Mail Archives: geda-user/2015/07/09/09:22:24
On Jul 9, 2015, at 6:15 AM, gedau AT igor2 DOT repo DOT hu wrote:
> 4. And here comes the hard part: gschem would also have two concepts, the current state of the schematics and the desired state and would also indicate mismatches - for connections maybe with similar rat lines and shorts. This is the part that'd involve UI changes too.
Please keep gschem clean.
I’ve tried this with one of the big $$$ tools and it’s messy. Automation doesn’t work right, and changes that involve relationships between pages are chaos.
However, I recently did this with a gEDA/Osmond design in a text-based way that turned out to be simple and straightforward. The layout guy had actually made some pretty extensive changes, eliminating a microcontroller and shoving its functionality into its companion FPGA. He gave me a rough sketch of what he’d done, and I started by redrawing the schematics to match the sketch.
Then, I worked on getting the details precise. Osmond PCB can export a netlist from the layout. A couple dozen lines of script suffice to extract the essential data from an Osmond netlist in a sorted, canonical form. I then just took a diff between the canonicalized netlists from the schematics and the layout.
There were a bunch of discrepancies. There had been some informal communication between the board layout designer and the FPGA designer. The document they gave me for the new FPGA pin layout didn’t match the layout-derived netlist. My procedure detected this, of course, and after some checking we determined that the document was wrong. Good review, and we fixed the document.
So, it was just fix what the diff shows, regenerate schematic netlist, iterate until diff shows nothing. Easy, really. I’d rather see a clean diff than a GUI mess for this kind of thing.
To generalize in the future, it might be nice to define a common canonical netlist format (others have suggested this). For each layout tool, we then need a script to convert its netlist export to the common format. Same with BOM (if separate from netlist), as changes might include footprints.
For comparing the topologies, I had to remove the network names from the netlists because nearly all of the unnamed nets changed their names in this case. The instability of the names of unnamed nets across revisions has been trouble in other contexts as well. I’m considering a gnetlist plug-in that would order the pins an unnamed net connects to and then base the assigned name on the first pin in the list.
John Doty Noqsi Aerospace, Ltd.
http://www.noqsi.com/
jpd AT noqsi DOT com
- Raw text -