Poor linearity of a opamp push-pull buffered

[Kleinstein]

The circuit with diode switching and separate transistors does not work well. The amplifiers following the 1 Ohms path would add to the input noise and bias. So low leakage switching with relays makes sense.

[Marco]

When the 1 Ohm leg is not being driven D2 is reverse biased by a couple mV, it's not going to leak much current. Noisy or otherwise.

When it is being driven the opamp determines the noise.

[magic]

The amplifiers following the 1 Ohms path would add to the input noise and bias.

[Marco]

Oh, the amplifier to buffer the outputs ... yeah I guess, throw another couple fA from some LMC662s on the pile. Don't think it matters much though. It will stay in the same order of magnitude.

[Vovk_Z]

I would better use bipolar transistors for 100 mA or even two powerful (50-70 mA) opamps in parallel (one dual opamp as a buffer).
I used mosfets too (for 15 A output current) but added 3.3-3.6 V bias voltage to each with zeners (and resistors of cause).

[MrYakimovYA]

Hi!

Excuse me for upping an old topic. But I have a question: what about an additional noise that brings a FET or BJT output stage? How to calculate it? Maybe there are some guide lines in such configuration of TIA? I'm developing a wide range picoammeter (20 nA ... 20 mA) with 4.5 digit resolution. And I fear that the upper range (20 mA) can't be done without efforts. Should I make the upper range some like 10 mA and use pure op amp (LMC6482AIMFX) output (without busting the op amp)? I have no experience in the field (especially in noise calculation). I'm learning. But there so many questions and traps in the picoammeter/electrometer field...

[magic]

Noise of an output buffer is divided by open loop gain of the controlling opamp at given frequency, it shouldn't be an issue at DC to a few Hz.

Primary noise sources are the Johnson noise of the feedback resistor Rf and voltage noise of the opamp. Both appear like a random voltage in series with Rf and generate a spurious Rf current equal to Vnoise/Rf.

Low frequency noise of your opamp is probably a few µV p-p, noise of the resistor will only matter at high resistance, but then it contributes little noise current because of dividing by Rf.

IIRC noise been discussed at length here:
https://eevblog.com/forum/projects/picoammeter-design/

[Marco]

The big limitation for a multirange TIA will be the capacitance as the Analog link shows, though by keeping the unused paths biased with a buffer that could be reduced IMO (which the Analog link doesn't show).

You'll probably want to buffer the input voltage any way so you can use a more heavy duty input protection than the opamp input diodes, something which can divert enough current for long enough to blow a fuse. Though just sacrificing the circuit if you make a mistake is always an option too, the components aren't that expensive.

[MrYakimovYA]

Noise of an output buffer is divided by open loop gain of the controlling opamp at given frequency, it shouldn't be an issue at DC to a few Hz.

Thank you! It sounds very good to me

Primary noise sources are the Johnson noise of the feedback resistor Rf and voltage noise of the opamp. Both appear like a random voltage in series with Rf and generate a spurious Rf current equal to Vnoise/Rf.

Yes. There is some compromise between low and big value of Rf. In the one case we have lower voltage noise. In the other case we have lower current noise.

Low frequency noise of your opamp is probably a few µV p-p, noise of the resistor will only matter at high resistance, but then it contributes little noise current because of dividing by Rf.

Ok. I simulate my circuit in Micro Cap. It calculates the noise density for me because it's difficult to me to do it manually.

[MrYakimovYA]

The big limitation for a multirange TIA will be the capacitance as the Analog link shows, though by keeping the unused paths biased with a buffer that could be reduced IMO (which the Analog link doesn't show).

I rely mostly on "The Art of Electronics: The X Chapters" book. There is a whole paragraph about TIA. I also like a nice (a fine!) circuit of Keithley 480 and Keitley 485 picoammeters. The first one is manual range swithing. The second one is autorange device. I read both circuits carefully. There are still some strange and weird things to me. For example it's not clear to me the input protection circuit in K485 on every range.

You'll probably want to buffer the input voltage any way so you can use a more heavy duty input protection than the opamp input diodes, something which can divert enough current for long enough to blow a fuse. Though just sacrificing the circuit if you make a mistake is always an option too, the components aren't that expensive.

I want an output power stage in case of input current of 10 mA and higher. In this case op amp output must give exactly the same current to Rf (to have a virtual null in summing point). I believe the op amp will be hot and its parasitic parameters (input bias current, input offset voltage and so on) will drift. So, during a measurement I will have an error. That's why I want to extract the hot stage from precision op amp.