PSN-L Email List Message

Subject: Re: Tests on folded pendulum instrument
From: ChrisAtUpw@.......
Date: Mon, 17 Mar 2003 23:59:50 EST


In a message dated 16/03/03, dyouden@......... writes:

> I got a chance yesterday to run some tests on my instrument.
> Here's what I did:
> 
> I opened the loop and drove the force balance coil directly with a HP 
> function generator ( Model 3310B). I placed a 1000 ohm resistor in series 
> with the coil. I took my input signal from across the coil so I had a 
> sensitivity of 1000 volts/amp. The output was taken at the output of the 
> cap gauge circuit, after a 2 pole 35 Hz filter that would have been 
> inconvenient to remove. The output device was a HP 54603B memory scope. The 
> driven mass is pretty close to 500 grams.
> 
> I attempted to take phase and P-P position data at constant acceleration 
> from 0.1 Hz to 10 Hz while holding the current through the coil at 94 
> microamps, i. e. 94 millivolts across 1000 ohms. Why that value? Because it 
> was the maximum that I could use at 0.1 Hz and still maintain linearity of 
> the cap gauge signal.
> 
> Results:
> The force balance coil/magnet combination have a transfer function of 0.26 
> 

Hi Dave, 

       It should be possible to increase this force without too much 
difficulty using a soft iron rectangular chamber to enclose the magnet and 
the coil. This would also make the force nearly independant of the coil 
position. Other PSN equipment has attained 10 to >30 N/A. 

> I lost the signal in the noise at about 6 Hz. Obviously, I could go higher 
> with a larger input signal.
> 
> The phase between acceleration and displacement was less than 20 degrees 
> throughout the range.
> The P-P displacement was about 60 nanometers at 6 Hz.
> 
> The laws of physics still apply, as the P-P position signal was inversely 
> proportional to the square of the frequency. There was no sign of roll-off 
> due to damping from the cap gauge, but then the amplitudes were pretty low 
> at the higher frequencies.
> 
> Comment: My low pass filters are set to roll off at 5 Hz so data beyond 
> that point is academic.

       Measuring any damping due to the capacitative sensor plates in 
operation might be done by setting the pendulum frequency to say 1 Hz, 
disconnecting the feedback and putting a small step current through the force 
feedback coil. If there is no damping, the system would oscillate for some 
time. Dr.Blair used a photodiode setup for measuring the sensor 
characteristics. 

      The reason that I query this, is that one limitation on the use of 
capacitative sensors which use changing plate separations in air, is the gas 
pumping effect due to the moving plates. I would expect a significant effect 
with a 2" square plate and a gap of 10 thou. at 10 Hz. One way of reducing 
the problem is to drill an array of holes in the central plate to let air 
flow more easily between opposite sides of the plate. This need not greatly 
effect the capacity. 

      The damping force F = 3.n.u.A^2 / 2.Pi.x^3 for circular flat electrodes 
according to Jones, where n is the viscosity of air, u is the plate velocity, 
A is the plate area and x is the plate separation. The damping falls off with 
the cube of the plate separation and increases with the square of the Area, 
which suggests two ways of controlling any damping. The problem that I can 
see is that any air damping effects will be frequency and amplitude 
sensitive, which could limit the maximum seismic frequency that you can 
observe. 

       Your force feedback loop should dramatically reduce the amplitude of 
the armature swings and keep the mass nearly still. 
       Just adding critical damping to the response still allows quite large 
amplitude movements.

       What period is the pendulum set at?
       How did you determine the calibration and linear range of your 
capacitor sensor? What sensitivity and linear range do you get? The +/-250 
micron max possible movement range, with maybe only a fraction of this giving 
a linear output, may be a bit restricting.
       You might try using nylon monofilament fishing line as a spacer for 
setting the capacitor gaps.
       A rough calculation of the bridge suggests that the output becomes non 
linear for any movement greater than 1/3 of the gap. A bridge circuit may not 
be the most sensitive and linear option. Jones just used a charge amplifier 
in your reference. Linear Technology have several capacitance measuring 
circuits in their application notes.
       What amplitude do you measure for the 6 second ocean microseisms?
       I would not expect the test you describe to show significant damping. 
The damping force due to the air gaps depends on the velocity of the plate 
and this will be very low for micron movements and low frequencies. Real 
quakes could be a lot larger.
       This is why I suggested that you put a small DC current through the 
coil to deflect the arm and then track the deflection to measure the 
decrement of any oscillations. Maybe using your A/D and Winquake?  

       Some notes on the Circuit.

       The back to back diodes will not give a very stable oscillator 
amplitude over a wide temperature range. An XR8038 sine wave generator IC 
might be better and simpler, particularly if you add a two pole output 
filter. Circuits for very high stability Wein bridge oscillators are 
available and there are several other options.

       The INA121 is right at it's frequency limit at a gain of 100 at 50 
KHz. It's CMRR is only ~65 dB at this frequency. You might consider the high 
speed INA111?

       You might get improved results and less noise if you used an OP07 or a 
trimmed TL071 for the signal inverter U7A.

       Low noise analogue switches are available.

       For the lead/direct circuit, you can feed the signal into the 
inverting input with a R and a CR in parallel to separate out the two 
components. An OPA604 low noise fet amplifier will give about 40 mA max and 
could replace the OPA551.

       The integrator has a time constant of about 1 sec. You might wish to 
increase this to cover the lower part of the frequency range. If you take the 
output from the input to the feedback circuit, don't you get a velocity 
response without using an integrator? 
http://www.geophys.uni-stuttgart.de/seismometry/hbk_html/node24.html

       Hope that this is of some interest / help.

       Regards,

       Chris Chapman
In a message dated 16/03/=
03, dyouden@......... writes:


I got a chance yesterday to= run some tests on my instrument.
Here's what I did:

I opened the loop and drove the force balance coil directly with a HP fu= nction generator ( Model 3310B). I placed a 1000 ohm resistor in series with= the coil. I took my input signal from across the coil so I had a sensitivit= y of 1000 volts/amp. The output was taken at the output of the cap gauge cir= cuit, after a 2 pole 35 Hz filter that would have been inconvenient to remov= e. The output device was a HP 54603B memory scope. The driven mass is pretty= close to 500 grams.

I attempted to take phase and P-P position data at constant acceleration= from 0.1 Hz to 10 Hz while holding the current through the coil at 94 micro= amps, i. e. 94 millivolts across 1000 ohms. Why that value? Because it was t= he maximum that I could use at 0.1 Hz and still maintain linearity of the ca= p gauge signal.

Results:
The force balance coil/magnet combination have a transfer function of 0.= 26=20
Newtons/Amp.


Hi Dave,=20

      It should be possible to increase t= his force without too much difficulty using a soft iron rectangular chamber=20= to enclose the magnet and the coil. This would also make the force nearly in= dependant of the coil position. Other PSN equipment has attained 10 to >3= 0 N/A.=20

I lost the signal in the no= ise at about 6 Hz. Obviously, I could go higher with a larger input signal.

The phase between acceleration and displacement was less than 20 degrees= =20
throughout the range.
The P-P displacement was about 60 nanometers at 6 Hz.

The laws of physics still apply, as the P-P position signal was inversel= y proportional to the square of the frequency. There was no sign of roll-off= due to damping from the cap gauge, but then the amplitudes were pretty low=20= at the higher frequencies.

Comment: My low pass filters are set to roll off at 5 Hz so data beyond=20= that point is academic.


      Measuring any damping due to the ca= pacitative sensor plates in operation might be done by setting the pendulum=20= frequency to say 1 Hz, disconnecting the feedback and putting a small step c= urrent through the force feedback coil. If there is no damping, the system w= ould oscillate for some time. Dr.Blair used a photodiode setup for measuring= the sensor characteristics.=20

     The reason that I query this, is that one= limitation on the use of capacitative sensors which use changing plate sepa= rations in air, is the gas pumping effect due to the moving plates. I would=20= expect a significant effect with a 2" square plate and a gap of 10 thou. at=20= 10 Hz. One way of reducing the problem is to drill an array of holes in the=20= central plate to let air flow more easily between opposite sides of the plat= e. This need not greatly effect the capacity.=20

     The damping force F =3D 3.n.u.A^2 / 2.Pi.= x^3 for circular flat electrodes according to Jones, where n is the viscosit= y of air, u is the plate velocity, A is the plate area and x is the plate se= paration. The damping falls off with the cube of the plate separation and in= creases with the square of the Area, which suggests two ways of controlling=20= any damping. The problem that I can see is that any air damping effects will= be frequency and amplitude sensitive, which could limit the maximum seismic= frequency that you can observe.=20

      Your force feedback loop should dra= matically reduce the amplitude of the armature swings and keep the mass near= ly still.=20
      Just adding critical damping to the= response still allows quite large amplitude movements.

      What period is the pendulum set at?
      How did you determine the calibrati= on and linear range of your capacitor sensor? What sensitivity and linear ra= nge do you get? The +/-250 micron max possible movement range, with maybe on= ly a fraction of this giving a linear output, may be a bit restricting.
      You might try using nylon monofilam= ent fishing line as a spacer for setting the capacitor gaps.
      A rough calculation of the bridge s= uggests that the output becomes non linear for any movement greater than 1/3= of the gap. A bridge circuit may not be the most sensitive and linear optio= n. Jones just used a charge amplifier in your reference. Linear Technology h= ave several capacitance measuring circuits in their application notes.
      What amplitude do you measure for t= he 6 second ocean microseisms?
      I would not expect the test you des= cribe to show significant damping. The damping force due to the air gaps dep= ends on the velocity of the plate and this will be very low for micron movem= ents and low frequencies. Real quakes could be a lot larger.
      This is why I suggested that you pu= t a small DC current through the coil to deflect the arm and then track the=20= deflection to measure the decrement of any oscillations. Maybe using your A/= D and Winquake?  

      Some notes on the Circuit.

      The back to back diodes will not gi= ve a very stable oscillator amplitude over a wide temperature range. An XR80= 38 sine wave generator IC might be better and simpler, particularly if you a= dd a two pole output filter. Circuits for very high stability Wein bridge os= cillators are available and there are several other options.

      The INA121 is right at it's frequen= cy limit at a gain of 100 at 50 KHz. It's CMRR is only ~65 dB at this freque= ncy. You might consider the high speed INA111?

      You might get improved results and=20= less noise if you used an OP07 or a trimmed TL071 for the signal inverter U7= A.

      Low noise analogue switches are ava= ilable.

      For the lead/direct circuit, you ca= n feed the signal into the inverting input with a R and a CR in parallel to=20= separate out the two components. An OPA604 low noise fet amplifier will give= about 40 mA max and could replace the OPA551.

      The integrator has a time constant=20= of about 1 sec. You might wish to increase this to cover the lower part of t= he frequency range. If you take the output from the input to the feedback ci= rcuit, don't you get a velocity response without using an integrator? http:/= /www.geophys.uni-stuttgart.de/seismometry/hbk_html/node24.html

      Hope that this is of some interest=20= / help.

      Regards,

      Chris Chapman

[ Top ] [ Back ] [ Home Page ]