PSN-L Email List Message

Subject: Re: A/D bits, LP filters
From: Karl Cunningham karlc@..........
Date: Sat, 13 Jan 2001 10:07:49 -0800

Hi Sean-Thomas --

You're right.  I didn't mean accuracy.  But I'm not sure I really meant
resolution either.  Perhaps "precision" (or distortion level) might have
been a better choice of words.

What I was trying to say (and please correct me if I'm wrong) is that it
isn't necessary for the system to faithfully record the waveform to within
15ppm (16 bits) for a full-scale signal.  For very small signals you want
to get as many bits of good data as possible.  But if the signal is large,
perhaps allowing an imprecision (in other words distortion, including
aliases) of something like 8 bits (0.4% of the peak-to-peak signal at the
time) may be good enough.  Of course, it depends completely on what the
data is to be used for.

I was suggesting that the compromise of allowing unintentionally aliased
out-of-band signals may be ok as long as they are below some fraction of
the overall signal level at that time, given the resultant benefit of
decreased circuit complexity.


At 03:00 1/13/2001 -0600, you wrote:
>Regarding the amplitude of seismic waves and digitizer range:
>I must disagree with Karl when he says 8 bits of accuracy, by which I
>assume he also means resolution, is sufficient. (Accuracy more properly
>describes the precision of the reference voltage that determines the value
>of a bit). Modern seismic studies use the entire content of the waveform 
>for source modeling, tomagraphy, path attenuation, etc, which are dependent
>on having an un-clipped and un-distorted signal for spectral analysis.
>A single clipped peak ruins the spectral analysis.
>Earthquake waveform amplitudes vary logarithmically, and the magnitude
>scales reflect this. A magnitude M 5 is 100 times more amplitude than a 
>M 3 (everything else being equal). An 8-bit digitizer has a range of
>256 counts, 12 bits is 4096, and 16 bits is 65536 counts (or half these
>values if a bit is used for the sign (+,-).  All our newer multichannel
>broadband stations use 24 bit digitizers which have a range of 16 777 216 
>counts for a 40 volt p-p signal from sensors with an output of about
>2000 volts/meter/second. While we infrequently (like monthly) see amplitudes
>that actually exceed the 16 bit range, large quakes greater than M 7+ do
>use much of the dynamic range. I have data from our stations exceeding 
>2 000 000 counts from South Pacific and Eurasian events. Of course a great 
>quake in the near field of the station would clip the 24 bit system, so most
>IRIS stations also have a 3 component FBA (Force Balance Accelerometer).
>The background noise (natural and instrumental) of our stations runs
>about 200 counts at about -160 db PSD. The least count is 2.384 microvolts.
>But for my digitizer here at home, I am still using the 12 bit multimeter
>at 1 sample/second, with a full scale of 200 mv, and a least count of 
>0.1 mv connected to the STM seis with an output of 1640 v/m/sec. 6-second
>microseisms run 3 to 15 mv, and it clips on most events greater than M 7.0.
>And a comment about filters. Doug is quite correct about the natural
>attenuation of frequencies above a few tens of hz. For all out short
>period stations (1 hz seismometers) we have a 4-pole Bessel filter in
>the pre-amp (plus the usual single 30 to 50 hz low pass poles (to reduce 
>amplifying 60 hz noise) that are created by by-passing the high value gain 
>(feedback) Rs with a capacitor, as shown in the pre-amp schematic on my
>After the telemetry (VCO, FM transmitter, FM Receiver, and Discriminator PLL)
>another 17hz 4-pole filter provides anti-aliasing for the 20 sps digitizer.
>At my CCM station, the 1 hz high gain system for the CTBT (Comprehensive 
>Test Ban Treaty) uses a 24 hz low pass filter for its 40 sample/sec channels.
>Although the rules for anti-aliasing are exact, the content of the seismic
>signal allows some convenient variation; as Doug points out, the actual
>noise is only a small percentage of the full scale signal.
>Public Seismic Network Mailing List (PSN-L)
>To leave this list email PSN-L-REQUEST@.............. with 
>the body of the message (first line only): unsubscribe
>See for more information.

Public Seismic Network Mailing List (PSN-L)

[ Top ] [ Back ] [ Home Page ]

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