Beginner Magic layout question: if I add dummy dev...
# chipalooza
b
Beginner Magic layout question: if I add dummy devices by adjusting the x/y tiling multiples for a FET or a resistor structure, is there anything special I need to do to ensure LVS passes? It sounds like Magic/Netgen have some intelligence where they just look at the number of structures connected to a net and make sure the series/parallel combination is accurate. But I want to check before I dig myself into a hole.
m
@Brady Etz you’ll need to add the dummy devices to the xschem schematic.
b
Thank you, David. I'm using the attached image as an example of two P-channel dummy fingers sized 4um x 2um. Can a dummy device be "consolidated into" a single device in layout by adding fingers, as long as the B/G/S/D terminal connections match the schematic?
m
netgen merges unconnected nodes, so you should be ok. Are the source/drain connections on the dummy devices truly unconnected? It seems that at one would be shared with an active device.
b
No, you're right. At least one will be connected with the same net as its neighbor. I just threw that together quickly. I appreciate the guidance!
@Mitch Bailey Follow-up question now that I'm deeper in: I have a PMOS taking a gate voltage developed by a constant-gm bias generator. The PMOS is in the top-level circuit and the bias generator is a subcircuit. In Magic, I would like to combine these devices into a set of PMOS fingers (including dummy fingers) and interleave them to improve matching. Does Netgen care about the cell names in layout not matching the original hierarchy, as long as the flattened/overall cell count is the same? Does this require special handling in the Xschem schematic or Magic layout?
m
@Brady Etz netgen automatically flattens unmatched hierarchies, so you should be ok. In magic, be sure that your device specific layers are in the same hierarchy. You can’t put nwell, diffusion, and poly for the same device in different overlapping cells. (you can sometimes do this with deep nwell though). subcircuit names do not need to match. no special handling (generally).
b
Thank you! On "you can't put nwell, diffusion, and poly for the same device in different overlapping cells..." I've attached an annotated screenshot. Assuming that the well potential for all transistors is the same but the 4-finger device and 10-finger device are from different subcells, my understanding is: Red: Somehow migrate the 4-finger device into the 10-finger device's n-well, sharing the n-well. Connect G/S/D terminals to match schematic. Forbidden. Blue: Overlap the blue transistors. Forbidden. Pink: Add 4 fingers to the 10-finger device. Delete the 4-finger device. Connect all G/S/D terminals to match schematic. Allowed.
m
Pink is your safest choice. Blue. Interestingly, overlapping hierarchal cells with the same devices will extract twice the expected device.
1