Morning, I would like to know why sky130 cells hav...
# analog-design
n
Morning, I would like to know why sky130 cells have several instances for the same port? [Fig 1] Is it necessary or could I have only one? [Fig 2]
t
If lined up with the routing grid, each text rectangle is centered on a point where a contact can be made cleanly down to the pin. But that is not something that GDS files are used for. All other vendors I've seen have just one label per port. So my best guess is that it is just some artifact of the way that SkyWater generated the cell layouts, or it might have been used by some in-house tool used by Cypress.
n
@Tim Edwards one doubt regarding the layer associated to a port. Why port A is only associated to the upper licon [Fig 1] and not the whole area [Fig 2]? what is causing that split?
t
Magic uses a tessellated plane database (read John Ousterhout's original paper; it's on the opencircuitdesign.com/magic/ web page). When you do an initial selection with the
s
key, it selects a single tile, which will not include more than one layer (because each layer is represented by a tile). If you want to select an entire net, type
s
several times without moving the pointer in between. If instead of
what
you typed
getnode
, it would return "A"---it does understand that the label attaches to the entire net. The
what
command only returns what is in the immediate selection.
n
Got it, thanks Tim. I'll add the paper in my Todo list 🤝🏼
t
It's pretty eclectic reading, but it does help to understand how magic works, especially for understanding why magic doesn't work the same way as a lot of other layout tools do. The specific link is here: http://opencircuitdesign.com/magic/archive/papers/Magic_corner_stitching_original_paper.pdf
🤝🏼 1
n
Some time ago I was reading the website and it mentions that magic makes pretty easy to design cells (absolutely!).