|
|
Magic Mailing List |
|
From: Graham Petley (graham.petley AT vlsitechnology DOT org) Date: Sun Oct 12 2003 - 16:03:47 EDT
Hi all, I've come back from vacation, and in between the rugby world matches it's time to look at the DRC errors reported by Charles on September 27. It's my opinion that the vsclib Magic cell has a DRC error. When I examine the .MAG file around the VDD wire, I find that the Magic cifin has split the original VDD metal rectangle into four separate rectangles which leaves a hole right at the location of the genericcontact. If I do a ;cif write, in the output CIF file I see that at this genericcontact location there is no metal1. With the scmos.tech file, I think there is no DRC error because the metal1 width and spacing rules are 2 lambda, the same as the genericcontact cut. But with the SCN5M_DEEP.12 tech file, which requires metal1 width and spacing rules of 3 lambda, a DRC error is found at every genericcontact. At this point, I must admit some confusion about the Magic design philosophy. I am using magic-7.2.43, but also have the source files for 7.2.24 and 7.1. None of these contains the SCN5M_DEEP.12 technology. The file I have has been copied from UNC Charlotte, thanks to Google. I don't know why this file isn't in the Magic deliverables nor whether I have a "legal" copy from UNC Charlotte. I made the MAG files using the scmos.tech file that came with magic-7.2.43. I think this is a version 28 file, as there is no version suffix (ie not scmos.tech28) but the file header has a line format 28 which I assume to mean the same thing. As they say, assumption is the mother of all screwups, so I wish I was more secure about this point. The tutorial documentation has instructions on making a tech file, but when I try the flow seems broken. The source files for the scmos.tech, which I think come from MOSIS, have a header with a copyright dating back to 1995 (file scmos.tech.in). I don't think this has been used for the latest scmos.tech files. How then have the latest tech files been made? In a time honoured tradition, I text edited the MAG file for the iv1v0x1 to put back the metal lines as defined in the input CIF files. This file is attached, and no longer gives DRC errors with the SCN5M_DEEP.12 tech file. A preferred solution is for the Magic cifin to produce this file directly rather than fracture the metal lines in an undesired fashion. Can anyone advise me on this? In any case I will work myself on the problem. Best Regards, Graham Petley ===== Graham Petley ASIC Specialist and Consulting Engineer Metzstrasse 14, 81667 Munich, Germany Phone +49-89-6349-7630 Mobile +49-177-282-5072 http://www.vlsitechnology.org/
|
|
|
|