Stats

  • Locked Locked
  • Replies 2
  • Subscribers 65
  • Views 8492
  • 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

CDF callback discrepancies between design variables and hard-coded parameter values

VincentCheung

Hello,

please see attached images (pic 1.png, pic 2.png, pic 3.png, pic 4.png):

  • pic 1.png shows the simple testbench setup.
    • I am comparing IDS_N1 with IDS_N2.
    • N1 has its width parameter assigned with a design variable, Wn.
    • N2 has its width parameter assigned with a hardcoded value that is equal to whatever Wn is set to.
    • I run a simple transient simulation (or DC, doesn't matter) and do the comparison stated above.
      • Example: set Wn=120nm, then hardcode N2 to have width = 120nm. Compare IDS_N1 with IDS_N2.
      • Ideally, IDS_N1 = IDS_N2.
  • pic 2.png shows the results from the testbench above.
    • In the table: IDS_N1 = Ids_sat_n in the "Equations (Design Variable)"
    • In the table: IDS_N2 = Ids_sat_n in the "PCELL (Hard-coded)"
    • For Wn <= 145nm: there are significant discrepancies between IDS_N1 & IDS_N2.
      • In the table: as, ad, ps, pd, nrd, nrs are CDF parameters for the NFET device - there are significant discrepancies between design variable vs hard-coded.
    • For Wn >= 150nm: there are 0 discrepancies between IDS_N1 & IDS_N2.
      • In the table: as, ad, ps, pd, nrd, nrs are CDF parameters for the NFET device - there are no discrepancies between design variable vs hard-coded.
  • pic 3.png shows the edit cell properties for the NFET with W=Wn.
    • You can see that the 6 CDF parameters boxed in red all have equations that are dependent on Wn (you don't actually see Wn in the screenshot because the textbox cuts it off).
  • pic 4.png shows the CDF editor for the NFET cell:
    • You can see that the 6 CDF parameters boxed in red all have a CDF SKILL callback function: tsmcN65lo_nch_macCB().
    • I believe this CDF SKILL callback function is the root cause and main culprit of this issue. 

Question: Does anyone have any how to debug and resolve this issue? My best guess is that the issue lies within the SKILL code for tsmcN65lo_nch_macCB() but I don't know where to locate this code? 

Pic 1

Pic 2

Pic 3

Pic 4

Thanks!

Best regards, Vincent

  • Dear Vincent,

    you could try pretty printing the SKILL code by executing pp(tsmcN65lo_nch_macCB) in the CIW.
    If that works you can investigate the code and look for the root cause.
    I suppose that will not work, because foundries typically encrypt their CDF callbacks.

    You could also try to compare the netlists (Design Variable vs. Hard-coded) and look for discrepancies.

    Anyhow, I suggest that you contact customer support of TSMC.

    Kind regards

    Matthias

  • in reply to mschw

    Thanks Matthias! The pretty print definitely confirmed that tsmcN65lo_nch_macCB is definitely encrypted. 

    I will take a look at the netlist and do a comparison. Thanks again!

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.