smb@MIMSY.UMD.EDU (Steve M. Burinsky) (09/11/87)
If I'm not mistaken, the way MOSIS derives Spice parameters for a given run is to adjust various model parameters to fit parametric curves from devices on the test structures. Thus, the parameters they distribute may have very little in common with those of the actual devices on a given run -- they just happen to produce similar parametric curves.
daasch@psueea.UUCP (Robert Daasch) (09/14/87)
In article <8709102152.AA05876@mimsy.umd.edu> smb@MIMSY.UMD.EDU (St >given run is to adjust various model parameters to fit parametric >curves from devices on the test structures. Thus, the parameters >they distribute may have very little in common with those of the >actual devices on a given run -- they just happen to produce similar >parametric curves. Agreed with a but... It mystifies me how any parametric fit to test structures would produce counter intuitive results. Generally it is intuition from the physics of the device that bounds the fitting procedure. So when I see the p device with a VT0 > 0 and junction depths several orders of magnitude out of whack, I don't use that set of device parameters. Maybe the fitting utility doesn't write out a model deck and MOSIS people play rock, paper and scissors for "who TYPES in the models this week." Rob D. ..!tektronix!psueea!daasch.uucp P.S. He says sheepishly, a couple runs we did do weren't completely bogus.