how to estimate Senseamp offset voltage to use montecarlo ?

[daedu]

i'm so confuse about methodology about offset estimation.
Sram must have required offset voltage for sensing right memory data
Please help me Someone who knows about methodology. and how to make offset standard normal distribution

 

[shlooky]

Hi,

If you are referring to sense amplifier in SRAM block used for data reading...

You need to design the transistors with certain W and L, so in Monte-Carlo analysis the input offset voltage of your sense amp will remain under a specified value. Maybe read some material on mismatch, pelgrom's plots and pelgrom's coefficients.
Offset voltage usually has normal distribution. You are basically tuning its mean value and standard deviation by W and L of the transistors.
Sense amps are usually very simple and symmetrical, so it should not be too much work

Good luck with your design

 

[daedu]

Hi,

If you are referring to sense amplifier in SRAM block used for data reading...

You need to design the transistors with certain W and L, so in Monte-Carlo analysis the input offset voltage of your sense amp will remain under a specified value. Maybe read some material on mismatch, pelgrom's plots and pelgrom's coefficients.
Offset voltage usually has normal distribution. You are basically tuning its mean value and standard deviation by W and L of the transistors.
Sense amps are usually very simple and symmetrical, so it should not be too much work

Good luck with your design

Hi,
Thank you for applying my question
I read
https://designers-guide.org/analysis/comparator.pdf
After i read this , I think offset voltage is estimated by inverse function about cumulative standard distribution.
So I was wondering how to generate normal distribution about offset.
It means x-axis is offset, y-axis is probability density.

 

[shlooky]

Hi,

What you get out of Monte-Carlo analysis (out of many other things) is a histogram of a scalar parameter (i.e. input offset voltage) created by a number of simulations of your design with mismatch and random process variations.
The very basic parameters of your histogram are shown in the attached picture. It is an example of Monte-Carlo simulation results. I simulated 500 simulations with three different temperatures and put the numbers for both output edges together = 3000 values of input offset voltage.
The picture tells me that the expected value of my input offset voltage is 2.66 uV, sigma = std. deviation = 686 uV, 3*sigma = 2.05mV. These numbers are directly related to W*L area of your transistors, overdrive voltage level = VDD - VTH in digital circuits (dependent on technology used), drain current and layout quality.

Your PDF shows some Matlab script and the math background of Monte-Carlo analysis. But in circuit design these are not that important. Plus you get your data from the circuit simulator.... such as *SPICE, Virtuoso/Spectre, etc...

Shlooky