Component selection for special purpose amplifier

[Greg-ulate]

Hello all,

I have an amplifier in mind, you could think of it as an ac coupled differential amplifier, where the coupling capacitance is extremely small (<1pF) and you can assume the input is a sinusoidal differential voltage function. The amplitude of the source is over 1 Million volts.

The conclusion that I've come to is that I want to measure the charge that moves back and forth from one capacitive input to the other over each cycle, which is something like

[tex]10nC \star Sin[\omega t] \hbox{ where } \omega \hbox{ is about 240Hz}[/tex]

The most important goals are long term stability, temperature stability, common mode rejection, and low noise. In my latest prototype the amplifier has two stages. A dual-supply op-amp integrator on each input to give a voltage signal proportional to the charge, and then the output goes into a fully differential amplifier and then to the ADC. The signal is already relatively large compared to the noise, but I am attempting to achieve better than .01% accuracy. There is also a badly distorted 60Hz component in the noise, which i am trying to combat by adding more CMRR and symmetrical inputs.

I would like to produce a newer version of this with a single supply, and a completely differential signal chain.

From reading a few other posts on the forums, I have learned a little bit more about the parameters of amplifiers, but let me ask a few questions to be sure I have it right.

At first I thought that a very high impedance instrumentation amplifier would be the best choice, since I am measuring a high impedance source. I was looking at the INA116, with an input bias current of and considering all sorts of guard ring and cable shielding possibilities. Alternatively, I have looked at the AD8231 zero-drift programmable instrumentation amplifier. I like the AD8231 better because its got better drift characteristics and CMRR.

Each input of the instrumentation amplifier would be biased through a 10-100Mohm resistor to a mid-supply temp stabilized voltage reference. The inputs would have a capacitor between them to produce a voltage of about +- 30mV. The amplifier would be configured for a gain of 32. After the instrumentation amplifier everything will be low impedance and relatively easy to manage. With respect to the first input stage, is there anything that will end up biting me later if I go with the AD8231 vs. any other arrangement?

While examining this problem I have noticed certain things about the different breeds of amplifier. Considering op-amps, fully differential amplifiers, and instrumentation amplifiers, can anyone elaborate on qualities that are favored for each type? For example, the input bias current on most fully differential amplifiers is enormous compared to many op-amps and in-amps, like 3 microamps compared to 20 picoamps.

The purpose of this project is to regulate the voltage of a electrostatic particle accelerator.

 

[Floid]

So, your overall goal is this:

You want a circuit that can be used to remove noise from a 1MV sinusoidal signal. The output of the circuit should be what voltage range? (from the gain 32 statement you want a 32MV sinusoidal signal?).

If I understand what you are trying to do correctly, the biggest problem you are going to run into is that ICs can't handle voltages anywhere close to the 1MV range. For example, the AD8231 is only takes inputs up to +/-2.5V. The INA116 is much better, but you are still limited to +/-18V.

 

[jim hardy]

The 1pf is what couples the high volts into your measureing circuit?

in electric power they use capacitive coupling to measure transmission line voltages in hundreds of KV range so that approach is not unheard of.

search on "capcitive coupled potential device"

http://books.google.com/books?id=Su...q=capacitive coupled potential device&f=false

a million volts across one picofarad at 240 hz is on order of 1.5 milliamps?
With that much current why do you need high impedance? Couldn't the current just go into summing junction of LM324 with 1k feedback to give a nice 1.5 volt signal?

what did i miss? That's an honest question.

but 0.01% accuracy? How will you get such precise capacitor in picofarad range?



If it's just current you must measure , an audio output transformer might do a nice job of isolating your 1.5ma and providing a ground-referenced signal .

A sketch would help.

 

[Greg-ulate]

Sorry for the confusion. Please see the attached circuit diagram sketch. Although the symbol for the AD8231 is slightly incorrect, the idea should be clear. The high voltage forces a charge of about +10 nano Coulomb into the + input, and -10 nano Coulomb into the - input at the peak of the cycle. The presence of this charge develops a voltage across capacitor C. I will choose C so that I get about +- 30milliVolts (not Mega) across the capacitor, and then amplify that by 32 so that I get about +-1V out of the amplifier.

Yes, I believe that the current is just shy of 1mA rms, Jim. The question you ask is pretty much the one I am trying to answer myself. I am becoming more and more convinced that high input impedance is not critical, but with a 1k feedback resistor the device would become a high pass filter with a cutoff frequency at 160Mhz... (which does nicely attenuate the MegaVolt signal to a nice 1 Volt) This is similar to my very first attempt to perform this measurement and I am more comfortable considering a Capacitive Divider and measuring the charge rather than the current. The signal is not a pure sine wave, there are other harmonics present and they would be attenuated differently.

The 1pF capacitance is not a component of the amplifier but it is part of the detector, called a Generating Volt Meter (GVM). It is basically a parallel plate capacitor with an area of a few square inches and a distance of over a meter! it might be closer to .1pF. Attached is an illustration of the GVM, which shows that its not an AC voltage coupled by a fixed capacitance. Instead it is a fixed voltage and the capacitance is varied by blocking the electric field with a spinning grounded rotor.

I think I like the AD8231 the best because at a gain of 32, the Temperature drift is almost non-existant and the CMRR is very high. Additionally, internally selected gain means one less external component that I have to worry about drifting with temperature.
The two C/10 capacitors are shown because I have seen it suggested by instrumentation amplifier manufacturers. I might make them C/100 or omit them so their impact is lessened. The last order of business is to find a capacitor for C that is the most temperature stable one I can find.

 

[jim hardy]

That's an interesting project.

Hobbyists use a similar gizmo to measure Earth's electric field, it's on my "Do List"
here's old Scientific American article
i have seen more detailed article someplace, will try to find it and see what electronics they used. He just says "transconductance amplifier".

I improvised a field mill by taking two steel cake pans and cutting out a dozen equally spaced 15-degree wedges (the slots) from their circular bases. To rotate one of the pans, I used a surplus high-speed electric motor. These motors typically deliver between 1,000 and 7,000 revolutions per minute. At those rates and with the cake pans cut with 15-degree wedges, the Earth's field generates nanoampere-size current surges in the ground wire at frequencies between 200 hertz (for 1,000 rpm) and 1,400 hertz (for 7,000 rpm). Such a signal can be observed easily with a circuit containing a transconductance amplifier and a peak detector. In fact, my homemade instrument can readily detect shifts a mere thousandth of the ambient field. Furthermore, a computer analyzing the data will be able to follow the fluctuations with a performance rivaling that of professional instruments.

http://www.scientificamerican.com/article.cfm?id=detecting-the-Earth's-elec

I guess you'd have to convert currents from both sides of source into voltages and take difference ? Balancing the converters might be touchy...

but it looks to me like imbalances those coupling picofarads (area of plates?) plates could make common mode voltage at your measuring capacitor.

Might you replace your C with a small transformer and use secondary single-ended?
Then you'd have your choice, place C across secondary and Earth one end so as to measure its voltage more easily, or let secondary drive a current to voltage op-amp.

To get a stable capacitor you might use a combination of polypropylene (negative temperature coefficient) and polycarbonate or polystyrene (both positive).. see page 15 here, top left graph..
http://www.vishay.com/docs/26033/gentecin.pdf


I've never used that generating voltmeter - thanks for the introduction !

old jim

 

[Greg-ulate]

Oh man that is so cool! exactly the same principle. I bet it the E field goes crazy during a thunderstorm. I wish mine went at 1400 hertz because then my software could calculate the voltage faster.

Thanks for the interesting ideas. I would really like to be able to measure the current at the same time, because with the added information of the current I could have just a little bit more information with which to do my analysis. I calculate that the rms current should be about 15uA, which is not a terribly difficult current to measure I guess, but adding that info to the analysis is a bit out of the scope of this endeavor.

I will try the two different types of capacitor. This has been a major concern.