimation is the autocorrelation or autocovariance function c of x which is defined as:

c(k) = sum(i=k,N) x(i)x(i-k+1)

The number of elements of c (plus one), or, equivalently, the number of covariance non-zero lags computed, is called the order of the autoregression (AR) model.

Whitening and Phase Detection are both accompished by the FORTRAN subroutine grbeamfep, and is driven by the routines rungrbeam (in Xphase) and rungrbeam2 (in Xest). Both take an option argument that tells the program how far to progress: iopt = 1 for stacking only; iopt = 2 for whitening only, and iopt = 3 for whitening and detection.

Data whitening

Data whitening is accomplished by calculating a filter which estimates the inverse spectrum of the noise. A sample of noise is provided by the user (for example, by specifying a noise window). The covariance vector (y) is expanded into a Toeplitz type matrix (a) and then the filter is produced by calculating a^-1y. This filter is then applied to the time series in the user defined data window.

Besides the noise and data windows, the only variable of significance that the user supplies for whitening is the AR order (or number of non-zero covariance lags).

Phase Detection

Phases are detected by first grabbing the data within the user-defined data window, whitening it with a filter computed from the user-defined noise window, and then passing a moving window through the whitened data. At each position, the autocovariance function is calculated for all points within the moving window (=mvwind) at a user-defined number of lags (or AR order =arm). The values of the detection function are the chi-square measure of the variance in the autocovariance vector (also called the dispersion of the AR residuals), defined in this case by

chisqvar = sum(i = 1, m+1) c(i)**2

where m is the AR order. This number (chisqvar) is assigned to the midpoint of the moving window.

The most important user-definded variables for detection are the moving window length ("mvwind") and the lag number/AR order ("arm").

Onset Estimation

Onsets are estimated by defining a data window and calculating a similarity function for all points to the left and right of any given point in the data window. As with phase detection, the user defines a data window and an AR order ("arm"). Additionally, the user can set a space between the ends of the data window and the times over which the estimation function is generated ("arwind"). The functions eston1d and eston3d then compute autocovariances for windows to the left and right of each point in the estimation window and a similarity estimate is computed from the ratio of the dispersion in the covariances. The point of maximum dissimilarity is taken as the onset.

Phase picks can be adjusted using waveform correlation. Here's a step by step procedure:
1. You must have an existing pick for each of the waveforms you want to correlate. It doesn't have to be a good one, but of course the better it is the easier the job will be. Read these in as a pick file or generate picks on the screen using any of the many techniques for doing so.

2. Change from Absolute to Relative time (Using the appropriate Canvas Activity, accessed through the "A" key, or just hit the "a" key. This will line the perferred phases up (default is P phases, but you can change this in the OPTIONS window).

3. Define the correlation window by setting the Data Window to include the suspected phases. The acual picks don't have to be in the window (but I hope your judgement is good enough so that they should be within the window).

4. Select the channels that you want to correlate. You must select at least two.

5. Hit the "CORREL" button. The program calculates cross correlation functions for each of the pairs of selected channels. Some statistics, like the correlation coefficient and the optimal time shift, are printed in the shell screen.

6. After the corrections have been computed, the program adjusts the picks and attempts to replot the screen at the original scale with the revised times (relative frame). Unless you have really huge offsets, this should work fine (I haven't broken it yet).

7. You can try "CORREL" again to see if any additional improvement is possible. Consider redefining the Data Window if the offsets have been significant to reduce the effects of noise. When you are satisfied, write the picks to a file.

Some notes: Correlations are calculated by FFTs of data windows, followed by multiplitions of conjugates and retransforming into time domain. The time offset is derived from the maximum of the resulting function. Offsets for multiple channels are determined in a least squares sense following the algorithm of VanDecar and Crosson (BSSA, 1990) although only the nonweighted inversion is implimented at present.

The time corrections are adjusted to the nearest sample, which means that no "subsample" corrections are allowed. This is facilitated by keeping the first waveform pick constant and moving the remaining picks relative to it. If you are interested in absolute time, therefore, make sure you have a pick you believe at the top of the stack.

A useful aid in picking S arrival times is to compare S-P times of different stations, typically using a station with clear S and P arrivals as a reference. A simple estimate of the target S time can be made as as follows: let V_p/V_s be an estimate of the ratio of P to S wavespeed for the medium. It follows that the P travel time for the reference station (T_p) can be estimated from the S-P times as T_p = (t_s - t_p)/(V_p/V_s - 1). If we have a P pick at the target station (t_p') then the travel time to that station is just T_p' = T_p + (t_p' - t_p), and the S travel time (T_s') is T_p'*(V_p/V_s). The S arrival time is then t_s' = t_p' + (T_s' - T_p'). In the program this procedure is implimented in the "S-EST" function. Given a reference station, the estimated times for three V_p/V_s ratios (min, max, and pref) are calcuated and drawn on the screen as horizontal brackets with vertical ticks indicating the estimated S times.

The characteristic functions of particle motion are similar to those described by Cichowicz for his S picking algorithm. The basic idea is to form the cross power matrix M from a window of three component motion ( u_z, u_e, u_n) as follows:

	|	{uz, uz}	{uz, ue}	{uz, un}	|
M =	|	{ue, uz}	{ue, ue}	{ue, un}	|
	|	{un, uz}	{un, ue}	{un, un}	|

where {} denotes the dot product. The eigenvectors of the above matrix give the components of principle motion, hence a window about the P arrival will allow calculation of the direction of the L component (corresponding to the largest eigenvalue). There are several useful functions that can be derived from the eigenvalues (L₁ > L₂ > L₃) and eigenvectors for samples within a given window:

Rectilinearity:

The concentration of motion in a single direction at sample i can be characterized by

k(i) = (L₁ - L₂)/L₁

Planarity

The concentration of motion in a plane at sample i can be characterized by

k(i) = (L₁ + L₂ - 2*L₃)/(L₁ + L₂)

Polarization

The polarization of motion at sample i can be characterized by

k(i) = ((L₁ - L₂)² + (L₁ - L₃)² + (L₂ - L₃)²)/ (2*(L₁ + L₂ + L₃)²)

Deflection Angle

If the eigenvector corresponding to the L (P wave motion) direction is v_p and v(i) is the eigenvector corresponding to the largest eigenvalue at sample i, then the deflection of the particle motion from the L direction can be measured by

k(i) = 2/pi * arccos{v_p,v(i)}

Fraction of Energy perpendicular to the L component

After determining the eigenvectors corresponding to the L (P wave motion) and perpendicular directions, we can rotate any motion at sample i into this frame by multipling the (u_z, u_e, u_n) vector by these eigenvalues to give (u_L, u_Q, u_T). The fraction of energy perpendicular to the P wave motion can then be characterized by

k(i) = ( u_Q² + u_T²)/ ( u_Q² + u_T² + u_L²)

Product of Characteristic functions

Finally, we can emphasize the S arrival by multiplying together those functions that are sensitive to S:

k(i) = Planarity * Polarization * Deflection Angle * Fractional Transverse Energy

Note that each of the characteristic functions has 0 < k < 1, to the product is between 0 and 1 as well.

Note that of the characteristic functions discussed above, the last 3 (Deflection Angle, Fraction of Energy perpendicular to the L component, and the Product of Characteristic functions) require information about the eigenvectors associated with the P arrival. This is implemented in Xphase by providing a Canvas function for eigenvector estimation, after which these functions may be calculated. The functions can be calculated using the "CICH FUNS" button, which uses the current data window, or by the canvas functions "Current Cichowicz", which calculates the selected function, and "All Cichowicz", which calculates all of the functions, after the choice of a data window.

Automatic Detection and Onset Estimation

Several different algorithms have been tried for automatic detection and estimation to minimize the efforts spent by an analyst to create a catalog. None of these are perfect and will require some operator intervention at some point. Certainly the number of algorithms still embedded in the main routines is testimony to the trial and error nature of this endeavor.

The first part of an automated procedure is always the detection of a phase, using the algorithm discussed above. The trick is then to figure out which parts of the phase detection function might actually be an earthquake phase. At the moment all algorithms use threshold exceedance to decide if a phase exists. The second trick is how to tell which, in a multitude of phases, are P and S arrivals. With the level of information we have, this is not a trivial matter, and when we discriminate between the two we rely on two general observations: (1) the S will arrive after the P and (2) generally (but of course not always - that would be too simple) the phase detection function has much larger values for S then for P.

Once we figure we have an arrival, we need to put a window around the detection and estimate the onset within this window. The size of this window is user-defined ("winsize").

At the moment there are two algorithms used in Xphase to do automatic detection and estimation: AUTO EST/auto_estimate2 and DETECT_EST/det_est. The main difference between these is that auto_estimate2 tries to discriminate between P and S waves while det_est picks everything that looks useful without trying to figure out the phase. Also, AUTO EST/auto_estimate2 was designed to work on triggered data, DETECT_EST/det_est with continous data.

Xest can run in a "detection only mode" (Xest -d) in which case only detections are saved (more discussion below). If the -d argument is missing, Xest currently runs an equivalent algorithm to det_est called det_est2. Xest contains a batch equivalent of auto_estimate2 called auto_estimate3, the call to which is currently commented out (rather than make this user selectable - actually a trivial thing to do).

Details on how these different algorthims work follows.

AUTO EST/auto_estimate2 (Xphase) and auto_estimate3 (Xest)

Steps:

1. Run the data within the data window through the phase detector to produce the detection function (as discussed above).

2. Determine the detection "background" level. In the current implimentation, this is done by finding the maximum of the detection function within the user-defined noise window. This number is assigned to the variable "noise_max". NB: the idea here is that a routine has an intelligent way to determine a noise window, such as the first several samples of a triggered record. This is not useful for continuously recorded data.

3. Determine the maximum of the detection function within the user-defined data window. This number is assigned to the variable "detect_max". The philosophy is that if an S wave is present, it probably is related to this maximum.

4. Set thresholds for P and S wave arrivals. Because we expect the S detection to be much larger than the P, and S to be related to detect_max, we define these as:

   	p_threshold = arg.pfact*noise_max;
   	s_threshold = arg.sfact*detect_max;
   

   arg.pfact might take a value like "100" and arg.sfact something like "0.9". In other words, if we see something exceeding the maximum of the noise by 100, we'll call it a P, and something that gets within 90% of the data window maximum we'll call an S.

   Note that if the P threshold is greater than detect_max, it gets knocked down to 95% of detect_max, and if the the P threshold is greater than the S threshold, the S threshold is set equal to the P threshold.

5. We now begin marching though the data window looking for P arrivals. Once we find a point that exceeds the threshold, we put a window (+-winsize/2) around it and compute and estimation function. Exactly how this is done depends on the "mode" specified (see the table for modes in the Preferences and Defaults section)

   NB: this is a case where channel order is important! Channel 1 is presumed vertical; 2 and 3 horizontal.

   Basically the choices for P are (raw or whitened) vertical only, or (raw or whitened) 3 components simultaneously.

6. Once the estimation function is calculated, we need to determine the maxima in the function. Note that because we are looking specifically for a P wave, we have to allow for the possibility that an S wave be within our window, so we look for the FIRST local maximum of significance rather than the global maximum for the entire function. A problem here is that the estimation function can be "bumpy" with a lot of insignificant local minima. Therefore, we smooth the function by taking a running sum of the function over a fraction of the window (set by the variable "wfrac"; default value is 0.1). If this smoothed envelope starts to change direction, we designate a local maximum of significance and call that the P wave.

7. Once we have the P wave, we now go back to get the S wave, using the S threshold. We also begin the detection window at one sample beyond the P pick.

Note: Once a P and S wave have been found, the program stops looking at the current data window. It does not try to look for other events within the same window. This kind of algorithm works better with triggered data, because a noise window is easier to define (just the first several samples after the trigger).

DETECT_EST/det_est (Xphase) and det_est2 (Xest)

Again, det_est is more suited to continuous data and simply goes through and picks everything that looks like it might be an arrival.

1. Run the data within the data window through the phase detector to produce the detection function (as discussed above).

2. Determine the detection "background" level. In the current implimentation, this is defined as:

   	noise_level = [(sum(i=1,n) y(i)**0.25))/n]**4.0
   

   The fourth power is used to deemphasize excursions.

3. Determine the maximum of the detection function within the user-defined data window. This number is assigned to the variable "detect_max".

4. Determine the threshold. In this case, we take

   	threshold = arg.pfact*noise_level
   

   or, some factor above the ambient noise level (like, for example, 100*noise). If the threshold is greater than detect_max, we terminate (no viable detections present).

5. As before we now begin marching though the data window looking for P arrivals. Once we find a point that exceeds the threshold, we put a window (+-winsize/2) around it and compute and estimation function, using whatever mode we've selected. Note that if we select one of the three channel modes, the algorithm makes a single pass through the time series. If we select a dual mode (e.g., P on vertical and S on horizontals), the algorithm makes two passes through, the first one using the vertical channel and the second using the horizontals.

6. Once the estimation function is calculated, we need to determine the maxima in the function. As discussed above, there can be several local maxima of significance, and this algorithm uses the same smoothing approach to define them. The difference is that det_est determines ALL of the local maxima, rather than just stopping after the first one.

7. Once all the maxima are picked, the program moves the window up to one sample beyond the last threshold exceedance. If the estimation and threshold functions are complicated, this often is too small a move, the program makes sure that the window has moved at least half its length. We then start back at step 5 and loop through until we reach the end of the detection function.

PREFERENCES AND DEFAULTS

Several important parameters are given reasonable default values in the program. These values may be overridden by the user prior to any processing step simply by entering alternate values into the interactive windows that appear before the processing actually begins (Note that the "Option Windows" toggle must be set to "Show" for these windows to appear). These values may be defined by the user at startup through a file called Xphase.prefs located either in your home directory (defined by the environment variable "HOME") or in the working directory. Xphase will attempt to read files of this name when it begins, looking first in the working directory and then in your home directory. (If the file Xphase.prefs exists in both directories, the settings in the working directory will supercede those in the home directory). Variables are specified by including within the file a line with the variable name and its value (or 3 values, in the case of specifying colors). Only those variables which you wish to reset need be included. The items that Xphase tries to read are as follows (with default values):

arm	The order of auto-regressive model. This essentially is the number of points used to predict the value of the following point. The larger the order, the slower the program. My experience is that order 3 works pretty well, and this is the default value (arm =3).
arwid	Width, in samples, of the start of the window used in covariance modelling for estimation. The actual estimation takes place some distance inside the data window, because a few samples must be used in order to calculate the covariance matrix. The actual data window thus begins arwid samples after the defined beginning, and ends arwid samples before the defined end. The actual size of the data window is thus (arstop - arstrt) - 2*arwid. Default value is calculated from the size of the data window.
arstrt	Start of data window, in samples. Default = first sample.
arstop	End of data window, in samples. Default = last sample.
winsize	Size of window, in samples, around a phase detection for onset estimation. Default value is calculated from the size of the data window.
mvwind	Size of moving window, in samples, used for phase detection. Default value is calculated from the size of the data window.
pfact	P threshold factor (* noise_max) Default = 100.
sfact	S threshold factor (* detection_max) Default = .9
wfrac	Fraction of winsize used to determine the onset local estimation maxima. Default is 0.1.
nstrt	Start of noise window for phase detection, in samples. Default = 1st sample
nstop	End of noise window for phase detection, in samples. Default = Last sample
perr	Fallback P uncertainty (in seconds). Default is the time increment (delta).
serr	Fallback S uncertainty (in seconds). Default is the time increment (delta).
unthr	Uncertainty threshold, in fraction of estimation function maximum. Default = 0.9.
xmin	Sample minimum for plotting. Default = Current screen dimension.
xmax	Sample maximum for plotting. Default = Current screen dimension.
pickwin	Number of points used to determine polarity. Default = 2.
colors	Color scheme used, either black on white (0), or alternate (1). Default = 1. If alternate is chosen, the default is a yellow on blue scheme, but all of the colors in the alternate scheme can be modified by the user by specifying the next several preferences.
pcolor	Alternate color for P arrivals. Specify red, green, and blue values. Default is (255, 69, 0) which is an Orange-Red.
scolor	Alternate color for S arrivals. Specify red, green, and blue values. Default is (255, 255, 255) which is Pure White.
ocolor	Alternate color for all other arrivals. Specify red, green, and blue values. Default is (85, 107, 47) which is Dark Olive Green.
bgcolor	Alternate background color. Specify red, green, and blue values. Default is (0, 0, 53) which is dark blue.
fgcolor	Alternate foreground (waveform) color. Specify red, green, and blue values. Default is (255, 255, 0) which is Pure Yellow.
tmcolor	Alternate Time Mark color. Specify red, green, and blue values. Default is (255, 50, 200) which is a Pinkish color.
maxcolor	Alternate Maximum Value color. Specify red, green, and blue values. Default is (224, 255, 255) which is a Light Cyan color.
dotcolor	Alternate Time Dot color. Specify red, green, and blue values. Default is (204, 204, 255) which is a Light Grey color.
dwincolor	Alternate Data Window color. Specify red, green, and blue values. Default is (0, 255, 0) which is Pure Green.
nwincolor	Alternate Data Window color. Specify red, green, and blue values. Default is (255, 0, 0) which is Pure Red.
pthrcolor	Alternate P Threshold Line color. Specify red, green, and blue values. Default is (255, 69, 0) which is an Orange-Red.
sthrcolor	Alternate S Threshold Line color. Specify red, green, and blue values. Default is (255, 255, 255) which is Pure White.
nlvlcolor	Alternate Noise Level Line color. Specify red, green, and blue values. Default is (0, 255, 0) which is Pure Green.
view_dwin	View Data Window, either view (0), or don't view (1). Default = 0.
view_nwin	View Noise Window, either view (0), or don't view (1). Default = 0.
view_pick	View Picks, either view (0), or don't view (1). Default = 0.
view_picklab	View Pick Labels, either view (0), or don't view (1). Default = 1.
view_errs	View Pick Uncertainties, either view (0), or don't view (1). Default = 1.
view_tmark	View Time Marks, either view (0), or don't view (1). Default = 1.
view_tmarklab	View Time Mark Labels, either view (0), or don't view (1). Default = 1.
cfwin	Window length (number of samples) for computation of cross power spectrum when calculating Cichowicz characteristic functions. Default = 15.
max_view	The maximum number of datasets visible on the viewable part of the screen. If more that max_view data sets are viewed, they can be accessed by the scroll bar. Default = 10.
vpvs_min	The minimum value of Vp/Vs used by the S arrival time estimator. Default = 1.72.
vpvs_max	The maximum value of Vp/Vs used by the S arrival time estimator. Default = 1.90.
vpvs_pref	The preferred value of Vp/Vs used by the S arrival time estimator. This is the Vp/Vs used to estimate Tp for the first selected waveform. Default = 1.78.
mode	Control of operation for automatic detection. Possible choices for mode are: 0 Manual (interactive). This is the default for Xphase. 1 Raw P on vertical Raw S on biggest horziontal 2 Raw P on vertical Raw S both horizontals 3 Raw P on vertical Raw S both horizontals 11 Raw P on vertical P Coda Whitened S on biggest horizontal 12 Raw P on vertical P Coda Whitened S both horizontals 21 Whitened P on vertical P Coda Whitened S on biggest horizontal 22 Whitened P on vertical P Coda Whitened S both horizontals 33 3 Components, Whitened

THE BUTTONS ON THE TOP PANEL

When executing Xphase the first window to appear on the screen is the main window titled Xphase with the version number and the name of the directory from which Xphase was executed. The main window is divided up into two sections: a panel along the top containing various buttons, and a canvas onto which the time series are plotted.

The buttons on the panel and their functions are as follows:

READ

Brings up a window for inputing the names of pick files, ASCII time series, and FIR acausal filter files to be read in. The file name and directory are typed in the first two input rows. When the information is correct select "OK", or, if you want to abort the READ, select "CANCEL". Up to 50 data files can be read in. Note that once a time series file is read in it becomes the "Current" data set or model. After a time series file is read in, the extremal values of Y (amplitude) and X (sample number) are shown in the top panel in the Data Xmin:, Data Xmax:, Data Ymin:, and Data Ymax: entries.

WRITE

Writes the current file out as either an ASCII or SEGY file. Useful for saving Detection and Estimation files for later viewing or plotting.

DRAW

Draws selected time series and other marks on the screen. You can contol what is plotted by using the VIEW FILE and/or VIEW PREFS options.

VIEW FILE

Select which datasets to plot on the screen (or to delete when selecting DEL FILE later).

DEL FILE

Delete some set of data sets, with a choice of either only the current data set of all of the selected data sets (the dark ones in the "VIEW FILE" window).

OPTIONS

The View Preferences section allows you to select which variables to plot on the screen as vertical lines. If selected:

Data Window	appears as dashed green lines
Noise Window	appears as dashed read lines
Estimate Maxima	appear as dashed blue lines
Picks	appear as solid lines with phase dependent colors, depending on the background color scheme (these can be redefined by the user in the preference file). For the black on white scheme: P Pink S Royal Blue Other Olive Green For the Yellow on Blue scheme: P Orange Red S White Other Olive Green

Uncertanties appear as dotted lines in the color corresponding to the pick type.

The Pick Type section allows you to choose which phase to align when working in relative time mode. Default is the P phase.

SEL DAT

Select which datasets to use in processing (only time series files will appear, unless others are read in at the beginning). Note that you cannot select a dataset for processing unless it is viewed (i.e., plotted on the screen). Note that it is possible to select data using the Selection Boxes that appear to the left of each waveform (see discussion below).