[volt-nuts] How are femto-amps measured?

John Devereux john at devereux.me.uk
Sat Jun 6 08:08:34 EDT 2015


Jan Fredriksson <jan at 41hz.com> writes:

> As described, the pico-femto amp meters are "pass-through", current
> goes in AND out, at 10uV burden voltage. They are explicitly not shunt
> and as far as I understand, can't be trans-impedance amplifiers
> either.
>
> The very lowest leakage OP-amp I know of, LMP7721 is rated 3 fA typ,
> with a specified limit of ±20 fA at room temperature, ie much higher
> than the sub-femto.
>
> The best candidate discreet DIFETI know of, LS320, has a specified
> input leakage resistance of 100Gohm which would be equivalent to a
> gate current of 10fA at 10uV, ie numbers in the right ballpark. Still
> that does not explain the current pass through.
>
> So still it makes me wonder how they do it. Even the old HP 4140B
> (model fell out of my last post) has a specified resolution to one fA.
>
> The only possibility I can think of is to use selected FETs and let
> the test current flow through D-S and control the gate voltage to the
> specified D-S voltage drop and correlate/calibrate test current to
> gate voltage.
>
> I have a Keithley 6485 and guess that out of pure curiosity I'll have
> to make a tear down at some point :-)
>
> Any thoughts?

The 6517A electrometer just uses a commodity opamp, LPC661 I
think. Selected, presumably. Most mosfet input opamps are extremely high
impedance but would take a very long time to test on a production line
so are not specified to their true potential. The LMP7721 you mention is
just better tested I expect.

In use the bias current can be measured and compensated for, leaving the
current noise which is much smaller (though still objectionable and is
what limits the measurement precision and time).

The 6517A has a resolution of 10aA IIRC but it is very noisy and drifty
on the last two digits unless you use a lot of averaging and long
measurement intervals.

Anecdotally a 2n7000 gate leaks "electrons per second". 

<http://lu7abf.org.ar/news/read.php?server=freenews.netfront.net&group=sci.electronics.design&artnum=77452>

It would be fun to try this (configure as follower and leave the gate
open circuit after charging it to a suitable bias voltage).


-- 

John Devereux


More information about the volt-nuts mailing list