Process corner of Resistor and capacitor

[prasanta_g]

Hi,

I am working in TSMC 65n technology.
Can any body help me about how to choose the process variation of resistor and capacitor with NMOS and PMOS process variation?
For example, let I am using rppoly_m resistor and crt_mom(because my load is crt_mom) capacitor for miller compensation. Driver of the second stage is NMOS.
To compensate 2nd pole Rz=(1+CL/Cc)/gm
Now Cl/Cc is constant for all process and voltage.
So Rz should vary with gm.
Now Rz and gm is two independent process parameter.
That is why I am asking for some help about the choosing of process corner of resistor and MOS.

 

[ZekeR]

prasanta_g,

You can make g_m vary with R using a standard constant-g_m biasing circuit. This will help to account for variations in R by making g_m vary inversely with R.

However, this won't compensate for variations in the MOSFET (such as t_ox and doping variations).

You can scheme up a way to trim R to account for the MOSFET's variations if it's really so important. However, usually it isn't important enough to do this, so we simply design the circuit with generous margin so that regardless of how much the zero moves around due to MOSFET variation (which might be ±30% or so), the circuit still functions as planned.

 

[prasanta_g]

prasanta_g,

You can make g_m vary with R using a standard constant-g_m biasing circuit. This will help to account for variations in R by making g_m vary inversely with R.

However, this won't compensate for variations in the MOSFET (such as t_ox and doping variations).

You can scheme up a way to trim R to account for the MOSFET's variations if it's really so important. However, usually it isn't important enough to do this, so we simply design the circuit with generous margin so that regardless of how much the zero moves around due to MOSFET variation (which might be ±30% or so), the circuit still functions as planned.

First of all thank you for your suggestion.
But my main query is what will be the process corner(slow or fast) of resistor (rppoly_m) if nmos is slow and pmos is fast? I don not know whether there exist any relation betn. process of resistor and MOS.
Because resistor can vary ±30%(min) and gm can vary ±30% so overall variation is huge.

 

[ZekeR]

prasanta_g,

There should be some amount of poly doping variation that affects both the MOSFETs' threshold voltage (by changing the poly gate's work function) and its effective transconductance (due to poly depletion), and assuming that your resistors are made of poly, this will affect the resistor values too. However, since the poly is probably silicided, the doping variation will contribute very little to the resistors' actual resistance (since its resistance is made much smaller by silicide), meaning that most of the resistors' variation would come from other factors. So, the factor that links MOSFET corners to resistor corners is most likely negligible; for this reason, I think your library's model corners won't show any relation between resistor and MOSFET corners.

To test it for yourself, you can pump a current through a resistor while stepping through MOSFET corners and see if the voltage changes. This would indicate whether the resistance variation with MOSFET corners has been modeled. Then, you can pump a current through a diode-connected MOSFET while stepping through resistor corners. If the voltage changes, it would indicate that that variation has been modeled.

My bet is that you can treat them as independent variations.

 

[prasanta_g]

Thank you.
I did the simulation with various process corner with mos ,resistor and cap and get different results.

For ex. I take a corner -"FSTT"
F-Fast NMOS
S-Slow PMOS
T- Typical Resistor
T-typical Capacitor

and take another corner "FSFT"

I get different AC result for miller compensated circuit.

 

[ZekeR]

I get different AC result for miller compensated circuit.

There are many things going on in a complicated circuit like this, making it difficult to narrow down exactly what corners cause what variations. You should try simulating an extremely simple circuit, such as a single MOSFET or a single resistor, to know with certainty what's happening.