Stats

  • Locked Locked
  • Replies 2
  • Subscribers 125
  • Views 2168
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

How does the correlation constraint "cc" affect the correlation of mismatch parameters

Matthew CG

Hello,

I am using the correlation constraint to tie together Monte-Carlo mismatch parameters for a number of instances of an ADC.

Example syntax 

statistics {
correlate dev=[ I0.* I1.* ] cc=1.00
}

When cc=1 I can see that the parameters are all identical across instances.

When cc=0 the parameters are uncorrelated and I have full Monte-Carlo dispersion has per the model distributions.

Are values between 0 and 1 valid?

I would like to reduce the amount of mismatch between certain instances with the idea being that my blocks aren't too far away from each other so perhaps my mismatch should not be maximum.

Is this a valid approach? I haven't yet found any documentation that describes the values for "cc" and in my testing it's not obvious to me how cc affects the correlation for values between 0 and 1.

Thanks for your help.

Matthew Cordrey-Gale

  • Andrew Beckett

    Hi Matthew,

    A quick answer. First of all, you might want to take a look at this older reply from me (there are a few posts on this, but this was one I found which summarised it fairly well). Put simply the "mismatch" naming we use might better be described as "local" variation, and so it's describing the residual local variation. Usually foundries have not characterised how that variation changes with proximity, alignment, presence of dummy devices and so on (although we do have some "litho" tools that aim to do this). So setting the correlation coefficient (which absolutely can be any value between 0 and 1) is rather hard (and possibly pointless, depending on how the models have been characterised). Usually the models are characterised assuming good layout practice, but asking the foundry would be the best solution here - as for setting the correlation coefficient, that's very hard to know what is reasonable unless you have measured data to help determine the correlation with various layout structures. The meaning of correlation coefficient is.

    In practice I rarely see people trying to set the correlation coefficient between devices - largely because the models from foundries don't really support it in a meaningful way.

    Andrew

  • Matthew CG
    in reply to Andrew Beckett

    Andrew,

    That confirms some of the thoughts we had. Like you say we should perhaps bring up the subject with our foundry and see what they think.

    Many thanks for the quick reply and useful advice.

    Matthew

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.