Electronics Forum / Design / July 2008

Sounds pretty simple: a good opamp and an analog switch to flip the
gain from +1 to -1, then a lowpass filter. At 10s of Hz, charge
injection shouldn't be a big deal. The resistors will need to be very
good to hold 1 ppm gain. VERY good.

1 ppm of 10 mV is of course 10 nV, so I hope you can amplify ahead of
the synchronous detector stage.

John

Do you really mean stability?
Stability over what?

or resolution?

or accuracy?

Mark

On Jul 24, 9:26 am, Spehro Pefhany <speffS...@interlogDOTyou.knowwhat>
wrote:

TO GET THIS STABILITY    YOU NEED  A TEMP CONTROLED  FLUKE
METER AND A LAB AT 68 DEGREES F.       GOOD LUCK.

On Jul 24, 12:26 pm, Spehro Pefhany
<speffS...@interlogDOTyou.knowwhat> wrote:

Firstly, your input (preamp, wideband amp...) has to be nondistorting
at the low frequency (which will cause thermal fluctuations that
generally
CANNOT be ignored).  A transconductance amp (output is a current
source into pseudo-ground) is preferable here, as the output bias
currents of a voltage-output op amp can be troublesome.

For a uA741 op amp, chip temperature is a bigger noise source than
input offset voltage at about 4 kOhm input resistance... the
change in temperature modulates the input currents that much,
for low frequency signals.

Secondly, you have to use a mixer for your filter, and for the kind
of accuracy you want, the dynamic range of an ADC is not enough.
Mix down before converting to digital data, then use digital
signal processing to deal with rejecting out-of-band signal.
Transconductance type mixers should work out well, even
an MC1496 can do this.

Your reference sinewave has to  be as accurate as possible, of
course.  I'm not sure how one makes a high-purity 10 Hz sinewave,
but triangle and square wave generators are easy enough.  The
feedback scheme on a Wien bridge oscillator gets tricky at
low frequency (maybe you can use sample/hold for determining
the amplitude, but it takes a good phase determination to
pick the gate closing time).