- Author : Christopher Torng
- Date : June 7, 2019
This is a 45nm ASIC design kit for mflowgen, a modular ASIC/FPGA flow generator:
This kit uses FreePDK45 and the NanGate Open Cell Library.
- FreePDK45: https://www.eda.ncsu.edu/wiki/FreePDK45:Contents
- NanGate: https://projects.si2.org/openeda.si2.org/help/group_ld.php?group=63}
The FreePDK45 kit is an open-source generic process design kit (PDK) (i.e., does not correspond to any real process and cannot be fabricated) that allows researchers and students to experiment with designing in a modern technology node without signing restrictive non-disclosure agreements or paying for licenses. The PDK allows you to use commercial full-custom layout tools (e.g., Cadence Virtuoso) to design both analog and digital circuits.
The Nangate Open Cell Library is a generic open-source digital standard-cell library designed using the FreePDK45 kit.
This package was assembled from these kits to be used as an ASIC design kit for designs built with mflowgen. See more information about mflowgen at its github repo.
Changes to make Calibre LVS pass more robustly:
- calibre-lvs.rule
We added case sensitivity to the commands:
SOURCE CASE YES
LAYOUT CASE YES
The tools sometimes create wires like "n52" and "N52" in the same module, and they are meant to be separate nets. If case sensitivity options are set to "NO", this turns into multiple drivers from Calibre's viewpoint.
This change reduces the number of unexpected LVS issues for most designs.
Changes to allow a small combination design (GcdUnit) to pass Calibre LVS:
- adk.tcl
- rtk-stream-out.map
- stdcells-bc.lib
- stdcells-wc.lib
- stdcells.cdl
- stdcells.gds
- stdcells.lef
- stdcells.lib
- stdcells.v
New WELLTAP_X1 stdcell added. The rtk-stream-out.map now has layer numbers that match the Calibre DRC/LVS rule decks and stdcells.gds layer numbers.
Change:
- rtk-tech.lef: One-line change to set MANUFACTURINGGRID to 0.0025 instead of 0.0050
OpenRAM is an open-source memory compiler:
To interact nicely with OpenRAM, we modified the routing technology LEF (i.e., rtk-tech.lef) such that "MANUFACTURINGGRID" is "0.0025" instead of "0.0050" (a one line change). Otherwise Cadence Innovus complains that the SRAM LEF has pins that are not on the manufacturing grid.
Since this is a fake technology, we think this is okay.
Actually, the DRC rule deck also says that the grid should be 2.5nm:
Grid.1 {
@All shapes must be on a 2.5 nm grid
OFFGRID active 5 5
}
Change:
- calibre-lvs.rule: One-line change to set precision to 10000 instead of 2000
In order to run LVS, the precisions have to match between the LVS rule file precision and the GDS database precision. Otherwise Calibre says this:
ERROR: Rule file precision 2000 is not consistent with database
precision 10000 in input file design_merged.gds
The stdcells.gds has a precision of 10000. For example, here is a message printed from calibredrv when reading the gds for the minimum size inverter:
% calibredrv -a layout filemerge -in stdcells.gds -topcell INV_X1
-out dummy.gds
Output file precision will be 10000, based on the lowest common
multiple of the input file precision(s).
To run LVS successfully, our solution is just to make the LVS rule deck expect a precision of 10000.
Initial commit
Base set of files from FreePDK45 and NanGate Open Cell Library assembled into an ASIC design kit (ADK) for use with mflowgen