Home Forums Bug Reports Otimization

Viewing 1 post (of 1 total)
  • Author
    Posts
  • #617
    zhu
    Participant

      Optimization itself works very well owing to a fix in the new release. Nelder-Mead is extremely robust and fast despite that generally it is regarded inferior to gradient type optimizers. The rest of the tools are very slow.

      Highlights:
      – adding an Octave support environment complex parameter extraction/optimization problems can be solved with ease of archiving and plotting.
      – internal variable export from sophiticated Verilog-A device models can be implemented with a few program lines and such “strobe” devices can be placed in the binary library for a convenient use
      – nodes of “voltage” nature in the Verilog-A compiler are highly appreciated
      Lowlights:
      – at RF optimizations using s-parameters output=bias implies halt (bug_DC.bmp).
      – output=dc works but currents can be strobed only by CCVS-s. Currents can be used in plots but not in Eqn-s (bug_DCfix.bmp).
      – at DC optimizations VS drives imply halt. Norton equivalents fix the issue (bug_norton.bmp).
      – in Y=stoy() Y[1,2] and Y[2,1] are interchanged. stoh() looks fine, the rest was not tested. (ypars.png)

      Thank you very much for this “poor men’s” IC-CAP!

      Best regards,
      Zoltan

      Attachments:
    Viewing 1 post (of 1 total)
    • You must be logged in to reply to this topic.