Ok, I would to make some precision resistors with Skywater 130. I'm using the resistor component but I think that it is convenient to use mosfet for to do this. I would need 2 values ​​in particular: 10Kohm and 12Kohm. To obtain them via mosfets, which mosfets should I use and how to arrange the parameters (L and W) of the same?

Usually, precision resistors are not possible to make in integrated circuits. The resistor variation is on the order of +/-20%. What is your goal? Usually, we try to make circuits that rely on ratios of resistors as this is relatively stable across process, voltage and temperature

In my case I need a voltage divider. The Ratio is as indicated (10k/12k) and I am keeping the resistor values ​​high to decrease the current as much as possible. I thought it could be done using mosfets. Is this possible?

message has been deleted

In the meantime I had read from various sources that polysilicon resistors are very precise with very low tolerances. This is in contrast to what you declare to me here. What is true in what I have read? EDIT: But your mosfet in the voltage divider are Depletion Mosfet?

The poly resistors have very good (low) temperature coefficients and local mismatch variability as a function of the area, compared to diffusion resistors. For process variability, it has 20% variability. It's not a problem for voltage dividers, as what matters is the ratio between the resistors, not their nominal value. The voltage divider made with PMOS transistors connected as diodes would be made of 11 transistors in series. For the open source techs, they would certainly operate in the weak inversion, with currents in the pA to few nA range. You couldn't have the same currents with integrated resistors.

Thanks for the reply, but I have to say at this point that I am not very knowledgeable. What exactly do you mean that for voltage dividers it is not a problem because "it is the ratio of resistors and not their nominal value"? As far as I know, the ratio of resistors is just the ratio of their nominal value, isn't it? And then, I would be fine with using mosfets instead of resistors, but I imagine that these have a really high resistance (is there a way to quantify it?) and this could be a problem because the downstream circuit must be discharged through the divider and it must do so in a short time.

See, if you have a voltage divider of 10k/12k it does not matter if it is 12k/14.4k or 8k/9.6k. The output voltage will be the same. The current consumption won't. Normally, depending on your application, the output of a voltage divider will be connected to the gate of a transistor. An example is a LDO. If you're designing a DAC, you will connect it to a voltage buffer. The high resistance of a mos voltage divider is a problem for transient signals. It can take milliseconds to reach the steady state. I only recommend using them if you are looking into low power designs.

Ok, but I consider the hypothesis that there could be one resistor with +20% and the other -20% tolerance and in this case the ratio would become 12K/10K, but from what you wrote the tolerance is the same for all resistors, did I understand correctly? In this case it shouldn't be a problem. For mosfets it would be fine but my circuit has to switch in picoseconds (Currently it does it in about 500ps and it's quite good, more than that would be a problem)

Nope. The 20% tolerance for PVT is all transistors at the same time. This is global mismatch. Local mismatch is a function of the resistor area only. You should run a monte carlo sim using minimum size resistors and see the actual mismatch. Then you should increase the area until it meets your specifications. I guess that its mismatch is inversely proportional to the squareroot of the area of resistors, as for transistors. Anyway, you can be sure you can achieve an accurate voltage divider with reasonable area using poly resistors, even without trimming. You just need to run a MC sim first. Maybe someone has already made a resistive DAC and run a MC sim.