Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Hspice Convergency Issue

Status
Not open for further replies.

Willt

Member level 5
Joined
Mar 17, 2006
Messages
83
Helped
5
Reputation
10
Reaction score
5
Trophy points
1,288
Location
Hong Kong
Activity points
2,077
Dear friends,

I am now studying convergency issue in Hspice.
As all of you know, convergence is defined as the ability to obtain a solution
from a set of circuit equations within a given tolerance criteria.

Could anyone explain the above mentioned definition using a simple example?
Your help is highly appreciated!
Thank you :D

Will

P.S. I would love to click "help me" for your good explanation.
 

Hi
read any numerical computing books.

see known algorithms such as Newton-Raphson or Gause-Seidel for better realization.
Methodhttp://www.shodor.org/UNChem/math/newton/index.html

regards
 

Hi!
The non-linear characteristics of some components will lead to the convergency problems. If you want to get a solution in a shorter time, you will use the component model showing a smooth gradients.

Added after 6 minutes:

One more thing to say is the circuit matrix is to be pivoted so as to reduce the fluctuation.
 

Most of the tolls use iterative methods such as newton raphson method to arrive at solution.

For Newton Raphson methods to converge it requires to meet certain conditions.

Most of the convergence problems are seen for dc and transient analysis

regards
Sarfraz
 

hr_rezaee said:
Hi
read any numerical computing books.

see known algorithms such as Newton-Raphson or Gause-Seidel for better realization.
Methodhttp://www.shodor.org/UNChem/math/newton/index.html

regards
 

read numerical computing book by
gerald wheatley it is extremly amazing Book.
 

i agree with the people who say you should study some numerical methods concerning this ... newton raphson methods, traipezoidal methods are probably the ones mostly used in simulators. read about their deficiencies.

a simple example will be that you use a numerical method eg simpson's 1/3 rule to solve an integration problem, you can get the error to be 5% from the real value (that's when you solve the integration completely without numerical methods). if you use a more accurate method like the traipezoidal rule, your error will reduce to maybe 1% from the actual number. so while simpson's rule might be faster, it might not necessarily be more accurate. it all depends on how accurate you want the result.
 

normally,hspice Convergency problem has close relatoinship with
the time step ,because numerical computing is commonly need reasonable step.

if there is breakup in the circuit , or short of some parameters of the device, it will also meet convergence issue.
 

initial value is another factor
 

Some device model have issues too
 

some netlist create convergency isuue
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top