Low-pA TIA Design and Noise Floor

Question

I am currently an EE/Physics undergrad doing scanning tunneling microscopy research at my local university. For future experiments, it would be helpful to employ a higher gain preamp than we currently have, and given that other groups have built them, I was looking into doing the same.

We're hoping to measure current signals from about 0.05-10pA from DC to a few kHz (though more bandwidth with a low noise floor is certainly welcome), requiring a transimpedance gain on the order of 0.1-1 teraohm. This certainly seems possible as it was done at the University of Florence (circuit below), as well as in a number of other papers. A few other topologies are presented here.

^{simulate this circuit – Schematic created using CircuitLab}

I was able to perform a noise analysis of the circuit in LTspice, but I'm a little uncertain of how to translate the spectrum of uV/rtHz vs frequency into a average value for noise floor as it seems to blow up at low frequency with 1/f noise.

Additionally, the LTC6268 seems to have a very low input bias and reasonable noise characteristics, so I was considering using it in place of the above for the first stage.

Answer 1

Looking at the low frequency voltage noise (0.1 to 10 Hz) from the AD549J and the LTC6268 I would say the AD549 is superior at 4 uVp-p. The LTC6268 has a noise of 13 uVp-p. So this is something to be wary of if you are considering a change.

For the input current noise at low frequencies, nothing appears to be stated that is much below 10 MHz for the LTC6268 and that would be a concern. The AD549 is 0.7 fA RMS from 0.1 to 10 Hz by comparison.

I think the LTC6268 is better suited for high speed photodiode applications and not low frequency applications like ion beam measurement.

Answer 2

1/f Noise is a real problem in low frequency applications, at around 10Hz it increases until around 0.01Hz or 0.001Hz and accounts for the majority of noise in amplifiers and electronics.

This is tough because it's in your frequency of interest and it can't be filtered out.

If 1/f noise is a problem, then use a chopping amplifier. However, they may not have the open loop gain characteristics and input bias current characteristics you need. I don't think any from the major manufacturers have an input bias current characteristic that meets your low requirements.

However, There is a company that makes high gain low input impedance low noise amplifiers called Nanovolt EM Amplifiers with no 1/f noise and TΩ input impedance. But they are expensive.