The following questions and answers may help you in processing response calibrations.
The software was developed originally on Sun boxes running SunOS 4.x and has recently been ported to Solaris 2.x. Many of the components of the package will work under Linux, but not all.
The software is available at ftp://ftp.crustal.ucsb.edu/scec/sun. and will have a name in the form response-<version>.<OS>.tar.gz. Once you have the tar file, uncompress it and read the README file in the shells directory.
The naming convention is used to get some of the required info into the final printout that you will produce. The information would have to be entered in one of the database tables otherwise. Also, by requiring this format, the data is somewhat organized just by the names. You can look at any file and know a lot about that cal pulse.
It means that you need to provide another table to the response software. Usually this is one that contains manufacturers data specific to a type of sensor. You need to enter the filename manually into the datadir/caldir/sac/tablist file. This is one of the weakest portions of the whole process. I am trying to work in a nice GUI to make this easier, but I haven't worked out all the pieces yet.
When viewing the raw seismograms, you want to make sure that they all look like they are roughly the right shape and that each file contains a valid cal pulse. Then just hit the quit button.
When viewing the clipped seismograms, all the pulses should line up. The pulses should all start about one quarter of a second after the beginning of the record. The pulse should be complete. Sometimes the clipping routine chops a bit much off. This usually doesnt cause any problems, but you will see that when the clipped data is compared to the calculated model. Once you have looked over everything, hit the quit button.
When comparing the data and model. You will want to observe how closely the two signals overlay each other. A good match will look almost like one signal. Hit the next button till you have viewed all the signals. Xqr should automatically quit once you hit next on the last screenful of plots.
When comparing adjacent pulses, you are looking at the positive and negative pulses for the same component. If there is a huge difference in amplitude or shape it can indicate a number of things. If the component is a horizontal sensor, it may just indicate that the sensor was not leveled adequately. If the sensor is a vertical, it most likely means that the springs are starting to sag. Hit the next button till you have viewed all the signals. Xqr should automatically quit once you hit next on the last screenful of plots.
1 pulse is adequate in most cases. The key is to make sure that the sensor does not hit the stops when you apply the current to offset the mass. This will give you incorrect amplitude information. And while you don't want to put too much current into the coil, you also don't want to put too little current into the coil. If you do, the resulting signal may be too small and the signal to noise ratio may make it harder for the software to get a fit. The polarity of the pulse wavelet does not matter for the llnlid software. I prefer the pulse with the large positive going lobe, but that is just personal preference.
It is often useful to record a postive and negative pulse for each sensor. This allows you to compare the two and see if you have any asymmetry in the pulses. This could be an indication of a non-linear response characteristic possibly caused by tilt or fatigued springs. In fact this characteristic can be useful for calibrating the bubble levels on the cases of horizontal or 3d sensors. A storage oscilliscope can be used to overlay the pulses. This allows the technician to compare the waveforms and try to get them to match as closely as possible by changing the tilt. When both positive and negative match, that should be the true level spot. Once this spot is found, higher amplitudes should be used to verify that positive and negative pulses hit the stops at about the same current level. If they do not, the springs are most likely behaving differently on the postive and negative swings.
I believe that the llnlid program only uses the first 1000 samples of data. This poses problems in two different scenarios. The first is records in which the pulse is far into the record. The PBIC procedures attempt to crop the records closely around the signal pulse to avoid this issue. The second situation is when attmpting calibrations of long period instruments using high sample rates. The only workaround is to record the pulses at a low enough sample rate that allows enough of the pulse to be recorded in the first 1000 samples to get a good fit.