voltage source with gaussian distribution.

[yvk]

Hi,

Can we create a source(v or i) with gaussian distrib (and able to set its mean, std,correlation etc) in the schematic.
this would be handy to create macromodel for actual circuit which only can work while running monte carlo analysis.(analogous to using gm,r,c elements for creating smallsignal ciruit)

Thanks,
yvk

 

[Andrew Beckett]

On 09/20/12 09:57, yvk wrote:

Hi,

Can we create a source(v or i) with gaussian distrib (and able to
set
its mean, std,correlation etc) in the schematic.
this would be handy to create macromodel for actual circuit which
only
can work while running monte carlo analysis.(analogous to using gm,r,c

elements for creating smallsignal ciruit)

Thanks, yvk

You'll need to create an include file to describe the distribution.

Place an instance of the voltage or current source, and give the value
as a variable name e.g. "MYVOLTVAL". In ADE specify a value for
MYVOLTVAL (you can do Variables->Copy from Schematic to populate the
variables in ADE). This will be your mean value.

Then create a file, say called "statistics.scs" and put:

// the distributions I want
statistics {
process {
vary MYVOLTVAL dist=gauss std=10.0
}
}

You could potentially add a mismatch section too (see "spectre -h
montecarlo"), but you'd have to ensure that the voltage source is inside
a subckt, and then the mismatch would be between different instances of
the subckt (i.e. put it in a schematic, create a symbol, and then
instantiate that symbol).

You could also potentially make the std in the above statement a
parameter value and then add that as a design variable in ADE so you
could set both mean and std from ADE. The one thing that ADE doesn't let
you do is define the statistics block itself interactively.

Having created statistics.scs (make sure it has a .scs suffix so it's
treated as spectre syntax), use Setup->Model Libraries, and reference
the file.

Regards,

Andrew

 

[yvk]

Thanks for the solution Andrew.

On Wednesday, September 26, 2012 9:27:33 PM UTC+5:30, Andrew Beckett wrote:

On 09/20/12 09:57, yvk wrote:

Hi,



Can we create a source(v or i) with gaussian distrib (and able to

set

its mean, std,correlation etc) in the schematic.

this would be handy to create macromodel for actual circuit which

only

can work while running monte carlo analysis.(analogous to using gm,r,c

elements for creating smallsignal ciruit)



Thanks, yvk



You'll need to create an include file to describe the distribution.



Place an instance of the voltage or current source, and give the value

as a variable name e.g. "MYVOLTVAL". In ADE specify a value for

MYVOLTVAL (you can do Variables->Copy from Schematic to populate the

variables in ADE). This will be your mean value.



Then create a file, say called "statistics.scs" and put:



// the distributions I want

statistics {

process {

vary MYVOLTVAL dist=gauss std=10.0

}

}



You could potentially add a mismatch section too (see "spectre -h

montecarlo"), but you'd have to ensure that the voltage source is inside

a subckt, and then the mismatch would be between different instances of

the subckt (i.e. put it in a schematic, create a symbol, and then

instantiate that symbol).



You could also potentially make the std in the above statement a

parameter value and then add that as a design variable in ADE so you

could set both mean and std from ADE. The one thing that ADE doesn't let

you do is define the statistics block itself interactively.



Having created statistics.scs (make sure it has a .scs suffix so it's

treated as spectre syntax), use Setup->Model Libraries, and reference

the file.



Regards,



Andrew