[volt-nuts] How to keep voltage stable in the sub-100nV range?

[David]

Read the various datasheets and application notes for chopper
stabilized amplifiers, the highest precision bipolar operational
amplifiers, and low input bias current operational amplifiers.  Linear
Technology and Analog Devices are great sources for this information.
The major issues in rough order of importance are:

1. Thermocouples and temperature gradients - this is a huge problem
and special attention will need to be directed toward the layout and
maintaining an isothermal environment.  Careful design is required to
get the specified drift performance out of chopper stabilized (10nV/C)
and low drift operational amplifiers (100nV/C before trimming or
grading).

2. Leakage - even at low impedance levels, minor amounts of leakage
will cause significant errors.  100nV at 10k is only 10pA so this is a
real problem for RC filters where long time constants require high
impedance levels and capacitor leakage also needs to be considered;
avoid Mylar/polyester/PET and high dielectric constant ceramic
capacitors.  In the past, through-hole construction allowed "air
wiring" sensitive nodes but that is not practical with surface mount
parts.  Guards will need to be used in the layout which is impossible
with the smaller surface mount packages.  

3. Resistor self heating - even precision resistors become a problem
at this level of stability because of self heating which also
complicates the thermocouple problem.  High resistance values lower
self heating but make errors from leakage current and current noise
worse.

4. Dielectric absorption - this will be a problem if fast settling is
required.  Avoid Mylar/polyester/PET and high dielectric constant
ceramic capacitors.

5. Pink Noise - 1/f noise increases as frequency decreases.  Chopper
amplifiers have flat 1/f noise so are invaluable below about 1 Hz.

On Tue, 1 Nov 2016 13:14:16 +0100, you wrote:

>Hi,
>
>I have a "small" side project, which involves keeping a voltage stable
>to better than 100nV over the period of several seconds. Ie. a DAC
>produces an output and a chain of opamps and low pass filters feeds
>it to the consuming device. The absolute value and drift over more
>than 10-100s is not that important.
>
>As I lack a lot of knowledge in this field, I would like to ask
>whether someone can point me at literature or give me some terms
>to search for that help me to figure out whether this is actually
>feasible and how I could achieve that. I know the basic literature
>on noise and how to deal with that. What I am interested in are the
>real world problems, how big they actually are and how to deal with them.
>
>			Attila Kinali