Your input pins are swapped They are ok...

Hi I've used a testbench form the repo of @Weston Braun https://github.com/westonb/open-pmic and it worked for me so You can give it a try. In fact the design was silicon proven and worked.

Pretty sure inp and inm are swapped? If inp is > inm the current through [M1, M3, M5, M7, M8] will all be less, which will pull the output down.

So, the feedback and vin signal must be in the other input?

Your inp and inm pins are named incorrectly. They need to be flipped.

So, are you saying that order is this?

You have a pretty complicated feedback network. Just try with a gain of 1 non inverting amplifier or something.

This is my testbench of open-loop gain.

@wisla morais before doing a .ac simply run a .op simulation, that gives the dc voltage of all nodes. Check if all node voltages are reasonable. In DC the circuit act as a unity gain buffer so

out

should be close to

vcm

. From the images the input pins of the OTA are swapped. I blieve this is fixed now according to above comments. I suspect that 0.9V headroom (vdd -vcm) for

01v8

pfets is not enough. try to replace the input pair pfets with

01v8_lvt

and/or lower vcm value.

Yes, the inputs were switched. And now the dc level of the inputs and output is about 0.9V.

Try to replace 01v8 pfets of the input pair of the ota with 01v8_lvt and make them somewhat bigger (higher W). That will improve the gain.

Okay @Stefan Schippers. Thank you for the suggestion.

The OTA is biased at low current, 1.5uA, The 10MEG feedback resistors is not negligible if compared to OTA output resistance. This will reduce the overall gain.

Thank you very much for your help, suggestions and for clarifying some doubts.

@wisla morais see the problem. With 0.9V on input stage the differential is not working The sources of the input 01v8 transistors are at 1.798V, so the current source transistor is in linear region.. See the 1.5uA current is not mirrored into the differential amplifier, it is working with 37nA

I really appreciate your help @Stefan Schippers. I will consider all the improvements suggested by you in my project. I sized the transistors not to operate in the weak inversion, because I saw that PMOS transistor from the open-PDK behave strangely in this inversion region. Thinking about it, the sizing was not very good, the input pair was small. I will improve this with your tips. Thank you.

Yes, you should avoid weak inversion; the current models are bad in that region. We are waiting for a fix.

Right @Boris Murmann. Thank you for confirm for me.

This topology should use pseudo resistors. They have resistances close to teraohms for a small voltage swing (< 400 mV). See the article from Reid Harrison. https://ieeexplore.ieee.org/abstract/document/1201998

Yes @Luis Henrique Rodovalho thank you. I will study more about pseudo resistors. I'm trying to use them in my simulation. But I have a problem, because I'm using single-supply (1.8V) so I must consider the common mode voltage (VCM) and in the articles I saw the authors used split-supply.

Try this. For your case, if you use pseudo resistors, a resistor ladder is enough, but you should use a resistor ladder plus an opamp in buffer configuration. https://www.ti.com/sc/docs/msp/papers/1998/gustaf.pdf

So, do I need to use an buffer to isolate the non-inverting input and avoid loading it?

Yep. It is not really needed for this configuration with pseudo-resistors, because the current that passes through it would be in the order of pA. Even for your case, with 10 MegOhm resistors, a resistor ladder alone could be enough. Anyway, if you use anything like a non-inverting amplifier, or a differential voltage amplifier, you will need a buffer to make this virtual ground.

Right. I appreciate your help @Luis Henrique Rodovalho, thank you.