PSN-L Email List Message

Subject: Re: Digital / Analogue Inverse Filters
From: ChrisAtUpw@.......
Date: Sat, 31 Dec 2005 21:54:13 EST


In a message dated 31/12/2005, Bobhelenmcclure@....... writes:

>     If you should still want to try to extend the low frequency response of 
> a 4.5Hz geophone, I have implemented a digital inverse filter which can 
> operate on PSN Type 4 files. The application is "WQFilter.exe". To download it, 
> go to 
>    http://www.jclahr.com/science/psn/mcclure/wdq_utilities/index.html
> and download the file "seismic_dataq.zip". Extract "WQFilter.exe" and 
> "WQFilterHlp.txt" from the zip file.
>   
>     My personal opinion is that trying to use a 4.5Hz geophone is a pretty 
> futile exercise for most amateur seismologists, unless you have an interest in 
> local cultural noise or are sitting on top of an active volcano. I use 
> moderate period homebuilt sensors for my station. They range in natural period 
> from ~5 seconds to ~15 seconds, and for them, the use of "WQFilter" yields event 
> waveforms closely resembling those recorded by professional long period 
> instruments. I use oversampling and DC amplifiers as well for data acquisition.

Hi Bob,

       Thanks for the comments. The negative side of trying to do x10 digital 
frequency extension for 4.5 Hz geophones is two fold. The low frequency 
signals are inherently small and grainy and most 16 bit ADCs have 2 to 3 bits of 
noise on them. This noise can be averaged out by oversampling, as you mention, 
but you do need to do it with a processor on the ADC board. You need four times 
the number of samples for each bit of ADC noise that you want to average out 
-> 64 samples per reading for 3 bits of noise. Normal communication software 
does not let you take enough samples quickly enough, using just the main data 
logging computer. 24 bit ADCs have built in averaging, but you could still see 
1/f noise.

       Roberts' original circuit used a two stage low pass filter. This 
allowed the very low frequency 1/f circuit noise to be amplified along with the 
signal. Changes to the circuit design have removed most of the VLF noise 
sensitivity. The noise on the compensated signal is certainly higher than I would get 
with a new $1000 1 Hz geophone, but it is still significantly below the 
ambient + cultural noise that I experience. This analogue technique is viable and it 
can be made to work reasonably well. 

       I agree with your rather more pessimistic opinion of the digital 
technique for 4.5 Hz geophones. If you want a period extension of less than x4, the 
digital method should work OK.

       I wish everyone a less traumatic year for 2006!

       Chris Chapman
In a message dated 31/12/2005, Bobh=
elenmcclure@....... writes:

    If you shoul= d still want to try to extend the low frequency response of a 4.5Hz geophone= , I have implemented a digital inverse filter which can operate on PSN Type=20= 4 files. The application is "WQFilter.exe". To download it, go to
   http://www.jclahr.com/science/psn/mcclure/wdq_utilities/inde= x.html
and download the file "seismic_dataq.zip". Extract "WQFilter.exe" and "WQFil= terHlp.txt" from the zip file.
 
    My personal opinion is that trying to use a 4.5Hz geophon= e is a pretty futile exercise for most amateur seismologists, unless you hav= e an interest in local cultural noise or are sitting on top of an active vol= cano. I use moderate period homebuilt sensors for my station. They range in=20= natural period from ~5 seconds to ~15 seconds, and for them, the use of "WQF= ilter" yields event waveforms closely resembling those recorded by professio= nal long period instruments. I use oversampling and DC amplifiers as well fo= r data acquisition.


Hi Bob,

       Thanks for the comments. The negative s= ide of trying to do x10 digital frequency extension for 4.5 Hz geophones is=20= two fold. The low frequency signals are inherently small and grainy and most= 16 bit ADCs have 2 to 3 bits of noise on them. This noise can be averaged o= ut by oversampling, as you mention, but you do need to do it with a proce= ssor on the ADC board. You need four times the number of samples for eac= h bit of ADC noise that you want to average out -> 64 samples per reading= for 3 bits of noise. Normal communication software does not let you take en= ough samples quickly enough, using just the main data logging computer. 24 b= it ADCs have built in averaging, but you could still see 1/f noise.

       Roberts' original circuit used a two st= age low pass filter. This allowed the very low frequency 1/f circuit noise t= o be amplified along with the signal. Changes to the circuit design have rem= oved most of the VLF noise sensitivity. The noise on the compensated signal=20= is certainly higher than I would get with a new $1000 1 Hz geophone, but it=20= is still significantly below the ambient + cultural noise that I expe= rience. This analogue technique is viable and it can be made to work reasona= bly well.

       I agree with your rather more pessimist= ic opinion of the digital technique for 4.5 Hz geophones. If you want a peri= od extension of less than x4, the digital method should work OK.

       I wish everyone a less traumatic year f= or 2006!

       Chris Chapman

[ Top ] [ Back ] [ Home Page ]