PSN-L Email List Message

Subject: Filtering local noise
From: S-T Morrissey sean@...........
Date: Tue, 11 Apr 2000 22:20:45 -0500 (CDT)

Paul and Kareem have brought up a question that has not been
explicitly discussed, namely what to do about short period
local or "cultural" noise. Seismology would be a lot easier sport if
we didn't have to put up with ourselves and the noises of civilization.
This includes everything from the dogs romping through the house, 
the school bus or Mack trucks on the street, and the freeway a 
kilometer away. This is why professional seismic stations are hidden
in remote underground vaults rather than basements.

Such noise has generally been dealt with by using low-pass active
analog filters in the output of the seismometer.  If a digitizer is
used, its input circuitry probably includes a measure of "anti-aliasing"
filtering, where the filter period is set to twice the sample interval.
(A 10 hz waveform is minimally sampled by a 20 hz digitizer, which would
then require a 10 hz low pass anti-alias filter.)

For analog recording, the cosmetic appearance of the record becomes the
determining factor. You cannot see the small P-wave if the record 
is full of noise bursts from passing cars. So the idea is to try to
reduce the cultural noise to about the same level as longer period 
noises such as from tilting of the base of a horizontal or atmospheric 
pressure (vertical). Some amount of short period noise assures that the
instrument is working. Also, the 6-second microseisms should be present.
(this is true even for a long period sensor, not necessarily a broadband)
So it is a matter of customizing the filters to the site. For my monitor 
here of the leaf spring vertical operating at 90 seconds, I use a 
6-second passive notch filter to reduce the microseisms by about 100,
and about 6 poles of low pass filtering the cut the traffic noise.
I set the gain so I still see the local traffic noise at peaks of about a mm,
the microseisms when there is a storm, and barometric noise of storms of
10 minutes period or more, while not having a record that is too messy.

For utilizing low pass filters, simple one pole series resistor, shunt
capacitor combinations can be used prior to almost any amplifier stage.
More formal 2 and 4 pole active filters are generally employed. The
preferred response is the uniform phase delay of the Bessel configuration.
In a typical seismic system, the pre-amplifier at the seismometer usually
has two two-pole filters at a net cutoff of about 20 hz, which is more
than adequate for local short period instruments. The schematic of
this preamplifier is on the web site. The filter is easily changed
form 20 hz to 2 hz by using 10x the values of 4 resistors for use with
broader response long period sensors. (This filter can be built with regular
amplifiers like the LM308 if it doesn't have to be battery powered.)

Additional filtering is found in the output of the telemetry discriminator
(if used) and the digitizer input. Again, a phase linear response is 
used at about 20 hz for short period sensors like the L4-C, and 2 hz 
for a long period sensor. Most analog recorders provide several poles
of additional, often switch selectable, filtering. Here one can play
the tradeoffs of increasing the gain while lowering the frequency 
response to make the best of the cultural noise problem.

I have seen some of the schematics suggested for use with the Lehman
and SG designs, and they do include a measure of low pass filtering,
even if it is scattered about. A more focused filter arrangement would 
be preferable. On the other hand, I see a number of amplifiers with a
high value feedback resistor (for high gain) that is not bypassed with
a capacitor to limit the gain at the noise frequencies. Since all our
amplifiers operate at frequencies from DC to a few hz at most, we should
be careful that higher frequencies, which are still sub-audio, are not


Public Seismic Network Mailing List (PSN-L)

[ Top ] [ Back ] [ Home Page ]

Larry Cochrane <cochrane@..............>