( ESNUG 414 Item 10 ) ------------------------------------------- [06/18/03]
From: Tong Xu <xutong=user company=ict.ac.cn>
Subject: PhysOpt User Asks On extract2plib & Astro vs PhysOpt Correlation
Hi, John,
My question is about the extract2plib of PhysOpt. In our original plib
there are 6 routable layers M1-M6 and one device layer POLY1. In our
ITF file, there are 8 conductors: M1-M6, poly and OD (the thin oxide and
interconnect). When using extract2plib to enable 2.5-D extraction in
PhysOpt, it reports that
"There are less number of metal layers in plib file; No of metal
layers in tech file are: 7"
I tried by deleting OD layer in ITF and it did work. Then a Synopsys AE
told me to change OD layer from CONDUCTOR to DIELECTRIC, which also work
but the result was the same as deleting OD.
I have 4 questions:
1. At both successful occasions, in the verbose_file.txt there is no OD
layer info any more and Poly is regarded as field_oxide. Whether the
modifications will make the extraction inaccurate in PhysOpt? If yes,
how to solve it?
2. In the original plib, the metal layer res_per_sq is 0.101, while it is
0.078 in the ITF file, which makes the ultimate plib 0.078. Where can
such difference come from? Is this 20% difference natural?
3. Will RC calculation around large hard macros/RAMs still be inaccurate
as mentioned in ESNUG 399 #8 in PhysOpt ver 2003.03?
4. If the parasitics are not still well correlated between Astro and
PhysOpt, what can I do without rc_estimation?
I appreciate any help on these issues.
- Tong Xu
Institute of Computing Technology Beijing, China
|
|
