Introduction
velin was written by Bob Herrmann from St. Louis University in 1986. In it's initial incarnation it simply built a velocity trace from an input velocity function set. In 1991 several options were added to allow velocity type and time/depth conversion capability. In addition, several changes have been made to the routine over the years to handle a variety of input velocity function set formats and various flavors of data indexing. These changes were made by a variety of programmers with a variety of programming styles resulting in code that has become very difficult to maintain. At this point in time several new options have been requested, and since several old bugs need reparation, it has been decided to re-write the entire routine to both incorporate the changes and make the code much easier to maintain.
Bugs Fixed
- all memory allocation is now dynamic. No longer will the routine fail for small control datasets coupled with large velocity function sets.- the velocity type conversion routine will now handle the missing average velocity, depth output.
- the internal record number array is now dynamic and as such will no longer fail with more than 4096 record numbers.
- DISCO handvel cards may now be right or left justified.
New Features
The following upgrades have made velin compatible with MBS, ILIS, ProMax, DISCO and PPRED.- the first sample of the output USP velocity dataset is referenced to zero time [t = 0.0 seconds]. This is a change from the first sample being reference to the sample interval.
- the internal interval velocity reference has been moved to the base of the interval. This is a change from the top of the interval.
- a command line option -ilis has been added which results in a fifth order polynomial approximation to the velocity function set being generated. This is the same approximation routine used in the Ilis System.
The following upgrades enable the routine to handle very large datasets and grant the user greater facility in velocity input and conversion.
- if your velocity function set is a USP dataset all dynamic memory allocation is shut off. This enables velin to handle very large datasets [3D] which would previously not be possible.
- when performing sample unit conversions [time/depth] velin now calculates both the maximum time and maximum depth involved an reports the complete function conversion in the printout file [if verbose -V output is flagged].
- all trace header word reference is now accomplished using the trace header mnemonic [see the program scan man page]. This includes the -hw command line entry.
- a sorting routine has been added which sorts the input velocity function set to the same indexed order as the control dataset. This allows one to enter one's velocity control in any arbitrary order which is useful when supplying velocities in pick file -xsd format. One may then pick one's velocity spectra from best spectra to worst instead of left to right.
- a slope adjustment factor has been added for use with USP velocity function set input. If velocity type conversion is requested velin will decimate the input function based on a slope change algorithm. If the input velocity is rms or average, a velocity sample will be gleaned from the input dataset anytime the slope of the velocity curve changes by more than the amount specified by the input slope adjustment factor -saf. The value is supplied as a percent. For instance a -saf value of 0.1 would cause a velocity sample to be gleaned any time the slope changed by greater than 0.1 percent. This parameter affects the smoothness of the resulting velocity field.
The following upgrades allow for much better quality control of the velocity data used.
- an XSD pick file output option has been added which if activated will result in the creation of an XSD pick file containing the output velocity function set. This is useful when quality control of velocity analysis is undertaken. It is also useful [in combination with piktdfn] when an input velocity function set format change is desired.
- an effort has been made to clean up the error messages produced by velin. All error messages are output to the printout file. In addition any terminal condition is also reported to stderr. The error messages now contain references to record number and sample where an error occurred. This should aid in debugging problems with one's input velocity function set.
Features Removed
- the command line option -cline has been removed. This option made possible the inclusion of a tdfn input velocity function set on the command line immediately following the velin statement. If you have used this option in the past you will now have to put your tdfn format velocity function set into a disk file and supply the filename using the -v option. Also include the flag -tdfn on the command line.
Program Usage
Describing the subsurface in terms of velocity is a complicated endeavour fraught with pitfalls. Representing your subsequent velocity distribution using digital sampling adds a set of variables which if not handled correctly will cause significant errors in the resulting velocity dataset. Some of these errors will be obvious even to the uninitiated while others may be quite subtle but cause large errors upon application.Once you have determined the velocity function set to convert to a USP dataset the following should enable you to achieve that goal with confidence.
Indexing
The relative indexing between the control dataset and the input velocity function set is critical to the correct spatial interpolation of the input velocity field. It is also the area of greatest confusion to the average user. Confusion arises when the user's concept of how the line is indexed does not coincide with the information stored in the trace headers. This usually happens when the user confuses the sequential record number with the record index he/she is using to reference the velocity function set. For example, consider the following velocity function set [illustrated in flatfile format]:
- 4, 1500, 1000
There are two velocity functions contained here. The first is referenced to an index location of 1000 and the second to 1004. velin has the capability of relating the input velocity function indexing to any trace header word. The default header reference in velin is RecNum while other standard references include -shot [SrcLoc], -receiver [RecInd], -cdp [DphInd] and -pri [PrRcNm]. Of course any valid trace header mnemonic may be reference through the use of the command line option -hw.
Whatever the index chosen, it is entirely independent of the sequential record number, which is not stored in any trace header location, but is merely the sequential location of that record in the dataset on disk. The first record in a dataset is always sequential record one while the last record of an N record dataset is always sequential record N. The control dataset for the above velocity function set contains sequential records 1 through 5. Since this example is set up to use default indexing the trace header word that the velocity indexing is referenced to is RecNum. A quick scan reveals that sequential record one has a RecNum entry of 1000 (Table 1).
Regular Indexing
If the indexing in your control dataset is regular, that is if you can take the first index and increment by 1 from record to record, then you may make use of the command line option -R. If present velin reads only the first trace from the control dataset. The index contained therein combined with the line header entry for the number of records in the dataset, NumRec, fully defines the indexing for the output velocity dataset. The advantage here is that the program does not have to read the entire control dataset to build it's internal record index array and hence runs much faster.
Single Trace Records
In USP velocity datasets are constructed of single trace records. The number of records is equal to the number of records in the control dataset if more than a single velocity function is used. In the MBS world, velocity datasets are often [if not always] a single record of many traces. If you want to use such a velocity dataset for input to velin simply run
- utop -N YourData -O YourData -L1 -R ntrc
where ntrc is the number of traces per record of your MBS velocity dataset.
Velocity Mapping
Each trace of a USP velocity dataset is made up of a fixed number of samples corresponding to the number of samples per trace in the control dataset. The time [or depth] value at each sample is a function of the sample interval used. The method used to map the input velocity function to the output time/depth series is a function of the velocity type of the input function set.
Interval Velocity
- a given interval velocity will be assigned to all samples with a time/depth greater than that of the previous entry in the function and less than or equal to that of the current entry.
RMS or Average Velocity
- linear interpolation is done within the function. The first value is used above the first element of the function [if non-zero]. Where the last function time/depth is less than that of the control dataset an appropriate entry is calculated by extending the last defined interval velocity.Should the command line flag -ilis be present then the mapping is done using a fifth order polynomial approximation to the input function regardless of velocity type.
Velocity Function Completion
If not supplied with a velocity corresponding to the maximum time [or depth] of your control dataset velin will attempt to generate one. Your input function, if average or rms, is converted to interval velocity. The bottommost velocity is used to generate an rms or average velocity entry at maximum time [or depth]. When the input velocity function type is rms a problem arises. The solution for interval velocity involves a square root containing the expression [V(i) - V(i-1)]. In the presence of an rms velocity inversion the above expression becomes negative and the solution to the square root becomes imaginary. velin will fail at this point and report the occurrence to the printout file. If you have valid rms velocity inversions you may prevent velin from aborting by simply including an rms velocity value for the maximum time [or depth] of your control dataset in the input velocity function affected. Each element of the input velocity function set is handled independently so that you need only do this for the functions where the reversal occurs.
Velocity Type Conversion
In most cases velin is used to build a USP velocity dataset for normal moveout correction using rms velocity. In some instances however, the user has acquired velocity control in either interval or average velocity and would like to generate an rms velocity dataset. Alternately, one may wish to generate interval velocity from a pre-existing rms velocity dataset. These velocity type conversions are done inside the program using Dix equation and requested through the command line entries for input velocity -vin and output velocity -vout.Should the input velocity function set be a USP dataset in rms or average velocity velin will attempt to decimate the number of entries in the input function. The algorithm used examines the slope on the time/depth, velocity curve and gleans a sample any time the slope changes by more than the percentage specified by the command line entry for slope adjustment factor -saf. The preset defaults provide reasonable results, however, smoother interpolations, especially when the output velocity type is interval, may be had through experimentation with this value. Too large a value will result in too sparse an output function. Too small a value at worst will cause the output to contain the gleaned function to have the same number of samples as the input. Be sure to flag the command line option -V and examine the printout file to monitor the results.
Sample Unit Type Conversion
velin will do rudimentary time/depth or depth/time conversions. The method used is again a function of the input velocity type.
Interval Velocity
- a running sum calculation is used
Average Velocity
- an instantaneous calculation is used at each function location with linear interpolation between.RMS Velocity
-The function is converted to interval velocity followed by the running sum method. The limitations with respect to rms velocity inversions apply here also.When attempting a sample units type conversion it is important to get the output sample interval correct. If the interval used in the input control dataset is not desired then use the command line entry -osi to enter the appropriate override.
Quality Control
Without adequate QC there is a very good chance that the output from velin will contain some unwanted surprises which will cost you dearly before your processing flow is completed. The following minimal quality control procedure will save you hours of grief and is recommended any time a USP velocity dataset is generated.- scan your input control dataset to verify that trace header word used to index your input velocity function set.
- use the command line option -V to generate a complete listing of the input function read and the output function generated. Examine this listing and verify that the correct action was taken with regard to your choice of command line parameters. This listing will appear in the output printfile.
- plot the resulting velocity dataset using xsd or xmplot and verify that velocity interpolation was carried out correctly. Any glitches in the output velocity field will be very obvious on this display.
- scan your output velocity dataset to verify that the indexing used was correct. You may also use the command line option -np of program scan to output the actual sample values of the velocity dataset to verify that the correct velocity field has been formed. To examine a particular trace of your output velocity dataset you could issue the command
- sis_xy -N YourVelData -rs -re -ns -ne | xgraph
to get a graphical display of velocity versus time/depth.
- always read through the velin printout file as not all the warning are posted to stderr. Most of the policemen inside velin will provide you with the index location and function element where something bad happened.
Program Help
There are two sources of online documentation for this routine. Firstly you may issue the command uman velin which will print the velin UNIX man page to the screen. Secondly, you may obtain an abbreviated help by issuing the command velin -h or velin `-?' which will cause the program to execute it's internal help subroutine listing the command line parameters and their defaults to the screen. The man page is also available through the xuspman functionality browser and of course IKP.
Return to the USP Home Page.Frame Doc to html conversion by Tony Garossino