not too butt in but I believe it to be important
that you include either a digital or analog low
pass filter of like n8 or n6 that is maybe ((0.1 to 0.27)* Samp/sec)in Hz of your sample rate
before the analog signal hits the a/d converter or you will get artifacts that
fold back into the same frequency range in which you are looking as a
matter of frequency translation which is like a superhetrodyn receiver.
I know this both from military technical training as well as personal experiences.

unrelated note of interest:
Also I have noted that in order to like integrate the data in Winquake
you must first bandpass filter equalize the data so it stays at zero
and not a diagonal line across the screen. after integrating you will
still see noise if only there is noise but if like there is human activity
in the neighborhood the signals will appear as low frequency
artifacts. When you equalize the data you can have differing N
factors for high and low pass sides of your filter.
whatever takes the noise and creates a straight line data
set at zero baseline.
I have never seen anyone recommend this before.
also, I have seen the integrate function totally eliminate a valid signal and appear
only as noise.



----- Original Message ----- 
From: "Randall Peters" 
To: 
Sent: Thursday, February 19, 2009 5:42 PM
Subject: coil/magnet sensors


Larry,
      The primary reason for poor signal to noise ratio at long periods with a coil/magnet system involves the physics employed. 
Such a sensor functions on the basis of Faraday's law, which says that a time changing magnetic flux gives rise to a voltage.  The 
key word in this process is 'changing'; the output depends on the flux rate.  If the boom of the seismometer is swinging 
sinusoidally at a frequency f, then the amplitude of the output voltage of the sensor is proportional to f-which is the very nature 
of the time derivative of the sine function, involving the chain rule.  Because the signal is proportional to f, the signal level 
decreases by the same amount as f decreases.  Regardless of the nature of the noise, this means there is a significant reduction in 
the signal to noise ratio for accelerations having a frequency below the natural frequency of the instrument.  When operating 
without the high-pass filter with your S-G instrument, your output is not rate sensitive; therefore it isn't afflicted with this low 
frequency loss.  The 'wandering' you experience is actually representative of the various changes occurring.  Although some of those 
changes are oftentimes the undesirables of instrument thermal coefficient or whatever, there are also significant variations 
associated with the Earth itself.  When you high pass the output, any earth motions having frequencies lower than the cutoff  are 
suppressed.  I operate the VolksMeter with a high pass and also low pass filter of recursive type in WinSDR.  The primary reason is 
so that the helicord record is well behaved for purpose of seeing earthquakes.  But for viewing tides, or magnetoelastic phenomena, 
or earth hum, or .... It is necessary to look at the unfiltered output, which is what I save.
      The obsession with 'velocity' sensing is almost universal.  Even force balance instruments of the highest dollar type use a 
network that causes the seismograph to behave just like a magnet/coil system with a low corner frequency, usually about 30 s.  Thus 
the greatest advantage of the capacitive sensor is actually lost, for signals having frequencies lower than the design corner.  For 
signal variations having a frequency below the corner, the output from such an instrument is not 'velocity'; it is instead the 
derivative of acceleration, called the 'jerk'.  Only for drive frequencies above the corner can one use the term 'velocity' 
appropriately.
    Randall

__________________________________________________________

Public Seismic Network Mailing List (PSN-L)