Can a Liberty file have CCS data and not NLDM data?

[janthonym]

I've seen liberty files that just have NLDM (Non-linear delay model) data. For the CCS (Composite current source) liberty files I've seen, it always contains NLDM data.

Can a liberty file contain only CCS data and not NLDM data? If not, what is the reason why CCS liberty files must have NLDM data?

 

[ranaya]

I've seen liberty files that just have NLDM (Non-linear delay model) data. For the CCS (Composite current source) liberty files I've seen, it always contains NLDM data.

Can a liberty file contain only CCS data and not NLDM data? If not, what is the reason why CCS liberty files must have NLDM data?

Hi, for the power/delay calculations you definitely need NLDM lookup tables in your .lib file. However the library does not necessarily need to have CCS noise information of the cells. So during the characterization of cells, either you can exclude CCS Noise/Power information or combine it with NLDM. What you usually see as *_ccs.lib/.db s are the libraries with this additional CCS information.

Anuradha

 

[janthonym]

Hi, for the power/delay calculations you definitely need NLDM lookup tables in your .lib file. However the library does not necessarily need to have CCS noise information of the cells. So during the characterization of cells, either you can exclude CCS Noise/Power information or combine it with NLDM. What you usually see as *_ccs.lib/.db s are the libraries with this additional CCS information.

Anuradha

Thank you for the information Ranaya. Question, why do we need the NLDM data? For example, in a liberty file, why can't we only have CCS delay values and not NLDM delay values? Is that a limitation of the characterization?

 

[ranaya]

Thank you for the information Ranaya. Question, why do we need the NLDM data? For example, in a liberty file, why can't we only have CCS delay values and not NLDM delay values? Is that a limitation of the characterization?

Btw do not misunderstand that NLDM here means look up tables generated from traditional voltage curve based measurements ! Nowadays most industry standard libs use CCS measurements to generate reliable delay and power look up tables in .libs. So what you referred to as NLDM data is this delay/power information of cell arcs measured based on CCS curves. However these lookup tables do not provide any information related to the circuit noise/cross talk related information. For this purpose, CCS modeling has been extended to CCSN which includes additional entries in the .lib to account for noise immunity of the cell. I think that is what you refer to as CCS data in the .lib !