quick question on monte carlo simulation

[evilguy]

Hi, i'm a bit puzzle in the term of sigma (σ) in the monte carlo simulation. what is σ, 2σ and 3σ represent when we do monte carlo simulation. σ in statiscal term represent standard deviation of random data. Is this term also similar in monte carlo simulation?
Then 2σ = 2*σ? and 3σ = 3*σ?

my other question is, if we know the parameter variation will not exceed 20% from the nominal value, how to present this variation interm of σ? thanks

edit: i think i've found the answer
https://en.wikipedia.org/wiki/Standard_deviation
In monte carlo analysis, how can we model two SPICE parameter so that both will distrubute similarly when we do monte carlo analysis.

-evilguy-

 

[DenisMark]

Yes, sigma is parameter of statistical distribution. 2σ = 2*σ and 3σ = 3*σ. There 99,7% spead in values lies in 3σ. 20% means extreme range of parameter variation. The propability of +20% or -20% occurrence in parameter value usually is less than 0.1%.
How many sigma in 20% u can define as 20%*MeanValue / σ.

 

[evilguy]

thanks. according to the paper i've read. the 20% parameter variation is equal to 3σ value. how i can use this information when i want to do monte carlo simulation. i mean the sintax in hspice.

 

[DenisMark]

20% = 3σ isn't a rule. I cann't help u with hspice, i use spectre. Look in hspice's documentation.
One more problem u don't clearly explain what parameters u are going to change during monte carlo simulation. May be information what u had isn't enough for correct simulation.
Give us more details and somebody helps u with hspice.

 

[evilguy]

sorry about that. actually i want to varies the TOX parameter. according tu paper, the variation of oxide thickness in the range of 10%-20% from the nominal value. So assumption is made the the variation of oxide thinkness follows normal distribution. that is why the term 3σ=20%. means that the most variation to occur is 20%.

However the MOS model that i want to modified is based on BSIM4 model which the oxide thickness is model to TOXE, TOXP, and DTOX which

DTOX=TOXE-TOXP

if i want to model the variation in TOXE to use in monte carlo i need to include the variation+the nominal value so that it can be used in monte carlo simulation. However, i'm not sure how to include the variation in the spice model parameter. Hope my explanation clear something here.

 

[sowmya005]

how is the montecarlo simulation done using the variation block?