OCV mode optimization in encounter

Yes, Encounter does support OCV mode both for timing analysis and for optimization.  To enable it, use the setAnalysisMode command:

setAnalysisMode -analysisType onChipVariation

However, it is crucial to understand that if you set up the tool with max and min libraries via a .conf file and enable onChipVariation timing that the software will use the min library for timing the capture clock paths.  This is probably not what you want (the difference between the max and min .libs is likely quite large which will lead to very large clock skew).  From a timing analysis perspective, this quandry can be resolved by pointing to only the max .libs for setup analysis and only the min libs for hold analysis.  For optimization purposes, this can be resolved by establishing the timing environment using MMMC (mulit-mode/multi-corner) mechanisms- even if you have only a single setup corner and a single hold corner.

This approach is elaborated on in detail in documentation (and a complete flow) in the 8.1 release of Encounter as part of the newly released Foundation Flows.  Log in with your Sourcelink account to access more information here:

http://sourcelink.cadence.com/docs/files/Release_Info/Docs/flatImpl/flatImpl8.1/flow.html

I'm not sure I understand the last piece of your question regarding derating libraries.  I've not come across this before myself.  Maybe someone else can provide additional information on that subject.

Hope this helps,
Bob

 It is rare that we actually get OCV libs, so I think what you are looking for is the set_timing_derate command. For example, if we want 5% OCV, we would use something like this:

set_timing_derate -early 0.95 -late 1.00 -delay_corner ss_125c_cmax_125c
set_timing_derate -early 1.00 -late 1.05 -delay_corner ff_m55c_cmin_m55c

(The -delay_corner option is part of the MMMC setup which Bob has already recommended to you.) Please check the docs for all the available options of set_timing_derate to make sure you are derating exactly the way you want to.

- Kari 

Thx for the reply Bob, Kari

I understand that encounter has MMMC flows mode is supported by derated commands. Going into a more intricate level my timing analyzer works in this way for OCV:

Uses only 1 library and uses early and late derates .( as does encounter) The other thing it does is the following : Lets say we are doing a SETUP analysis with worst case libraries : When ever the tool finds multi input gates it propogates max slews for delay calculation in the launch path and min slews for delay calculation in the capture path , thus adding an extra element of pessimism . Would encounter do the same when the derates have been specified ? Consider that fact that in most designs clock tree is  not comprised of sime buffers or invertors but a lot of combo logic as well hence the difference turns out be significant. 

The goal here is to understand if by some black magic I could create derated libraries and do the setting

setAnalysisMode -analysisType onChipVariation

Would there be better correlation in timing as seen from encounter and seen by my STA tool ?

Since I mention the word correlation ,  I would mention that we are looking into RC correlation by running ostrict hence I would request the discussion to be specific to OCV mode optimization of encounter.

Yes, Encounter timing would reflect differences on launch/capture paths due to min/max slew when in onChipVariation mode.  Other similar effects include:

• Pin cap ranges in .libs
• Crosstalk pull-in on capture clock paths