Circuits to detect low voltage level signals

Measuring picovolts at DC is extremely hard at room temperature. Just about everything has thermoelectric potentials of nanovolts to about a millivolt per degree C.

It can be done if you convert it to AC using some modulation scheme, amplify there, and convert back to DC, but it does not happen by accident.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

Agilent and Keithley make nanovoltmeters, and they have manuals and probably appnotes online. Just this morning I designed a cable to connect one of our gadgets to an Agilent 34420A. The input connector, a custom Agilent thing, costs $112.

Picovolt measurements aren't easy or even practical in most situations. Dissimilar metals, or even the same metals with impurity differences, can generate microvolts or nanovolts of signal from small temperature gradients. Just moving wires in the earth's magnetic field can pin a nanovoltmeter.

Why would you want to measure picovolts?

John

Phil,

Thank you so much for your response! I would wonder if there would be some sort of monograph on these types of measurements. It seems tricky.

Oh, and BTW, I can't wait until your new book is released. I'm going to pick up a copy as soon as it becomes available here in Canada.

Nicholas

It's next to impossible. Measuring picovolts near DC requires a SQUID.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

I'm going to have to take a look at the Agilent website, and do some exploring on the Keithley website. BTW, it seems to be a pricey connector.

I'm interested in this type of thing from a research perspective, and stumbling across the idea of measuring picovolts led to me to consider the possibility of being able to better measure outputs from a transducer. Right now I am thinking of using a 24-bit ADC, and I thought about the possibility of measuring more than just microvolts.

Anyway, it makes some interesting discussion, and helps to feed my addiction for electronics--what else a better way to enjoy my time on a newsgroup?

Thanks, John.

I vaguely remember something about superconductivity in the application of a SQUID. So this is the reason why working with an AC signal is more desirable?

10...

John said, "Just moving wires in the earth's magnetic field can pin a nanovoltmeter."

Hmm, I'm looking at the noise voltage from a 100k ohm resistor down the bottom of a 6" brass tube. (one end of the resistor is soldered to the bottom of the tube and the other to the one side of a shielded twisted pair.) There is then 18" of cable connecting the probe to a metal preamp box. This works great unless you 'bang' on the tube. Banging on the tube with hard objects, (such as ice cubes in an ice bath.) causes spikes in the noise voltage from the resistor at acoustic frequencies. (2.5kHz and higher.) The spikes can increase the noise measured in a 100kHz bandwidth by a factor of 2 or more.

If this is caused by motion of the conductors in the earth=92s B field can I reduce it by orienting the wires along the field?

Perhaps I should mention that the other half of the twisted pair and the shield around the twisted pair is being driven by the output of the first opamp. A driven shield.

(Friday thoughts after a few beers)

George H.

Could also be microphonics (in any insulators, in the resistor, in the geometry itself). For instance, if the inside wire has a DC potential, its capacitance and therefore voltage varies as it's moved around inside the shield.

You can try a mu-metal tube instead, see if that has any effect. In the mean time, it might be worthwhile adding some acoustic dampening inside the tube to at least reduce the resonance, if not necessarily its amplitude.

Tim

-- Deep Friar: a very philosophical monk. Website:

Hmm, I'm looking at the noise voltage from a 100k ohm resistor down the bottom of a 6" brass tube. (one end of the resistor is soldered to the bottom of the tube and the other to the one side of a shielded twisted pair.) There is then 18" of cable connecting the probe to a metal preamp box. This works great unless you 'bang' on the tube. Banging on the tube with hard objects, (such as ice cubes in an ice bath.) causes spikes in the noise voltage from the resistor at acoustic frequencies. (2.5kHz and higher.) The spikes can increase the noise measured in a 100kHz bandwidth by a factor of 2 or more.

If this is caused by motion of the conductors in the earth?s B field can I reduce it by orienting the wires along the field?

Perhaps I should mention that the other half of the twisted pair and the shield around the twisted pair is being driven by the output of the first opamp. A driven shield.

(Friday thoughts after a few beers)

George H.

The wiring (if there's a local B field) can induce voltage from motion, but there are lots of other suspects; if that 100k ohm resistor is a carbon composition item, piezoconductivity is likely as well. If it were a wirewound resistor, it'd be piezoresistivity to worry about (probably a smaller effect). In insulators, one finds triboelectric effects from bending of the insulator, and the list goes on.

To some, the word 'electronics' means a kind of engineering with off-the-shelf components. Those people don't measure picovolts. To others, 'electronics' means everything that electrons, positrons, even ions and other charged particles get up to, in solids, gases, etc. Those aren't engineers, they're physicists. And they don't generally know much about picovolts, either.

Some of the cheap chopper opamps have drifts in the nv/K range. It wouldn't be hard to temperature-regulate one of them, or several working in parallel. A cmos analog switch could zero-check them. Then the problem becomes noise and getting the signals into the amps without thermal and magnetics problems.

Some of the cmos chopamps that I'm using have rather large offsets - microvolts - that seem to depend on the capacitance seen by the input pins. Some charge injection effect. I had to add a really tiny (0805 size) trimpot to one plug-in 6-pole active filter board to normalize dc offsets to below a uV.

Picovolt signals, like picosecond signals, are inherently low impedance.

John

Yes. Techs and scientists are often both bad circuit designers; one trusts too little in theory, and the other too much. The horrors I've seen!

We need a web site that posts the schematics of world-class-bad commercial products. badschematics.com and badschematics.org are both available.

John

n

ld

,

=3D10...

If you hit a metal object, you will inevitably displace electrons, hence generating a voltage between points that were at the same potental. There doesn't seem to be anything when Googling "piezo electric effect in metals", strangely.

: > : > Nicholas

: It's next to impossible. Measuring picovolts near DC requires a SQUID.

: Cheers

: Phil Hobbs

SQUID based picovoltmeters indeed exist. These guys make cooled transformers for pV measurements IIRC:

Regards, Mikko

Sorry, the resistor is a 1/4 watt metal film, Xicon. This isn't really much of an issue. I was actually surprised that I could ground the resistor to the outside of the 'can' and still measure its Johnson noise. There's a bit of a spike when I place it in an ice bath, and then the noise drops to a lower value. Actually a more pertinent question, for me would be; what=92s the thermal conductivity of a 1/4 watt metal film resistor. One side of the resistor is at zero C and the other side is connected to Room Temp via a few inches of #24 copper wire. I measure noise temperature of about 10 C. (It=92s summer and RT is about 26 C.)

George H.

Yup, I'm more physicist than EE. Wiggling the cable doesn=92t have much effect. It=92s tapping on the brass tube that makes the noise voltage wiggle. I see spikes at audio frequencies as measured by an SRS

760? (100kHz spectrum analyzer.) I could imagine wire=92s moving in the earth=92s field.

George H.

In which direction?

John

I often wondered how to get rid of the large offsets. I suppose that calibration is essential in this type of precision stuff.

Thanks, John.

I guess that measuring picovolts is something in-between engineering or physics. Perhaps it could be considered something of an "art"?

As an environmental physicist and an engineer, I second this recommendation.