What makes the AVM-2000 nullmeter have such low input noise?

[MegaVolt]

The instruction states that the device has: "Equivalent noise resistance ~ 25Ω for any given filter setting."

The AD8221 input amplifier has a much higher noise floor.

How do they manage to get such low noise?

[Kleinstein]

Thesuggested noise test with  25 ohms resistor at the input is a bit dubious: With added resistance there is also more effect of current noise. Also requiring less than doubling with the resistor suggest this is a test for the input current noise. Having the noise current of a 25 Ohms resistor is more like a bad sign and high noise current.

The schematics does not suggest really low noise: in addition to the Ad8221, there are 2 resistors of 1 K each (likely for over voltage protection). These alone would contribute quite a bit of noise. With the given circuit they can not reach the claimed noise - not even the same ball park. It may need much more filtering (100 seconds time constant). 

The 2  resistor should result in some 600 pV_RMS (some 3.6 nV_PP) alone for the suggested bandwidth of some 0.06 Hz.

Maybe there is addition preamplifier for the nV ranges, not shown.

[guenthert]

  I think this was discussed before.  No one from AVM spoke up (and we didn't inquire there), but the AVM2000 uses relays in the input section (and some users complain about them being audible), presumably in a low-frequency discrete chopper manner with heavy signal processing in the following digital section.  This can dramatically narrow the noise bandwidth and with that effective noise.

[Kleinstein]

The relays used for "chopping" only tends to add to the noise: they can't switch very fast so that 1/f noise from the amplifier can come in, and even if switched only every few seconds there would be some power loss and possible thermal EMF.  Digital signal processing can not do more than filtering (or cheating, e.g. by some sticky action).

[MegaVolt]

They are talking about 400 pV.

[Kleinstein]

Yes they claim 400 pV_pp, but the circuit is more like some 12 nV_pp for the 0.06 Hz BW which seems to be standard  für the +-100 nV range. This also makes about sense for the analog movement to have some 6% FS für the PP noise in the lowest range.
Not sure how they got the 400 pV number - maybe from some 25-30 Ohms "needed" to double the noise.

[MegaVolt]

This also makes about sense for the analog movement to have some 6% FS für the PP noise in the lowest range.

Why is this necessary? Why is noise needed?

[Kleinstein]

One does not need noise, but it does not make much sense to add more gain if the noise is already at 6%. If the noise would really be at 400 pV the analog meter would like more gain, like a 10 nV range.

[MegaVolt]

One does not need noise, but it does not make much sense to add more gain if the noise is already at 6%. If the noise would really be at 400 pV the analog meter would like more gain, like a 10 nV range.

What's the magic number 6%? Where did it come from? Is there a story?

[Kleinstein]

6% is just the 12 nV / 200 nV ( for +-100nV). There is nothing special other wise, this only about the noise level were one can see it with a mechanical movement.