differential notch filter with high out of band impedance

With those impedances, I don't know if you _can_ do this with passive components; certainly this will be a challenge if you want to maintain the > M-ohm input impedance on the amplifiers.

Why don't you want to do this with active amplifiers? Because of the ratio of signal strengths? The circuit that jumps into my head is to preamplify both signals as much as possible within the limits of linearity, and applying the result to a twin-T active notch filter. Then gain as necessary to pull your signal at f2 out.

No matter what you do, you're trying for a huge dynamic range; this is going to make distinguishing the signal at f2 from all the hash generated by rejecting the signal at f1 exceedingly difficult. At those relative signal levels you are going to be fighting not only the nonlinearities of your active components, but you are going to be learning fascinating things about nonlinear effects in capacitors and maybe even resistors, as well.

A twin tee can do good service here; the notch depth depends on component matching, so a bit of work with a LCR bridge is usually required. It's possible to use a preamplifier with a twin tee in the feedback loop as a frequency-selective gain element.

The easiest to tune solution would be a switched capacitor filter, but that's not passive.

It's possible with transformers to put the filters in series, so the input loading is the series combination, which might help the loading situation.

signal amplitudes are very small (10 nV), so the amplifier would have to be very low noise. With an amplifier, I'd have to deal with interference, ground loops, noise from the power supply. Furthermore, I am interested if it could be done with a passive circuit, and how one would approach the design.

That's the reason I'd like to keep it simple. I have some active amplifiers with filters, but they create their own problems and are not easy to troubleshoot. I don't need any tuning, I can shift the frequencies a bit.

What about a twin tee filter driven with a voltage follower,and have a voltage follower after the filter? You could drive the common part of the tee with a part of the output follower (use a pot) to adjust how narrow the filter acts.

ve

you mean a voltage follower in front and after of the filter? probably that's what I would have to do, because otherwise the loading of my source (1 k) will be too large, or I'll get too much Johnson noise from the resistors in the tee.

A passive notch filter wont give you gain so your going to need an low noise, high PSRR, isolated supply, pre-amplifier no matter what.

At low frequencies, the inductors will be huge things. I think you are better off to do the filtering with an active filter.

You also may be able to do without any filter at all if you are free to make the lock in amplifier much better. At these frequencies, the sorts of ADCs used for audio can be used. This will get you 24 bits per converter. You can do better if you have multiple ADCs. You can phase shift the signals a little and convert them in a second converter. This tends to blur out any artifacts that the converter may have.