Repository navigation
8.3.693
commit 8.3.693
"Tagging version 8.3.693"
See documentation at https://github.com/RTimothyEdwards/magic/blob/9fd4ee57/appimage/7/README.md
See documentation at https://github.com/RTimothyEdwards/magic/blob/9fd4ee57/appimage/8/README.md
See documentation at https://github.com/RTimothyEdwards/magic/blob/9fd4ee57/appimage/9/README.md
See documentation at https://github.com/RTimothyEdwards/magic/blob/9fd4ee57/appimage/10/README.md
Change Log (since previous tag):
9fd4ee57 Reworking a change made back on or around August 1. After reports about incorrect sidewall parasitics, I (with much help from Claude Opus 5.5) revisited the solution to the problem of why magic was over-calculating the capacitance of the interdigitated capacitors in IHP sg13cmos5l (specifically, my cap_cmomf layout). After an investigation, we resolved that (1) I had made a calculation error in my early analysis, (2) Magic had been correctly handling the sidewall coefficient being twice the value for a single edge (but was calculating it in a place I wasn't looking), (3) The real reason that the interdigitated caps were wrong was that FasterCap needed a much finer mesh for sidewall calculations than was being set by my "capiche" analysis tool, and that successive refinement brought the value down by almost exactly the amount of the difference between magic's (former) output and IHP's simulations using OpenEMS and AWS Palace. There are now multiple steps in fixing the problem: (A) Update magic (which I'm doing now), (B) Update the sidewall coefficients in all of the magic techfiles for IHP and GF180MCU, which used "capiche" to compute values, and (C) implement "cap_cmomf" and "cap_cmomi" devices properly in IHP SG13CMOS5L, such that the extraction uses IHP's model and not magic's extracted value (which should now be close, but never as good as a proper model). Then (D) magic and the open PDKs need to be updated in "ciel" and in IIC-OSIC-TOOLS.