With a signal of up to 20 Hz and a 10 times higher amplitude noise of 50 Hz, it sounds like a difficult task.

I was wondering about that too. If the 50Hz noise only needs to be blocked in order to protect the rest of the system, that's one thing, but if you need to filter it in such a way that you can still detect the smaller signal, that would be a daunting task!

Honestly, the more I think about this problem, the less certain I am that I actually understand the user requirements. In addition to my earlier question about AC and negative voltages, I'm still not clear on whether or not analog voltage measurements are critical here - is the signal just a pulse train? Do we only need to capture the frequency of the signal, or an analog voltage measurement, both at the same time, or maybe a complete capture of the full waveform?

If we're just protecting an input, some of those details may not matter, but if we want to read the signal through the noise, we'll need to understand the critical parameters of the signal much better!

 

A different version of the input clamping scheme could be to use germanium diodes, such as the 1N34, which have a much lower forward voltage drop.. And I am wondering if your data acquisition input is subject to damage from a small over-voltage, or just incorrect operation? Does it have any internal over voltage protection, or not? Most of the systems that I have used did have some protection, but they also needed current limiting on the inputs for over voltage protection. So I am wondering.

 

The datasheet does not say +/- 0.3 Volt absolute voltage, it says -0.3 Volts below AVss and 0.3 Volt above AVdd.

With AVss = 0 Volt and AVdd = 5 Volt, the analog inputs must be within -0.3 Volts and +5.3 Volts.

And the analog inputs can withstand up to 10 mA current at over voltage at the input.

 

The datasheet does not say +/- 0.3 Volt absolute voltage, it says -0.3 Volts below AVss and 0.3 Volt above AVdd.

With AVss = 0 Volt and AVdd = 5 Volt, the analog inputs must be within -0.3 Volts and +5.3 Volts.

And the analog inputs can withstand up to 10 mA current at over voltage at the input.

GOOD NEWS INDEED!! Now we know what the input device is and we see the specifications and it becomes clear that the more challenging part will be preventing any negative input voltage. But that clamping is quite possible by using a normal fast diode with the cathode biased up so that it will conduct as soon as the input starts to go negative. The downside of this is a few more parts and a slightly higher power consumption. But there would be no additional active components to affect the gain, offset, noise, or linearity. So the approach is worth considering.
So many thanks to the TS, UltraV, the TS, for the clarification of the actual situation.