A nanovoltme...nanovolt amplifier

[sahko123]

So somehow I managed to decide to make a nanovolt amplifier around the same time everyone is making nanovolt meters instead. (thought the nanovoltmeter thing came about in September so quite a while ago actually)

I'm basing my amplifier off of the ADI application note linked here (https://www.analog.com/media/en/analog-dialogue/volume-49/number-2/articles/low-noise-inamp-nanovolt-sensitivity.pdf). As part of it there is a circuit which connects 4 of the AD8428 which is my weapon of choice in this case. The circuit shows the inputs of the in-amps all hooked to each other to reduce noise but for some reason the outputs aren't connected together. Obviously there is a reason to not connect the outputs as well so I'm here to ask for that reason.

What would the effect be if the outputs were also connected be it through resistors or directly to each other?

Also hypothetically if I were to add a capacitor between FIL- and FIL+ while all the in-amps are tied together to achieve the same frequency response would I need to quadruple the capacitance or keep it the same?

I will also post updates on design/constructions in this thread.

[Kleinstein]

The parallel connection can be done at the outputs or at the filter connections. It should also be OK to combine the ouputs and thus use 4 of the output stages. With high gain the output part contributes very little to the overall noise. It should make very little difference whether the filter parts are connected or not in addition to the outputs.

With 4 chips on parallel with the filters one would have resistors for the signal in parallel and would thus also need 4 times the capacitance to get the same cross over.

[sahko123]

Actually going over the datasheet it may not be as good an idea to tie the outputs together because they have different gains so that could cause currents between the outputs

[Kleinstein]

One would not directly tie the outputs togehter, but use series resistors (e.g. 100 Ohm-10 K) to get some averaging. So also combining the ouputs also comes with an disadvantage.

[KE5FX]

The idea is probably that most of the noise comes from the first stage.  So if the filter nodes are tied together, it's reasonable to take the output from only one of the chips. 

It makes sense, even if that's not how it's normally done.

[guenthert]

Low noise is nice and all, but I thought the bandwidth of 3.5MHz would be just so-so.  I misread the specifications thought, it's not 3.5MHz GBW product, it's the bandwidth:

"The AD8428 is one of the fastest instrumentation amplifiers
available. The circuit architecture is designed for high bandwidth
at high gain. The AD8428 uses a current feedback topology for
the initial preamplifier gain stage of 200, followed by a difference
amplifier stage of 10. This architecture results in a 3.5 MHz
bandwidth at a gain of 2000 for an equivalent gain bandwidth
product of 7 GHz. "

Holy smoke.