In a message dated 25/03/2006, gmvoeth@........... writes:

Can  someone out there run down a list of best ways to increase the signal to 
 noise
ratio of a seismic sensor?
Hi Geoff,
 
    Wow, a BIG subject! Difficult to cover  everything.

Things  like:
1. in a mass spring system use a larger mass
    Above about 35 gm, the intrinsic kT noise of the  mass is less than 
natural noise. For 'safety', use maybe 5x to 10x this as  minimum.
    For a spring / mass system the ends of the spring  need to be straight 
wires effectively clamped. The same applies to suspension  systems. Good clamp 
design is important. 
    The mass should preferably be non magnetic.
    No horse shoe or open pole magnets should be  mounted on the armature. 
This is just asking for trouble and you  will get it! Mount the sensor and 
damping magnets on the baseplate  and put the coils and damping plates on the arm. 
Note that carefully  designed 'potcore' magnets are used as the mass in some 
seismometers. These  do not have a high external magnetic field.

2. use a  phased array
    This would allow you to cancel much of the  local environmental noise if 
it is physically large, but it is a major  undertaking. 

3. use  low resistance's in the circuitry
    Choose the amplifier noise impedance to  roughly match the sensor 
resistance. 

4. use a  narrow bandwidth
    This can be fairly critical. However it varies with  the local noise and 
the range of the signal that you want to detect. For local  and near regional 
signals you need maybe 10 to 20 Hz. For teleseismic quakes you  may only need 
2 to 3 Hz, or less.
    In noisy situations you need properly matched  fairly sharp cut-off 
multipole filters. 
    You may also wish to provide a high pass filter to  reduce VLF 1/f noise.

5. use a  high Q
    Oscillators for driving sensor systems need to have  very good frequency 
and amplitude stability.

6.  shield everything against anything not a seismic signal.
    Agreed, but you do need a single effective earthing  point, usually 
linked to the input of the first opamp where you have the minimum  signal. 

7. a  high signal to noise ratio is not possible for the average amateur due 
to  cost
it might take a bunch of amateurs to pool their moneys  and talents to
create a decent seismic sensor, possibly such a  thing is only possible
at a university where both money and  talent exists.
    I disagree entirely. 
    Amateurs should be able to achieve instrument noise  levels below their 
local ambient seismic levels without great  difficulty. 
    The more likely problems are interference, RF,  magnetic or static 
electric, poor earthing, poor shielding, utility  supply signals and noise and local 
environmental noise commonly due to  wind, lightning and static electric 
discharge, human activity including ordinary  movement, cars, lorries and other 
road traffic, trains, road and building work,  quarrying, power plants, heavy 
machinery, mining.... You need to try to identify  the noise sources.
    It may be that you need to treat your site as  though it were a heavy 
engineering factory and use isolating transformers and  filters for supplying 
your system and the computer, sealed metal cases, common  point earthing and 
braided screened connecting cables.
    It is no accident that most seismic sites  are well away from human 
activity and fully encased borehole instruments  are often used, buried maybe 100 m 
below ground level. This eliminates a great  deal of the surface and weather 
related noise. 
 
    However, the design of sensor and damping systems  is very important. It 
is quite easy to increase the output of a system which  used a coil and a 
horse shoe magnet, by using a NdFeB magnet array and a smaller  coil. 
    
    For a Lehman, I use two 1/4" thick mild steel  plates, 3.5" long by 2" 
wide. The corners are drilled to take 1/4" mild steel  set screws 2.5" long with 
mild steel washers and nuts. The two plates are  held maybe 3/4" to 1" apart 
for a sensor, using three nuts on every bolt.  A bolt is firmly secured to one 
plate using a nut. The second plate is  mounted in between pairs of nuts on 
the free thread, to allow the separation of  the plates to be adjusted. Four 
NdFeB magnets 1/8" thick by 1" square are  mounted on the inside faces of the 
plates. A N+S pair on one face is  opposed by a S+N pair on the other. The 
sensor coil is mounted in the high  intensity magnetic field in the centre.  
    This construction gives quite an effective magnetic  and electrostatic 
screen around the coil. The external stray field is low. The  sensitivity is 
high due to the high field and the response is reasonably linear.  If you use 
rectangular instead of round coils, it can be made very highly linear  over 
+/-1/2" movement.
    For induced current damping, I  use four NdFeB bar magnets 1"x1/2"x1/4" 
thick in two opposing squares,  with the same 1/4" steel plate mounting. I use 
a copper damping plate,  1/16" to 1/8" thick as appropriate, a bit over 2" 
wide and 2.5"  free length. This allows the arm to swing +/-1/2" without the edge 
of the copper  plate overlapping the edge of the 1" magnet square. The N+S 
join of the magnet  pairs is set perpendicular to the direction of motion. The 
damping is adjusted  by varying by the length of copper tongue overlapping the 
magnet square and also  by varying the separation of the 1/4" mild steel 
plates and hence the magnet  separations.

Does any  know of a system out there with the
highest signal to noise ratio...if so  can we see it ?

More then likely it is in military hands and is secret  ?
    More likely to be a commercial secret of the  various seismometer 
manufacturers! It takes quite a bit of effort to get the  instrument noise below the 
minimum earth noise levels. However, unless you  are very fortunate with your 
site, you are very unlikely to see seismic noise  levels even approaching that 
low.

 
    Try either wedging your sensor in a fixed position  or substituting a 
resistor for it and running the whole system for 24 hrs on a  weekday. This 
should demonstrate the amplitude and timing of many interfering  signals. But it 
won't pick up sensor movement due to drafts, local magnetic  field changes 
(including the earth, your refrigerators, car, bicycle or mowing  machine), ground 
movement or tilt, insects or animals. Then you have the  interesting task of 
identifying these sources and eliminating the effects.  Compare the signals and 
timing with an active trace taken a seismically quiet  day.
 
    Get onto the www and see how ham radio operators  deal with earthing in 
your area, also the utility company. There are large areas  of the US where the 
rocks and soils are dry and have very poor electrical  conductivity. Your 
house or mobile home wiring may be sticking way up above the  effective local 
'earth' plane. 
 
    For noise coming in through the utility wiring, you  can provide a 
filter, but check that this has a good rejection from a few  hundred Hz up and is 
not simply an RF filter. The next stage is to use a 1:1  isolating transformer 
with a metal case and an electrostatic screen  in between the windings. 
Consider providing a separate Earth for  the screen, the case and the electronics. 
The more extreme alternative  is to use two batteries, one to drive the 
equipment and a laptop computer while  the other is charged on a separate power 
circuit. An alternative is to use  a large ferrite core transformer, similar to that 
in a TV set, to provide a 15  kHz isolated charging circuit. You can also use 
optical or radio links to  transfer data to and from the phone system.
 
    Hope that these comments / suggestions are of some  help.
 
    Regards,
 
    Chris Chapman





In a message dated 25/03/2006, gmvoeth@........... writes:
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>Can=20
  someone out there run down a list of best ways to increase the signal to=20
  noise
ratio of a seismic sensor?
Hi Geoff,
 
    Wow, a BIG subject! Difficult to cover=20
everything.
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>Things=20
  like:
1. in a mass spring system use a larger mass
    Above about 35 gm, the intrinsic kT noise of th=
e=20
mass is less than natural noise. For 'safety', use maybe 5x to 10x this as=20
minimum.
    For a spring / mass system the ends of the spri=
ng=20
need to be straight wires effectively clamped. The same applies to suspensio=
n=20
systems. Good clamp design is important. 
    The mass should preferably be non magnetic.
    No horse shoe or open pole magnets should be=20
mounted on the armature. This is just asking for trouble and you=
=20
will get it! Mount the sensor and damping magnets on the basep=
late=20
and put the coils and damping plates on the arm. Note that carefully=20
designed 'potcore' magnets are used as the mass in some seismometers. T=
hese=20
do not have a high external magnetic field.
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>2. use a=20
  phased array
    This would allow you to cancel much of the=20
local environmental noise if it is physically large, but it is a m=
ajor=20
undertaking. 
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>3. use=20
  low resistance's in the circuitry
    Choose the amplifier noise impedance to=20
roughly match the sensor resistance. 
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>4. use a=20
  narrow bandwidth
    This can be fairly critical. However it varies=20=
with=20
the local noise and the range of the signal that you want to detect. For loc=
al=20
and near regional signals you need maybe 10 to 20 Hz. For teleseismic quakes=
 you=20
may only need 2 to 3 Hz, or less.
    In noisy situations you need properly matched=20
fairly sharp cut-off multipole filters. 
    You may also wish to provide a high pass filter=
 to=20
reduce VLF 1/f noise.
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>5. use a=20
  high Q
    Oscillators for driving sensor systems need to=20=
have=20
very good frequency and amplitude stability.
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>6.=20
  shield everything against anything not a seismic signal.
    Agreed, but you do need a single effective eart=
hing=20
point, usually linked to the input of the first opamp where you have the min=
imum=20
signal. 
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>7. a=20
  high signal to noise ratio is not possible for the average amateur due to=20
  cost
   it might take a bunch of amateurs to pool their money=
s=20
  and talents to
   create a decent seismic sensor, possibly su=
ch a=20
  thing is only possible
   at a university where both money an=
d=20
  talent exists.
    I disagree entirely. 
    Amateurs should be able to achieve instrument n=
oise=20
levels below their local ambient seismic levels without great=20
difficulty. 
    The more likely problems are interference,=
 RF,=20
magnetic or static electric, poor earthing, poor shielding, utility=20
supply signals and noise and local environmental noise commonly du=
e to=20
wind, lightning and static electric discharge, human activity including ordi=
nary=20
movement, cars, lorries and other road traffic, trains, road and building wo=
rk,=20
quarrying, power plants, heavy machinery, mining.... You need to try to iden=
tify=20
the noise sources.
    It may be that you need to treat your site as=20
though it were a heavy engineering factory and use isolating transformers an=
d=20
filters for supplying your system and the computer, sealed metal cases, comm=
on=20
point earthing and braided screened connecting cables.
    It is no accident that most seismic sites=20
are well away from human activity and fully encased borehole instrument=
s=20
are often used, buried maybe 100 m below ground level. This eliminates a gre=
at=20
deal of the surface and weather related noise. 
 
    However, the design of sensor and damping syste=
ms=20
is very important. It is quite easy to increase the output of a system which=
=20
used a coil and a horse shoe magnet, by using a NdFeB magnet array and a sma=
ller=20
coil. 
    
    For a Lehman, I use two 1/4" thick mild steel=20
plates, 3.5" long by 2" wide. The corners are drilled to take 1/4" mild stee=
l=20
set screws 2.5" long with mild steel washers and nuts. The two plates a=
re=20
held maybe 3/4" to 1" apart for a sensor, using three nuts on every bolt.=20
A bolt is firmly secured to one plate using a nut. The second plate is=20
mounted in between pairs of nuts on the free thread, to allow the separation=
 of=20
the plates to be adjusted. Four NdFeB magnets 1/8" thick by 1" square are=20
mounted on the inside faces of the plates. A N+S pair on one face is=20
opposed by a S+N pair on the other. The sensor coil is mounted in the high=20
intensity magnetic field in the centre.  
    This construction gives quite an effective magn=
etic=20
and electrostatic screen around the coil. The external stray field is low. T=
he=20
sensitivity is high due to the high field and the response is reasonably lin=
ear.=20
If you use rectangular instead of round coils, it can be made very highly li=
near=20
over +/-1/2" movement.
    For induced current damping, I=20
use four NdFeB bar magnets 1"x1/2"x1/4" thick in two opposing squa=
res,=20
with the same 1/4" steel plate mounting. I use a copper damping plate,=20
1/16" to 1/8" thick as appropriate, a bit over 2" wide and 2.=
5"=20
free length. This allows the arm to swing +/-1/2" without the edge of the co=
pper=20
plate overlapping the edge of the 1" magnet square. The N+S join of the magn=
et=20
pairs is set perpendicular to the direction of motion. The damping is adjust=
ed=20
by varying by the length of copper tongue overlapping the magnet square and=20=
also=20
by varying the separation of the 1/4" mild steel plates and hence the magnet=
=20
separations.
<=
FONT=20
  style=3D"BACKGROUND-COLOR: transparent" face=3DArial color=3D#000000 size=
=3D2>Does any=20
  know of a system out there with the
highest signal to noise ratio...if=20=
so=20
  can we see it ?

More then likely it is in military hands and is sec=
ret=20
  ?
    More likely to be a commercial secret of t=
he=20
various seismometer manufacturers! It takes quite a bit of effort to get the=
=20
instrument noise below the minimum earth noise levels. However, unless=20=
you=20
are very fortunate with your site, you are very unlikely to see seismic nois=
e=20
levels even approaching that low.

 
    Try either wedging your sensor in a fixed posit=
ion=20
or substituting a resistor for it and running the whole system for 24 hrs on=
 a=20
weekday. This should demonstrate the amplitude and timing of many interferin=
g=20
signals. But it won't pick up sensor movement due to drafts, local magnetic=20
field changes (including the earth, your refrigerators, car, bicycle or mowi=
ng=20
machine), ground movement or tilt, insects or animals. Then you have the=20
interesting task of identifying these sources and eliminating the effects.=20
Compare the signals and timing with an active trace taken a seismically quie=
t=20
day.
 
    Get onto the www and see how ham radio operator=
s=20
deal with earthing in your area, also the utility company. There are large a=
reas=20
of the US where the rocks and soils are dry and have very poor electrical=20
conductivity. Your house or mobile home wiring may be sticking way up above=20=
the=20
effective local 'earth' plane. 
 
    For noise coming in through the utility wiring,=
 you=20
can provide a filter, but check that this has a good rejection from a few=20
hundred Hz up and is not simply an RF filter. The next stage is to use a 1:1=
=20
isolating transformer with a metal case and an electrostatic sc=
reen=20
in between the windings. Consider providing a separate Earth f=
or=20
the screen, the case and the electronics. The more extreme alterna=
tive=20
is to use two batteries, one to drive the equipment and a laptop computer wh=
ile=20
the other is charged on a separate power circuit. An alternative is to=20=
use=20
a large ferrite core transformer, similar to that in a TV set, to provide a=20=
15=20
kHz isolated charging circuit. You can also use optical or radio links=20=
to=20
transfer data to and from the phone system.
 
    Hope that these comments / suggestions are of s=
ome=20
help.
 
    Regards,
 
    Chris Chapman