Setup to measure input-referred current noise of a differential input output amplifier

The datasheet reports the noise, why don't use that ?

Also, be more clear on what amplifier you intend to use - part #

Hhm, figure 18 of datasheet of THS4120 seems to be wrong, as the front page says that the THS4120 is fabricated in CMOS?

Vahid, what frequency range are you interested in?

By the way, you have seen that the THS4120 provides 1fA / SQRT(Hz), haven't you? 1 femto is 1 x 10^-15. This is 0.001p or 0.000 001n or 0.000 000 001µ. Are you really sure you want to measure this by yourself?

Kai

Well these are both kind of archaic parts, there are better new ones available. 

Anyway, the THS4120 is indeed a CMOS device with essentially unmeasurable input current noise. 

In Any actual input noise extraction, all you can measure is the total output noise which will have the resistor noise terms and the amplifier terms all combined. So the job is to set up different external conditions that emphasize different terms, make the measurement and calibrate out the measurement path noise, and then eventually do a matrix solution for the input referred terms. You will not be able to measure the THS4120 input current noise. By the time you get the R values large enough to emphasize it, they dominate on noise and/or you have bandlimited the current noise gain or put the part into oscillations. And, that current noise term to the output is never dominant so ignore it. 

The THS4505 is Bipolar (Bicom2) - you can always tell these things by the supply range. That one had a poor extraction in the datasheet, that rolloff at higher F is just the rolloff in the gains for the input noise terms, not the noise itself. 

As you set up to measure diff to single output noise at the FDA outputs, keep in mind that channel should have adequate BW to show a flat response well above the expected 1/f corners of the DUT. 

Oh, and I was looking at the THS4505 curve, it is kind of unlikely those corners are below 100Hz, probably scaling error. One of the better noise extractions in this epoch of parts (early 2000 period in Dallas) was probably the THS4131 as it was focused on low noise. Oh, I now see the x-axis is in kHz, not Hz, so probably ok. 

Hi Vahid, 

If I understand your question correctly, you're looking for the hardware setup/circuit configuration used to measure input-referred current noise on an FDA. 

Well, the short answer is that we use a spectrum analyzer to measure output noise of an FDA over frequency, either differentially or single-endedly. The output noise value at every frequency is calculated using a noise model which takes into account all voltage, current and resistive noise sources, and calculating their individual contributions at the given frequency. For reference, please see section 10.1.1 of THS4551 data sheet; it defines output voltage noise of an FDA in terms of input voltage and current noise and resistive noise of the feedback resistors. 

Then, if you're working with a TIA, you would calculate input-referred current noise values across frequency by scaling down output voltage noise across frequency by the transimpedance gain of the amplifier. 

I hope this helps to broadly answer your question. Please let me know if you have further questions. 

Regards,
Vladimir