Chris wrote, "There is NO PROBLEM in this case. Didn't you check the
data sheet?"  =

I have to admit I was trying to take the lazy way out and just ask the
question and so I didn't check the spec sheet.  I went back and checked
out the data sheet for the 3515 and 3516, (27501.10B) and did not see
any statements about magnetic or non-magnetic leads - so where does that
information appear?  In any case, I decided to take on a longer
discussion of why I made some of the statements I did.  That discussion
follows answers to the other question, "What other companies are
producing devices similar to the MS3110, please?"   =

Analog Microelectronics GmbH in Mainz, Germany, has developed a family
of analog ASICs, the CAV404, CAV 414 and CAV424.  Downloadable spec
sheets at:
http://www.servoflo.com/analog/data_ho.htm
And two articles in Sensors magazine on these ICs.  =

http://www.sensorsmag.com/articles/0201/88/main.shtml
http://www.sensorsmag.com/articles/0402/capacitive/main.shtml

I seem to remember that SensNor had one also, but all I could find was:
http://www.isit.fhg.de/
which seems to be a custom silicon design house.  So make that one other
company making capacitive sense ICs for sale on the open market.  Some
companies have a captive design, for instance, SETRA company makes their
own sensor IC, but does not sell it alone, only in their instrumentation
products. =



So on to the theory.  The "New Manual of Seismological Observatory
Practice" which covers many topics is at:
http://www.seismo.com/msop/nmsop/nmsop.html
and discusses many areas of seismometer construction, but in particular
the noise floor discussion at:
http://www.geophys.uni-stuttgart.de/seismometry/man_html/node28.html
and in a discussion on electronic displacement sensors the statement,
"Their sensitivity is limited by the ratio between the electronic noise
of the demodulator and the electrical field strength; it can be a
hundred times better than that of the inductive type. The comprehensive
paper by [Jones & Richards1973] on the design of capacitive transducers
still represents the state of the art in all essential aspects." is at:
http://www.geophys.uni-stuttgart.de/seismometry/man_html/node21.html =

Part of the reason a capacitive, or even an inductive sensor, can
surpass the Hall effect or LED shutter types is that at the conversion
point from a mechanical quantity to an electrical one, there is no noise
in the capacitive sensor because it is reactive.  A bit of explanation. =

There can be several noise sources in a sensor application such as a
seismometer.  External, such as the one discussed by Sean-Thomas (in the
email mentioned below) set one limit.  If your electronics were perfect,
you can't do any better than your external noise floor.  But many times
there are additional local sources, which can amount to traffic in the
street, trees waving in the breeze, etc.  Again these can become the
noise floor.  But next comes the internally generated noise sources. =

The first one is thermal.  The arm mass of your seismometer has a
temperature and is "vibrating".  In addition the air around that mass
has Brownian motion and can push on the mass generating noise.  This
source is problem for miniature sensors such as the Analog Devices MEMs
types.  Barometric pressure changes "float" the mass which can translate
into movement, especially on vertical seismometers.  Next comes the
electronic noise -- the heart of the discussion that prompts this
reply.  A non-reactive (resistive) sensor such as a Hall effect sensor
or photosensor vs a reactive sensor such as capacitive or inductive
sensor suffer from at least two major noise sources, thermal and current
noise.  These are physical, inherent properties.  The thermal noise is
proportional to the square root of the product of the resistance,
temperature and BW.  Amounting to 1.29 nV for 100 ohms, 1 Hz BW, and 295
degrees K.  =

Current noise is proportional to current and inversely proportional to
the frequency of measurement.  I.e., it goes up as the frequency goes
down and is somewhat dependent on the construction/material of the
resistance.  =


So glancing at the A3515LUA, we find that this floor is 400 uV for a 10
to 10KHz BW in a +/- 2V range.  This limits your total range to 10,000
to 1, or about 80 dB.  Capacitive and inductive displacement sensors do
not have to have either of these noise sources since they are
essentially reactive, and the A3515LUA noise sources are due to
resistance.
A more subtle problem is another noise called the 1/f or flicker noise
problem in semiconductors.  This noise also rises as the frequency goes
down.  So the numbers for the A3515LUA will almost surely get worse as
we go down in frequency for a seismometer.  In particular, 10 Hz is
high, when we want to look at tele-seismic events, which is exactly
where we need good low noise performance.  =

One way around 1/f noise is the carrier amplifier, i.e., if you put the
information on a carrier frequency, amplify this carrier frequency, then
demodulate, you step around the 1/f problem.  Coincidentally, this is
exactly what happens in capacitive and inductive displacement sensors. =

A high frequency is impressed across the sensor capacitor (or
inductor).  This capacitor changes value according to the displacement,
yielding (modulating) that high frequency (carrier) which is amplified
by a carrier amplifier.  The output of that carrier amplifier is then
detected at a high level where the 1/f, thermal and current noises are
well below the output.  So, you'll note that Sean-Thomas used an
inductive sensor on his vertical seismometer.  That is one route.
Another is that you could easily build the design idea "Circuit resolves
0.1-fF change from 100 pF by Derek Redmayne, Linear Technology Corp,
Milpitas, CA  (from EDN Access Design Ideas1/6/2000)" at:
www.e-insite.net/ednmag/index.asp?layout=3Darticle&articleId=3DCA46462&pu=
bdate=3D01/06/2000
The major cost of the semis in quantities of one would be about $32, but
this includes a 24 bit digitizer sampling at 7.5 samples/second. =

Resolving a 0.1 fF change in 100 pF is a 1,000,000 to 1, or 120 dB
range.  One measure to decide if this is sufficient.  Sean-Thomas
mentioned in an email post of Nov 15, 2000 to the PSN list that, "The
normal background noise for a seismometer like a Lehman are the 6-second
microseisms, usually caused by storms off the east coast.  Away from the
immediate shore (100km) these run 2 to 4 microns peak-to-peak=85"  So for=

ease of computation, let's set a goal of 1 micron.  Therefore the
capacitor plates could be set apart by 1e6 * 1micron, or 1 meter and we
could still resolve the mechanical background.  Practical construction
might be to have parallel plates about 10 mm away on each side (~1/2
inch) from a moving plate on a SG pendulum.  Then for 100 pf between the
plates we need:
C =3D 2.249E-13 * Er * sq.in * (N-1) / (in of separation)
so rearranging, substituting and solving we get that we need 222 sq.in
of area.  Not good.  but lets reduce the spacing to about 0.05" and we
get 22 sq.in or a plate a bit over 4.5" square.  This is doable.  A
variation would be to salvage the parallel plates of a variable tuning
capacitor such as in a broadcast receiver.  Fully meshed, the large gang
was 365 pF.  So, mount so the pendulum meshes the group.  Now you have
perhaps a 1" range with about 300 pF change.  =

Another solution would be to implement a slightly different circuit such
as the one in the SETRA patents: 4,054,833: 5,194,819: and 6,316,948. =

Now the amplifier can provide gain, and the voltage tracks the ratio of
bridge caps, so small capacitance changes generate larger voltages and
therefore you could use smaller single plate sensors.
An example of this was the design idea "Bridge Measures Small
Capacitance" by Jeff Witt, Linear Technology Corp, in Electronic Design,
Nov. 4, 1996, pg. 110.  This circuit seemed to be similar to the SETRA
circuits and allow gain.  Unfortunately, it does not seem to be
available on the web or on Linear Technology's site, even though it was
submitted by one of their employees and again used the LTC1043.  This
final circuit is probably the best compromise of the lot.  Take it and
add the 24 bit A/D, LTC2400, and you have a powerful combination for low
cost.  Since the initial output is analog, you have a node that can be
fed back to the force coil on the SG and at the same time you get a
simple digital stream to feed your computer. =

If there's enough interest, I'll scan and make a PDF file and either
direct email it, or maybe Larry can warehouse it.

Anyway,enough, just some more food for thought.  I hope this was
clearer, but I often jump over the details from impatience.   Oh, yes,
an excellent reference for capacitance sensor design:
"Capacitive Sensors, Design and Applications," by Larry K. Baxter, IEEE
Press.

Regards,
Charles R. Patton
__________________________________________________________

Public Seismic Network Mailing List (PSN-L)