AnsweredAssumed Answered

Design Centering and Yield Optimization

Question asked by JERRY_MARKETOS on Nov 8, 2004
The information that follows is largely compiled from within Genesys HELP,
but we receive questions from time to time and so I thought I would post here.

If this post runs past the "size" limit, the remainder will be added as first reply.


One of the powerful features of Genesys (not the core subject of this post)
is optimization.  In simple terms, optimization is accomplished by adding a set
of optimization targets and defining variables for the optimizer to tune.  (The
variables can be direct component values or equation variables in Genesys)
When the optimizer is invoked, it will tune the variables in search of
a set of values that satisfies the defined optimization targets.

Genesys also provides the ability to go a step further.... (the core subject of this post !)

Once a set of optimized component values is obtained, we can ask the next question...

"Do we really have the best set of component values such that,
when those components begin to wander within their specified tolerances in production,
we maintain the best possible yield of target circuit performance ?"

This is where design centering and yield optimizaton come in.

(This part is directly from Genesys HELP files...)
"Design centering is a process which attempts to maximize yield by adjusting the nominal
component values so the objective function errors are equalized." (as the component goes through
its tolerance excursion)  That is, the result of design centering may reveal that a shift in the nominal
value of some components (compared to what the optimized value may have been) produces better
yield as all components vary through their tolerances.

Design centering is (was- read on...) a powerful way to apply modern EDA tools to further optimize
designs for manufacture. 

Things tend to progress as time goes by, and a new approach has been developed for pursuit
of element values that are optimized for mass production.  Instead of working to equalize the objective
function errors, Yield Optimization instead works by keeping track of assortments of element values
that satisfied the yield targets, then taking the average of the different data points for each element value.
Using Yield Optimization brings in a new parameter "depth of search".  A small depth of search such as "1"
confines the search to element values that don't differ too much from the nominal.  Yield optimizations with
small depth of search will run faster, but may not find the best possible optimization of yield.  A larger depth
of search such as "5" will offer more opportunity to find solutions that may offer best yield, but will take
longer to run since they incorporate more statistical "ground to cover".

More information on these topics can be found in Genesys HELP
(Main menu HELP, Contents, switch to the SEARCH tab and search for "design centering" or "yield optimization"
then click "list topics").