I am trying to make sense on the supported resistors in Sky130. In devices list I see different resistors listed (n-diff, p-diff, n-poly, p-poly, li). In the periphery rules I also see relevant rules for rpm, e.g. p+ poly resistor. In the magic tech file I see references to *RES layers.
Are there more resistors supported than the p+ poly resistor ? If so, what are their design rules ?
10/01/2021, 1:28 PM
Yes, I found it rather difficult to correlate between the design rules and the devices, sometimes resorting to the Calibre LVS rule deck to figure things out.
Yes, there are diffusion resistors (n and p), p-well (in deep nwell) resistors, and three kinds of poly resistor (bare poly, high-value resistance, and ultra-high-value resistance).
Check these rules for resistors: poly.3, poly.9, poly.11, diff/tap.14a, all the rpm rules, some licon rules like licon.1b and licon.1c.
Only the high- and ultra-high-res poly resistors have special additional layers and rules. The diffusion and p-well resistors and plain poly resistors are mostly just plain diffusion/well/poly with a resistor ID marker. Some of the associated rules require additional width or spacing to ensure the validity of the device model.
For the "plain material + ID marker" devices, the corresponding GDS ID marker layers are 65:13 (diffusion resistor), 66:13 (poly resistor), and 64:13 (pwell resistor).
10/01/2021, 1:42 PM
Does npc (nitride block) have any other use than for high-res poly resistor ?
10/01/2021, 1:53 PM
Yes. My understanding is that the NPC layer is essentially a silicide block, but apparently the silicide does not make a good contact, and the NPC layer is used under every poly contact to ensure a good contact between poly and local interconnect. I tried to work out the physics/chemistry of that interface but there isn't enough information about it. I'm guessing that there is TiN on top of the poly to act as a silicide but I couldn't find that explicitly stated anywhere in the documentation. The resistance of poly and poly resistor (48 ohms/square) is about right for a very thin layer of TiN, and this layer is stripped off for the xhrpoly and uhrpoly devices.
10/03/2021, 2:26 AM
That is useful. I was trying to understand the npc layer previously.