From: Dave NelsonTo: psnlist Sent: Thu, 27 Dec 2012 17:35 Subject: Re: Instrumentation Question I have been working extensively on period extension circuits for 4.5Hz geop= hones extending the period to 0.5 Hz.=20 Hi Dave,=20 ****The French School Seismometers use 4.5 Hz geophones extended to 20 seco= nds period. One of the commercial Lennartz seismometers uses 2 Hz geophones extended to= 20 seconds period. I used the Roberts' circuit to extend the period of my 380 Ohm 4.5 Hz geoph= ones to 0.5 Hz - about five~six years ago !=20 =20 The so called Lippman circuit is just a well known negative impedance conve= rter circuit (NIC) applied to geophones. With the original Lippman circuit= the output of the NIC is proportional to acceleration and subsequent circu= its shape the spectrum to give the desired velocity response with the 2 slo= pe roll off below the long period corner. I have taken it a step further a= nd modified the NIC to provide a velocity response at the output of the NIC= .. The result is that there is no point in the active signal path where the = signal is proportional to acceleration. The advantage is a significant imp= rovement in clipping margin for a strong local event and better DC stabilit= y. =20 =20 ****The Lippmann circuit is a current to voltage converter with a negative = impedance input. The output voltage is proportional to f, so it is not too = difficult to add 1/f compensation. Brett has created a spice model which has been very helpful in optimizing = the selection of the negative impedance load on the geophone. I am confiden= t the same circuit could be used to extend the period of a 1 second geophon= e to ~ 20 seconds. I was able to use the circuit equally well for 4.5 Hz ge= ophones with both 380 and 4000 ohm coils. For the 1 second geophone a hig= h resistance coil would be necessary to avoid impractical component values. =20 The DC gain of the NIC will become very high (potentially unstable) if you= attempt to match the coil resistance with a negative resistance resulting = in a near zero net resistance. There is an optimum negative resistance whic= h provides good DC stability and the desired frequency response.=20 ****It can do, but it is easy to be a "bit more clever". The resistance of = the copper coil is highly temperature sensitive, but fairly linear, so you = make the -ve input impedance ~equally sensitive using Pt resistance foil se= nsors stuck onto the geophone. Not difficult to do or too expensive ! The l= ower the resultant total resistance the higher the damping and the less the= coil movement.=20 =20 The major disadvantage of all period extension methods is long period noise= .. A look at the Lennartz noise curves illustrates that very well. It is a u= navoidable consequence of the technique, however, for some applications lik= e volcano monitoring it may not be important. Low noise op amps are essenti= al as you point out. The force feedback technique is dramatically lower noi= se so is the preferred method for periods longer than a few seconds. =20 ****You may need to be careful which Lennartz curves you reference. Lennart= z use both the Roberts and the Lippmann circuits for various sensors. For a period extension of x10, the Roberts circuit requires an ADDITIONAL g= ain of x100 at the low frequency corner.=20 For a period extension of x10, the Lippmann circuit ONLY requires an ADDITI= ONAL gain of x10 ! But, since the current output is so small, the overall a= mplifier gain needs to be much larger than for a geophone and a voltage inp= ut.=20 The problem with using the Roberts circuit is that to get a gain of x100 at= the low frequency corner, the DC gain needs to be about x500. This can gen= erate a lot of very low frequency noise. You can't use a two pole bass boos= t filter for period extension, since you then get a null at the roll over f= requency. You have to use two single pole filters in series. To control the= VLF noise adequately, you may need to use either a two or a four pole high= pass filter. I have been doing this work for a friend for a commercial application but= the circuits are available. We can put them on Brett's website since I do = not have a website.=20 Regards,=20 =20 Chris Chapman =20 =20 =20 From: Dave Nelson <davef= nelson@.......>
To: psnlist <psnlist@..............>
Sent: Thu, 27 Dec 2012 17:35
Subject: Re: Instrumentation Question
I have been working extensively on=20 period extension circuits for 4.5Hz geophones extending the period to 0.5 H= z.=20<= font face=3D"Arial" size=3D"2">
Hi Dave,
****The French School Seismometers use 4.5 Hz geophones ex= tended to 20 seconds period.
One of the commercial Lennartz seismometers uses 2 Hz geophones extended to 20 seconds period.
I used the Roberts' circuit to extend the period = of my 380 Ohm 4.5 Hz geophones to 0.5 Hz - about five~six<= /font> years ago !
The so called Lippman circuit is just = a well known=20 negative impedance converter circuit (NIC) applied to geophones.=20 With the original Lippman circ= uit the=20 output of the NIC is proportional to acceleration and subsequent circuits s= hape=20 the spectrum to give the desired velocity response with the 2 slope roll=20 off below the long period corner. I have taken it a step further and modified the NIC to provid= e a=20 velocity response at the output of the NIC. The resul= t is that there=20 is no point in the active signal path where the signal is proportional = ; to=20 acceleration. The advantage is a significant improvement in clipping=20 margin for a strong local event and better DC stability.
****The Lippmann circuit is a current to voltage converter with a negative impedance input. Th= e output voltage is proportional to f, so it is not too di= fficult to add 1/f compensation.
Brett has created a sp= ice model=20 which has been very helpful in optimizing the selection of the negative=20 impedance load on the geophone. I am confident the same circuit could be us= ed to=20 extend the period of a 1 second geophone to ~ 20 seconds. I was able to use= the=20 circuit equally well for 4.5 Hz geophones with both 380 an= d=20 4000 ohm coils. For the 1 second geophone a high resistance coil would=20 be necessary to avoid impractical component values.The DC gain of the NIC will beco= me very high=20 (potentially unstable) if you attempt to match the coil resistance with a= =20 negative resistance resulting in a near zero net resistance. There is= =20 an optimum negative resistance which provides good DC stability and=20 the desired frequency response.
****It can do, but it is easy to be a" bit more clever". The resistance of the copper coil i= s highly temperature sensitive, but fairly linear, so you make the -ve input impedance ~equally sensitive usi= ng Pt resistance foil sensors stuck onto the geophone. Not difficult to do = or too expensive ! The lower the resultant total resistanc= e the higher the damping and the less the coil movement.
The major disadvantage of all period e= xtension=20 methods is long period noise. A look at the Lennartz noise curves illustrat= es=20 that very well. It is a unavoidable consequence of the technique, however, = for=20 some applications like volcano monitoring it may not be important. Low=20 noise op amps are essential as you point out. The force feedback technique= =20 is dramatically lower noise so is the preferred method for periods longer t= han a=20 few seconds.
****You may need to be careful which = Lennartz curves you reference. Lennartz use both the Roberts = and the Lippmann circuits for various= sensors.
For a period extension of x10, the Roberts circuit require= s an ADDITIONAL gain of x100 at the low frequency corner. =
For a period extension of x10, the Lip= pmann circuit ONLY requires an ADDITIONAL gain of x= 10 ! But, since the current output is so small, the overall amplifier gain = needs to be much larger than for a geophone and a voltage input.
The problem with using the Roberts circuit is that to get a gain of x100 at the low frequency corner, the DC gain needs to be about x500. This can generate a lot of very low fre= quency noise. You can't use a two pole bass boost filter for period extensi= on, since you then get a null at the roll over frequency. = You have to use two single pole filters in series. To cont= rol the VLF noise adequately, you may need to use either a two or a four po= le high pass filter.
I have been doing this work for a friend for=20 a commercial application but the circuits a= re=20 available. We can put them on Brett's website since I do not have a=20 website.
Regards,Chris Chapman