wafer.space Community - Using gf180mcu_ocd

Mmm, I'm having an antenna violations when connecting to the SEL signal of the ocd IO but I don't see how that would be possible.

21:04

Violating rule is ANT.16_i_ANT.3

21:05

But (1) the Metal2 line that connects to that pin is really not that long, like 90 um or so. And (2) there is AFAICT an antenna diode on that input.

The detail on the error is :

ametal_eff: 202.5162
ratio: 401.817857143
agate: 0.504
max_ratio: 400
pmetal_factor: 0.54
adiodes: (0)
diode_factors: (800)
pmetal: 375.03

21:15

And indeed the gate area is 0.504. But there is an 1 um x 1 um diode on the same node, so the effective area should be 0.504 + 2 * 1 = 2.504 and thus the ratio is 80 which is way below the 400 threshold

Ah ... the DRC rule only consider nplus + COMP to be diodes.

21:32

But the IO have pplus + COMP

Looking at the doc : """This comp area contained the protection diode and any other drain or source junction connected to this node.""" so pplus + COMP is perfectly valid and should be included in the diodes so I'm pretty sure that KLayout deck is wrong. (edited)

21:44

(cc @Leo Moser (mole99) )

Thanks @tnt for the hint! From the DRM:

1. Possible repairs to metal antenna nodes, which have a high via count or antenna ratio. a. Add an NCOMP/Substrate diode for NMOS gate, PCOMP/Nwell diode for PMOS gate. b. Break the metal close to the gate and jog the metal to an upper metal level.

So instead of: diode = nplus.and(comp) I would do:

n_diode    = nplus.and(comp).not(nwell)
p_diode    = pplus.and(comp).and(nwell)
diodes     = n_diode || n_diode

(deep nwell not considered.) Does that look reasonable to you?

TBH I would think comp.and(nplus.or(pplus) would be enough.

07:52

Technically if a signal is connected to what's basically a tap, that's also a diffusion that will dissipate charge I would think.

07:53

But it's a rather weird case so the above would work just fine.