PSN-L Email List Message

Subject: Re: 'soft force feedback'
From: "James L. Gundersen" jgundie@.......
Date: Sun, 22 Feb 2009 11:24:39 -0800

Hi Geoff,

I'll put my two cents worth in.  I'm prejudice to using capacitive pickup as 
you may know.  Back in "75" when Barry and I built my first seismograph I 
choose to use a capacitive pickup because the coil and magnet combination 
signal amplitude drops off as the frequency goes down for a given amount of 
earth movement.  Today I believe I can build a seismograph where the noise 
level of the electronics will be below the earth's seismic noise levels at 
all frequencies down to and including DC.  I say "I believe" because I have 
not "had time" to build one yet but my analysis indicates so.

One thing that just occurred to me that's probably not been said about a 
capacitive sensor is it that it can be considered a "parametric amplifier" 
of the earth movement one is sensing.  This a very important advantage over 
the coil and magnet approach.  The coil basically creates a weak output 
signal from "weak movement" that gets 10,000 times (power wise) weaker when 
the frequency drops a 100 times.  (Power is the key factor for minimum 
electrical signal detection in a given bandwidth not voltage per se.)  The 
output level of the parametric (capacitive) sensor can be increased by 
simply increasing the drive level into the sensor (current into the 
capacitors) to improve the signal-to-noise ratio if needed.  In the case of 
capacitive sensors the effective current is the product of the p-p drive 
voltage level, the frequency of the drive level and the amount of 
capacitance of the sensor.  (There is more naturally like the capacitor 
plate seperation that's very important.)  So with a capacitive sensor if my 
memory serves me right you can detect changes in position of much less than 
a wavelength of light and I believe down to atomic dimensions.  I won't take 
time to find the numbers but I believe these dimensions (sensitivities) can 
be made much less the seismic noise level at all frequencies.  The key 
advantage is the "pumping" power level of the parametric can be increased to 
the power level of the detected signal to a far greater the power level than 
the magnetic sensor's corresponding signal level.  The reference noise level 
for the electronic is basically the thermal noise power level corresponding 
to the signals bandwidth hence the importance of power per se.

What prompted to write was your concern for RFI which I share.  With the 
capacitive sensor and associated electronic can certainly be susceptible to 
RFI or be a source of RFI.  My experience has been this can be controlled or 
eliminated by design and construction.  There is a lot to EMC 
(electro-magnetic compatibility, the other acronym) design in general but 
the key to not contaminating the rest of the world is to enclose the 
capacitive sensor and its electronics in shielding.  That is put the 
electronics in a metal box (it can be small now days with the size of ICs) 
and us shielded wiring (coax  shielded cable) to and from the sensor.  Then 
you can add some shielding to the sensor its self ( I've not done that but 
its a good idea).  This takes care of the emissions and the nice thing about 
EMC is that it by reciprocity it also eliminates the susceptibility concerns 
(or a least minimizes them).

An interesting note along this line is to assume for instance the oscillator 
is "too" near the sensor electronics and not shielded well enough (isolated) 
from the sensor electronics.  The coupling tends to create a DC error only 
which should be very stable and small.  This DC gets subtracted with all the 
other DC offsets when the signals are detected.

The thing I would like to do when I get done with the project I'm on for 
past year is build a seismograph using my favorite ADC.  It has a linearity 
of 0.0005% and a noise level at a 1 KHz sample rate of 1 uv (more than 120 
db below its full scale signal).  The big advantage of course of using 
digital electronics is I can integrate the signal for a minute or an hour or 
24 hours without the drift problems of an analog design. (Yes I can build an 
analog integrator than works beyond an hour but its difficult, expensive and 
easily compromised.)  Such integration can greatly reduce the electronic 
noise to get long period detection sensitivity.

I should mention also building the capacitor sensor electronic can be 
somewhat of a pain but it much easier today than in the past.

I also must say the capacitive sensor tends to be good for small movements 
but has limits for the large displacements one can get with the long period 
seismographs.  Hence in the past I've ended up detecting the acceleration at 
low frequencies and integrating to recover velocity and displacement 
information.  This puts a real significant limitation on the minimum on my 
minimum detectable earth movement argument but I think the numbers still 
work out.

Another comment on this type design without feedback is it will saturate on 
large signals (local earth quakes) and therefore rather than try to perfect 
the feedback approach I think I may simply build a second seismograph for 
strong signals probably based on acceleration and not displacement to avoid 
saturation.  In a strong earth the acceleration can exceed 1 g.  I'm 
expecting the big one some day here in Southern California; we're over due I 
suscept in someways?

Another thought on why one design can't do it all.  The dynamic range of 
electrical signal is limited by thermal noise level on the hand and maximum 
voltage or power level on the other.  Once electrical signal ranges approach 
levels like 120 db (a million-to-one) or 140 db the electronics can become a 
limit.  I believe if one studies the dynamic range of seimic signals they 
can exceed 140 db particularily if start including strong local activity 
versus the weakest long period signals and want narrow band (low frequency) 
minimum noise but also what the broad band (high frequency) local strong 
earth quakes.  I'll have to revisit this area sometimes because the digital 
processing can circumvent some of the some dynamic range problem along with 
active force feed back.


PS  I have no idea what your background is.  So excuse me if some things are 
too simple or esoteric.  Mine is electrical engineering.

----- Original Message ----- 
From: "GMV" 
Sent: Sunday, February 22, 2009 6:38 AM
Subject: Re: 'soft force feedback'

> Hello Randall Peters;
> Have you ever used an industrial control PID loop
> kind of thing with a Hall Effect sensor ?
> I would imagine it may work quite well to keep a mass
> locked in one position then look at the energy expended
> to keep the mass locked when anything tries to move it.
> You talk capacitive sensor but that is an active device
> purring out all kinds of RFI in an already saturated RFI world.
> I am interested in all forms of passive or baseband
> devices that do not need artificially applied AC to make them work.
> PID meaning proportional-integral-derivative.
> Amplifier/integrator combination.
> Possibly more complex.
> I have seen PID loops in auto cruise controls and
> in food service industry to control flow rates
> and temps. I imagine PID loops along with 20ma
> circuits can do about any kind of control.
> regards;
> geoff
> ----- Original Message ----- 
> From: "Randall Peters" 
> To: 
> Sent: Sunday, February 22, 2009 7:15 AM
> Subject: 'soft force feedback'
> Charles,
>   What you have indicated is indeed what I have used with a fully 
> differential capacitive sensor monitoring the displacement of my modified 
> Sprengnether (zero-length, Lacoste) vertical seismometer.  The output from 
> the sensor goes to an opamp integrator, whose output is a very weak 
> correction signal  (fed in turn to the original coil/magnet sensor, now 
> acting as an actuator) to keep the system from 'going to the rails' of my 
> capacitive sensor.
>     As I have noted previously, to operate with a PID feedback and then 
> use the (so called 'velocity' (really 'jerk' below the corner frequency) 
> output only-destroys low frequency response.  This 'pulls out the 
> frequency multiplier term' by the chain rule of differentiation, causing 
> the response to go to zero as the frequency goes to zero.  I teach my 
> students to recognize the important differences between differentiation 
> and integration when it comes to electronic signals containing noise. 
> The former is a 'noise enhancer' and the latter a 'noise reducer', as is 
> well known to anybody who has looked at their differences using an 
> oscilloscope.
>   About the differences between 'force balance' and 'soft feedback'. 
> Force balance is 'hard' in the sense that ideally there is no motion of 
> the seismic mass whatsoever.  The feedback signal is so strong that it 
> allows one to monitor the 'error' value required to eliminate motion-as 
> representative of what the mass would do if allowed to move in an ideal 
> Hooke's law oscillator.
>   Unfortunately, there are no Hooke's law oscillators.  It has taken me a 
> long time for the scientific community to begin finally accepting my 
> claims concerning mesoanelastic complexity.  There are two types of 
> anharmonicity, (i) elastic and (ii) damping.  Many of you know about (i) 
> since a big, close earthquake will cause anomalous response from any 
> seismometer, because it is afflicted (large motions) with a restoring 
> feature that is not perfectly harmonic.   When seismic disturbances are 
> 'low and slow', meaning low frequency as well as small amplitude, the 
> 'corrugation-like' features of the restoration potential come into play. 
> Engineers know about 'dithering' as a means to combat friction effects. 
> In effect, that is what I recommend.  It is advantageous to let the system 
> 'skate' over the metastabilties of internal friction type, some of which 
> can cause the system to be effectively 'latched' against being able to see 
> the low/slow signals.
>    For my Sprengnether, the time constant of the ompamp integrator was set 
> at several hundred seconds, so as you say, to integrate in a lower range 
> than the one of interest.  My approach to this is not the first.  Erhard 
> Wielandt mentioned at the IRIS Broadband Conference that a German 
> seismology team did effectively the same thing about a hundred years ago. 
> They used water (probably hundreds of gallons) in a feedback scheme to 
> alter the tilt of their seismic platform to keep the instrument from going 
> to the rail because of the adversities of (i) buoyancy of air pressure 
> changes associated with moving fronts, and (ii) temperature changes 
> altering the modulus of the spring.
>   Randall
> __________________________________________________________
> 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 ]