Read me for event gathers directory

3 February 1997

C. H. Jones

====================================
This directory contains the event gathers extracted from the continuous
data from the 1994-5 Colorado Plateau-Great Basin (CPGB) PASSCAL experiment
as well as event gathers from the U.S. National Seismic Network for
stations in the SW USA for the same events and gathers for many
of the same events from the permanent digital network of the University
of Nevada, Reno.  These latter two datasets were collected in order to
conduct a more thorough research program; we include these with our own
data as an aid to other researchers.  There are difficulties combining
all these data, as are noted below.

Overall, the data themselves are pretty well fixed and unchanging.  we
have been adjusting the databases to more accurately reflect the
instrument responses and gains.  For the CPGB data, barring significant
deviation from factory responses, everything should be complete and
fully accurate.  For the other datasets, noticeable limitations apply
as are discussed below.  In general, the databases for the "correct"
timing data for both CPGB and UNR datasets should be in better shape
than for the NoTime or _bad databases.

CPGB data:
  These are in directories orid00xxx, where xxx is the event number keyed
  from the .origin table (links to these directories are found in the
  ../Events_by_date directory using date/time names).  Files are in
  SAC format and should be nominal nm/s within the flat part of the
  instrument response.  Filenames include the year, Julian day, hour,
  minute, and second of the start of the record.

  Easiest access to this data is probably through the CSS3.0/JSPC
  database within this directory.  There are four databases that access
  all or part of these data:
    EventDB:  All data with known timing from CPGB.
    EventDB_NoTime: CPGB data with timing problems (usually < a few seconds)
    CPGB_EG_for_seedDB: Same as EventDB but including the more formal
      stage response files for SEED as well as network IDs, etc.  This
      was used to generate the SEED volumes submitted to the IRIS DMC.
    AllEventsDB: EventDB combined with US NSN data (below).
  Note that the EventDB and EventDB_NoTime require the presence of 
  the "current" db in the directory above this one.  If the current tables
  are not there, adjust the EventDB file accordingly (second line is a search
  list for missing tables).

NSN data:
  These are in directories sorted by response type (as defined by the
  NSN group at the time--some streams seem misplaced):
    orid00xxx.NSN.longper:  Long period data (usually about 1 sps)
    orid00xxx.NSN.broadband: Broadband data (usually about 40 sps)
    orid00xxx.NSN.shortper: Shortperiod data (between 40 and >100 sps)
  Response files from the long period and broadband sensors are referenced
  through the relavent CSS db in the NSN_responses directory, which
  also contains some scripts used in converting the response files that
  were returned with the data.  Short period data usually lacked response
  files.  Data files are in SAC format and are often close to nm/s
  but off some; the calib field in the CSS db .wfdisc tables corrects for
  this, thus the best way to access this data is probably through the
  CSS database.  Each stream has a database:
    NSN_Event_LP for the long period data
    NSN_Event_BB for the broadband data
    NSN_Event_SP for the short period data
  The NSNtodb script was used to generate the databases.
  Note that these use the ../current database for missing tables; this
  isn't strictly accurate and is mainly for use of the .origin table;
  the .network and .affiliation tables do not apply to these.  Also,
  the .instrument and .sensor relations do not apply to the _SP db.
 
  The NSN data are all combined with the CPGB data in the AllEventsDB
  database.  Limitations of the individual databases apply as well to
  this database.

  Problems:
    Overlap of event gathers caused some problems, so a small amount of
  data in the orid directories might not be present in the db.
    Responses provided are usually the factory responses and not
  anything measured.  Recording system responses are not included
  beyond the gain factors, though one assumes an anti-aliasing filter
  is present.
    Any blunders in the USNSN data retrival system for the time we gathered
  data (c. 1995) are propogated here.

UNR data:
  The UNR data is also saved as SAC files within an orid00xxx style
  directory tree.  In this case the files are in orid00xxx.UNR and
  possibly additional directories orid00xxxY.UNR, where Y runs up from A
  through B, C, etc.  This was caused by a desire to avoid altering the
  original data structure from UNR.  There are two databases:
    UNR_EventsDB:  Data with timing errors < 0.4s (usually a lot less)
    UNR_Eventsdb_bad:  Data with errors > 0.4 s
  Responses derived for these instruments are referenced in the sensor
  and instrument relations of the UNR_EventsDB.  Note that differences
  between different components have been observed.

  The UNR data was not included in the AllEventsDB because of some
  overlap with NSN stations and the difficulties with timing and gains.

  Problems:
    The UNR digital data is recorded in halfhour intervals through a
  custom transmission system developed at UNR.  It was found after the
  CPGB deployment was over that this system occasionally had dropouts
  of packets that would cause the timing to slowly vary from true
  timing.  Attempts at UNR to recover these errors met with limited
  success; thus the separation of data into those with timing errors
  under 0.4 s and those with errors greater than that.  Thus any
  techniques requiring absolute accuracy better than 0.4 s really should not
  use this data at all.
    Gains also appear to be a problem.  This might be due to the use of
  gainranging bits in the UNR digital transmission scheme (12 bits of
  data and 4 bits of gain ranging) or might reflect some misunderstandings
  of the instrument gains.  Regardless, gains should be taken with a 
  grain of salt.  NSN stations at BMN and MNA are colocated with UNR
  stations and thus provide some measure of gain accuracy for those
  events recorded in common.  Gains are as reported from UNR and are
  30-100 times away from nm/s in the raw SAC files.