In a message dated 22/12/2009, brett3nt@............. writes:

>1  I know this is an over simplification, but does  the  P arrives at an 
>angle (bottom to top) and this influences  the spring?   
What you suggest I believe is true, that P waves  may have some vertical 
component, though there might possibly be another  contribution related to 
something that Randall mentioned.  We  frequently see P waves on the 
verticals, sometimes quite strongly, which  might in part relate to the 
fact 
that when something gets squeezed in one  direction, it will expand 
somewhat 
to the sides.  
Hi Brett,
 
    It may have even more to deal with the curved paths  that the rays 
travel inside the spherical earth!

This  ratio of the sideways expansion to the original amount 
of compression is  called Poisson's Ratio, which for many materials is 
about 
1/3.  As a  horizontal P wave expands and compresses the earth, it may 
possibly be  causing additional vertical motion due to that effect, which 
you should  see on the vertical (though I don't know what Poisson's ratio 
might be for  dirt).  
    It is rock that you need to check.

>The  obvious advantage of a vertical is they see many earthquakes, the  
>disadvantage they are limited and also see a lot of other things which  
are 
>not earthquakes.  All of my verticals have been 1 to 2.5  seconds. 
> A typical vertical sensor recording will have a trace  of  20 
>mins vs. a 3 hours trace on the Lehman.
    This must be an observational error. The noise on  an uncased, 
uncompensated vertical may be over 20x that on a horizontal. The  noise on verticals 
is only less at longer periods since atmospheric noise  is excluded in 
contemporary seismometers by the hermetically sealed case.  

From  what I have seen, much of the surface-wave energy of large teleseisms 
is  in the region of 18 to 20 seconds, and sometimes longer.  20 seconds  
will see a lot, while 40 or 50 seconds can often see somewhat  more.
    You definitely need the sensor response flat  out to 20 seconds. A 
Lehman with this range will see 40 second waves at 1/4  their true amplitude. 
The roll-off in a vertical may be more rapid, depending on  the type / 
feedback.

One  problem is that there is so much microseism noise in the six-second 
region  and also around 12 seconds.  I generally use a 0.08Hz (12.5 second) 
4  or 6 pole low pass filter to cut out most of that noise, while allowing  
through much of the big-quake frequencies.  An instrument that can  only go 
down to 6 seconds will unfortunately be quite good at displaying  the noise 
while missing most teleseism surface waves.
    Yup!   I only see 10 second noise on  a few days each month in the UK. 
The six second noise batters us all the  time!

>  One's location has an influence on which sensor type  works for them

The only problem is that the movement of a seismometer  in response to an 
error force, such as from small temperature changes,  etc. increases as 
1/Frequency^2    
    Even more critical are air density changes  effecting a vertical sensor 
and also wind noise. Temperatures should change  quite slowly inside an 
insulating case, assuming that the sun does not  shine on it!

It is  way harder to make a stable instrument for 50 
seconds than for 10 or 20 or  6.  Long period verticals almost certainly 
need to use a feedback  design if they are going to be sensitive enough to 
see distant quakes  while at the same time, insensitive enough to 
temperature and other  variations to not wander off to maximum output.


Uh, Uh! Now define what you are calling long  period? 
 
    The Roberts' period compensating amplifier circuit  is probably the 
easiest technique for amateurs to use. It has constant gain from  the LP filter 
down to ~ the resonant frequency rf of the vertical sensor,  at say 1/2 Hz. 
See _http://jclahr.com/science/psn/roberts/index.html_ 
(http://jclahr.com/science/psn/roberts/index.html)    It was first used on geophones.  
Below this to about rf / 10, (= 1/20 Hz) the gain  increases as 1 / f^2, so 
compensating the f^2 falling output to give a flat  characteristic. So you 
can extend a 2 second period vertical sensor out to about  20 seconds quite 
easily. Extending the period much beyond x10 quickly runs into  noise 
problems with a coil + magnet velocity detector. You need two of  these stages 
with maximum gains of x10 (total x100 at 20  seconds) linked by a 2 pole high 
pass filter at ~ 1/30 Hz. Lennartz use a  system like this. So do I and it 
works fine.
    Direct digital period compensation probably works  best with 24 bit 
ADCs. The French use it on their schools system. 16 bit ADCs  may only show a 
few counts (1/100 the rest of the signal) at x10 period, unless  the gain is 
quite high and it may be partly masked by electronic  noise.
 
    Regards,
 
    Chris Chapman

 





In a message dated 22/12/2009, brett3nt@............. writes:
>1  I know this is an over simplification, but does=
=20
  the  P arrives at an 
>angle (bottom to top) and this influen=
ces=20
  the spring?   
What you suggest I believe is true, that P=
 waves=20
  may have some vertical 
component, though there might possibly be ano=
ther=20
  contribution related to 
something that Randall mentioned.  We=
=20
  frequently see P waves on the 
verticals, sometimes quite strongly,=
 which=20
  might in part relate to the fact 
that when something gets squeezed=
 in one=20
  direction, it will expand somewhat 
to the sides.  
Hi Brett,
 
    It may have even more to deal with the curved=
 paths=20
that the rays travel inside the spherical earth!
This=20
  ratio of the sideways expansion to the original amount 
of compressio=
n is=20
  called Poisson's Ratio, which for many materials is about 
1/3. =
 As a=20
  horizontal P wave expands and compresses the earth, it may 
possibly=
 be=20
  causing additional vertical motion due to that effect, which 
you sho=
uld=20
  see on the vertical (though I don't know what Poisson's ratio 
might=
 be for=20
  dirt).  
    It is rock that you need to check.<=
/DIV>
>The=20
  obvious advantage of a vertical is they see many earthquakes, the=20
  
>disadvantage they are limited and also see a lot of other things=
 which=20
  are 
>not earthquakes.  All of my verticals have been 1 to 2.=
5=20
  seconds. 
> A typical vertical sensor recording will have a=
 trace=20
  of  20 
>mins vs. a 3 hours trace on the Lehman.
    This must be an observational error. The nois=
e on=20
an uncased, uncompensated vertical may be over 20x that on a horizontal.=
 The=20
noise on verticals is only less at longer periods since atmospheric=
 noise=20
is excluded in contemporary seismometers by the hermetically sealed case.=
=20
   
From=20
  what I have seen, much of the surface-wave energy of large teleseisms is=20
  in the region of 18 to 20 seconds, and sometimes longer.  20 second=
s=20
  
will see a lot, while 40 or 50 seconds can often see somewhat=20
more.
    You definitely need the sensor response =
flat=20
out to 20 seconds. A Lehman with this range will see 40 second waves at 1/=
4=20
their true amplitude. The roll-off in a vertical may be more rapid, depend=
ing on=20
the type / feedback.
One=20
  problem is that there is so much microseism noise in the six-second 
=
region=20
  and also around 12 seconds.  I generally use a 0.08Hz (12.5 second)=
 
4=20
  or 6 pole low pass filter to cut out most of that noise, while allowing=
=20
  
through much of the big-quake frequencies.  An instrument that=
 can=20
  only go 
down to 6 seconds will unfortunately be quite good at displa=
ying=20
  the noise 
while missing most teleseism surface waves.
    Yup!   I only see 10 second noise&n=
bsp;on=20
a few days each month in the UK. The six second noise batters us all the=
=20
time!
>  One's location has an influence on which sensor=
 type=20
  works for them

The only problem is that the movement of a seismom=
eter=20
  in response to an 
error force, such as from small temperature change=
s,=20
  etc. increases as 
1/Frequency^2    
    Even more critical are air density changes=20
effecting a vertical sensor and also wind noise. Temperatures should chang=
e=20
quite slowly inside an insulating case, assuming that the sun does no=
t=20
shine on it!
It is=20
  way harder to make a stable instrument for 50 
seconds than for 10 or=
 20 or=20
  6.  Long period verticals almost certainly 
need to use a feedba=
ck=20
  design if they are going to be sensitive enough to 
see distant quake=
s=20
  while at the same time, insensitive enough to 
temperature and other=
=20
  variations to not wander off to maximum output.


    Uh, Uh! Now define what you are calling long=
=20
period? 
 
    The Roberts' period compensating amplifier ci=
rcuit=20
is probably the easiest technique for amateurs to use. It has constant gai=
n from=20
the LP filter down to ~ the resonant frequency rf of the vertical sen=
sor,=20
at say 1/2 Hz. See http://jclahr.co=
m/science/psn/roberts/index.html =20
It was first used on geophones.=20
    Below this to about rf / 10, (=3D 1/20 Hz) th=
e gain=20
increases as 1 / f^2, so compensating the f^2 falling output to give a fla=
t=20
characteristic. So you can extend a 2 second period vertical sensor out to=
 about=20
20 seconds quite easily. Extending the period much beyond x10 quickly runs=
 into=20
noise problems with a coil + magnet velocity detector. You need two=
 of=20
these stages with maximum gains of x10 (total x100 at 20=20
seconds) linked by a 2 pole high pass filter at ~ 1/30 Hz. Lennartz=
 use a=20
system like this. So do I and it works fine.
    Direct digital period compensation probably=
 works=20
best with 24 bit ADCs. The French use it on their schools system. 16 bit=
 ADCs=20
may only show a few counts (1/100 the rest of the signal) at x10 period,=
 unless=20
the gain is quite high and it may be partly masked by electronic=20
noise.
 
    Regards,
 
    Chris Chapman