Sean, Thanks for the input. I haven't tried anything very fancy to date. The scheme I'm using at the moment is reminiscent of the LTA-STA idea. I keep a running "LTA" with a shorter "STA" centered in the LTA window, and trigger on a specified number of samples that exceed a defined threshold value. I've played back the few event files that I have from my system along with some no-event files and it does an OK job of distinguishing between them. BUT several of my event files do not include the P-wave onset so I'm a long way from claiming any sort of victory. In the mean time, I'm willing to tolerate a high false alarm rate. If the false alarm rate is high enough, you wind up with a round the clock record (chuckle). My data collector does have some pre-detection memory, but it's only 5 minutes at the moment. Not long enough I've concluded. Disk space is cheap, missing the start of the show is disappointing. Larry ----- Original Message ----- From:To: Sent: Saturday, August 05, 2000 3:45 PM Subject: event detection > Larry, > > I have been trying to finness event detectors for years, and of course > have never found one that can sort out earth-related from man-made > events, only "something" from "not something". > > The most common is the "LTA-STA" process that can be implemented either > by analog filters or digital means. Basically, a short increase of short > period average energy is detected above a running long term average. > The STA is long enough to avoid spikes, like from radio telemetry. > Some digital detectors manage to use running FFTs before determining > the mean amplitudes. The detector parameters must be "tuned" to the > local noise situation, ideally not missing any events nor filling > the event file with junk. Of course, local and teleseismic detectors > need different parameters. I don't know where some code examples could > be found, but I would snoop around some major university seismo sites. > > And of course, all the methods involve a pre-event memory generally > long enough to catch the P-wave of a local event that triggers on the > S-wave. In the '70s, I used multi-channel analog tape recorders to > record the multiplexed FM telemetry carriers on a tape loop to provide > the delay; the loop output was written to the event tape (still telemetry > carriers) when the LTA-STA detector triggered. The tape loop had to be > replaced daily, and the splice in it often made glitches in the data. > Obviously digital delays are easier to do, as long as time tagging is > preserved in the delayed or pre-event data. > > The most competent detector I am familiar with is the one used in the > IRIS broadband stations. It was developed by Albuquerque, and is called > the MHH detector, or Murdock-Hutt-Halbert. It is quite elaborate, and > is set up or "built" by 11 parameters specified in the configuration file > for the station, with separate detectors for data of different sample > rates (derived by decimation and FIR filters from a single instrument). > Each MHH detector selects a particular IIR filter at its input. > It has proven quite robust in not missing events, and reports > detection quality info with each detection along with the period and > amplitude of the event and the background average counts. > > I believe that the code is available in C, and might be found > on the ASL web site (aslwww.cr.usgs.gov) . The IRIS stations run OS-9 > in 68020/30 multitasking processors, so I don't know if it is PC > compatible or if a stand-alone version is available. > > Regards, > Sean-Thomas __________________________________________________________ Public Seismic Network Mailing List (PSN-L)
Larry Cochrane <cochrane@..............>