Femtovolt

[glen herrmannsfeldt]

Continuing the previous question on femtovolt signals, consider
starting with 1 microvolt at 100 ohms, and assume that we
can detect that signal.

Now run it through a 1000:1 transformer, down to 1 nanovolt.
Through another 1000:1 transformer, down to 1 picovolt.
Now another, down to 1 femtovolt. To test that one can
detect the signal run it through three 1:1000 transformers back
to the original 1 microvolt 100 ohm signal.

The current in the original signal is 10 nanoamps, and increases
by a factor of 1000 at each step to 10 amps at 1 femtovolt.
(Assuming no transformer loss.) Each transformer winding should
have wire resistance much less than the signal impedance.
The femtovolt winding has 0.1 femtoohm impedance, so there better
be a lot of copper in that wire! One should arrange the transformers
carefully such that there is no coupling other than through the
windings. It would be nice to see the signal change polarity when
the femtovolt winding wires were reversed, but that will require
some careful work. (Switches with less than femtoohm resistance
are hard to find.)

-- glen

[Jerry Avins]

glen herrmannsfeldt wrote:
> Continuing the previous question on femtovolt signals, consider
> starting with 1 microvolt at 100 ohms, and assume that we
> can detect that signal.
>
> Now run it through a 1000:1 transformer, down to 1 nanovolt.
> Through another 1000:1 transformer, down to 1 picovolt.
> Now another, down to 1 femtovolt. To test that one can
> detect the signal run it through three 1:1000 transformers back
> to the original 1 microvolt 100 ohm signal.
>
> The current in the original signal is 10 nanoamps, and increases
> by a factor of 1000 at each step to 10 amps at 1 femtovolt.
> (Assuming no transformer loss.) Each transformer winding should
> have wire resistance much less than the signal impedance.
> The femtovolt winding has 0.1 femtoohm impedance, so there better
> be a lot of copper in that wire! One should arrange the transformers
> carefully such that there is no coupling other than through the
> windings. It would be nice to see the signal change polarity when
> the femtovolt winding wires were reversed, but that will require
> some careful work. (Switches with less than femtoohm resistance
> are hard to find.)

At room temperature, 50 ohms, and 1 KHz bandwidth, noise power is about
-144 dBm. A femtowatt is a millionth of that, -204 dBm. Might as well
try to measure a yoctowatt.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

[Richard Owlett]

glen herrmannsfeldt wrote:

> Continuing the previous question on femtovolt signals, consider
> [snip]

What *should* be considered:
1. what is the physical source to be measured
2. why measure it

Otherwise one can have "egg on face". The original question that
triggered this and other threads did not consider either.

Examples from personal experience.

1. As electronics tech for an Ivy League chemistry department I had a
post-doc ask me to procure an electrometer with a 10 Hz bandwidth
for a system such that it would be measuring 1/10 of an
electron/second.

The problem - he did not understand gain-bandwidth product. The
application was measuring the output of an early solid state electron
multiplier used as the detector on a high resolution mass
spectrometer.

2. [this one I was the silly one]
Just out of high school I was working as a gofer in a TV repair shop
and was being introduced to trouble shooting. The boss caught me
measuring the voltage across a cathode bias resistor on the most
sensitive scale of his new Simpson 260. He asked me to "think THEN
measure".

[Jerry Avins]

Jerry Avins wrote:
> glen herrmannsfeldt wrote:
>> Continuing the previous question on femtovolt signals, consider
>> starting with 1 microvolt at 100 ohms, and assume that we
>> can detect that signal.
>>
>> Now run it through a 1000:1 transformer, down to 1 nanovolt.
>> Through another 1000:1 transformer, down to 1 picovolt.
>> Now another, down to 1 femtovolt. To test that one can
>> detect the signal run it through three 1:1000 transformers back
>> to the original 1 microvolt 100 ohm signal.
>>
>> The current in the original signal is 10 nanoamps, and increases
>> by a factor of 1000 at each step to 10 amps at 1 femtovolt.
>> (Assuming no transformer loss.) Each transformer winding should
>> have wire resistance much less than the signal impedance.
>> The femtovolt winding has 0.1 femtoohm impedance, so there better
>> be a lot of copper in that wire! One should arrange the transformers
>> carefully such that there is no coupling other than through the
>> windings. It would be nice to see the signal change polarity when
>> the femtovolt winding wires were reversed, but that will require
>> some careful work. (Switches with less than femtoohm resistance
>> are hard to find.)
>
> At room temperature, 50 ohms, and 1 KHz bandwidth, noise power is about
> -144 dBm. A femtowatt is a millionth of that, -204 dBm. Might as well
> try to measure a yoctowatt.

Never mind. A femtowatt is -120 dBm.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

[John O'Flaherty]

On Mon, 8 Jun 2009 17:00:45 +0000 (UTC), glen herrmannsfeldt
<[email protected]> wrote:

>Continuing the previous question on femtovolt signals, consider
>starting with 1 microvolt at 100 ohms, and assume that we
>can detect that signal.
>
>Now run it through a 1000:1 transformer, down to 1 nanovolt.
>Through another 1000:1 transformer, down to 1 picovolt.
>Now another, down to 1 femtovolt. To test that one can
>detect the signal run it through three 1:1000 transformers back
>to the original 1 microvolt 100 ohm signal.
>
>The current in the original signal is 10 nanoamps, and increases
>by a factor of 1000 at each step to 10 amps at 1 femtovolt.
>(Assuming no transformer loss.) Each transformer winding should
>have wire resistance much less than the signal impedance.
>The femtovolt winding has 0.1 femtoohm impedance, so there better
>be a lot of copper in that wire! One should arrange the transformers
>carefully such that there is no coupling other than through the
>windings. It would be nice to see the signal change polarity when
>the femtovolt winding wires were reversed, but that will require
>some careful work. (Switches with less than femtoohm resistance
>are hard to find.)

The wire may be a problem. 10 ga Cu wire has 0.669 ohm / 100 m, with a
radius of about .9 mm. To get a factor of 10 lower than 0.1 f-ohm
impedance would take 10 ga wire about 1.5e-12 mm long - rather short
and fat*. Maybe a superconductor?
--
John

* 100m / 0.669ohm * 1e-17ohm * 1e5mm / 100m

[[email protected]]

On Jun 8, 1:00*pm, glen herrmannsfeldt <[email protected]> wrote:
> Continuing the previous question on femtovolt signals, consider
> starting with 1 microvolt at 100 ohms, and assume that we
> can detect that signal.
>
> Now run it through a 1000:1 transformer, down to 1 nanovolt.
> Through another 1000:1 transformer, down to 1 picovolt.
> Now another, down to 1 femtovolt. * To test that one can
> detect the signal run it through three 1:1000 transformers back
> to the original 1 microvolt 100 ohm signal.
>
> The current in the original signal is 10 nanoamps, and increases
> by a factor of 1000 at each step to 10 amps at 1 femtovolt.
> (Assuming no transformer loss.) *Each transformer winding should
> have wire resistance much less than the signal impedance.
> The femtovolt winding has 0.1 femtoohm impedance, so there better
> be a lot of copper in that wire! *One should arrange the transformers
> carefully such that there is no coupling other than through the
> windings. *It would be nice to see the signal change polarity when
> the femtovolt winding wires were reversed, but that will require
> some careful work. *(Switches with less than femtoohm resistance
> are hard to find.)
>
> -- glen

Hello Glen,

This certainly poses many questions. If we have a static voltage, then
we need to balance it and maybe something using a casimir force will
do. Of course the math for the casimir force assumes the plates are
uncharged. But maybe this can be worked out.

On the other hand let's assume the singal is alternating. If it is
oscillating like a sinusoid (regular frequency), then capture the
signal in a high "Q" (low loss) resonator until the collected energy
builds up to a point of being detectable.

The problem with low currents is of course shot noise. The nice field
equations don't show the quantum effects resulting from discrete
charges.

And I think Johnson noise will raise its head in all of this, so
femtovolt work will likely need to be done at very cold temperatures.

This will require some more pondering.

Just a few of my thoughts.

Clay