questions on the resistor for simulation

[John Xu]

Hi,
I have two questions on the rsistor variation for simulation.

First question:
If in my circuit, I used two resistors, the resistors are same type, e.g., all are nwell resistor. Then in full case simulations, can I assume they vary in the same trend? i.e., if one is maximum, another is maximum. It will not happen that one is min, another is max? Is it because that they have the same mask?

Second question:
If I use two different type resistors, e.g.,one is nwell resistor, another is P+ resistor,Then in full case simulations, need I combine them with different independent corners, i.e., one min, another is max, or one is max, another is min?It means we need more simulations to find the worst condition.

Can anyone pls. help on this?

Thanks in advance

 

[feier0251]

you should do the mentcarlo simulate.

 

[John Xu]

Montcaro analysis is useful for the assessment of performance for the resistor variation.
But, here, I hope to find the worst case condition. So I need combine different corners to find it. What I want to know is how to deal with the different resistors used in one circuit, should I consider them as different corners?

 

[Fom]

You have answered all your questions by yourself.
For the same type of resistor use the same corner. For different type of resistor- different corner, because they were done with different implantation on different time. So they should be considered as independent and can deviate in different direction.

 

[limingchou]

when simulation
u should simulate all corner of resistors. and u have a percent by the model parameter

 

[sunking]

careful layout.
and you can simulate in the same corner when change the ratio a little(0.2% to 0.5%)

 

[John Xu]

Hello, Sunking,
Thanks for the help.

You said "simulate in the same corner when change the ratio a little(0.2% to 0.5%) " , ir that mean when simulation, also considering the mismatch for a rsistor ratio up to 0.2% to 0.5%?

thanks

 

[huojinsi]

According to ur post,it is very difficult to simulate!

 

[Alan_Nesta]

I think you can use the res model in all corners, and better do post-simulation to it.

 

[neoflash]

Hi,
I have two questions on the rsistor variation for simulation.

First question:
If in my circuit, I used two resistors, the resistors are same type, e.g., all are nwell resistor. Then in full case simulations, can I assume they vary in the same trend? i.e., if one is maximum, another is maximum. It will not happen that one is min, another is max? Is it because that they have the same mask?

Second question:
If I use two different type resistors, e.g.,one is nwell resistor, another is P+ resistor,Then in full case simulations, need I combine them with different independent corners, i.e., one min, another is max, or one is max, another is min?It means we need more simulations to find the worst condition.

Can anyone pls. help on this?

Thanks in advance

I think process variation can be categoried into two types:
1. PVT variation
2. random mismatch

1st type can be viewed as group property of same lot (die). 2nd type can be viewed as individual behavior, as an add-on over group behavior.

Same resistor, same orientation, same geometry will give you very good matching on group behavior. However, they are not going to help you on indivudual behavior.

You need to run process corner simulation to cover group behavior, however, you also need to run monte-carlo simulation to cover individual behavior.

You might be a student in China, good luck if your teacher is really teaching you something.