open-source-silicon.dev

antennacheck-debug.png

So, late last night, I watched the chipalooza workshops--very nice, wish I'd got on board in time to play--and the antennacheck part got my attention.  Though I recall antenna checks from my use of openlane, a cohort just went through the zerotoasic analog course, and there was no mention of any of that.  Kindof assumed it was in the final DRCs or mpw_precheck scripts or something, apparently not.
So this today I did an
`antennacheck debug`
`antennacheck`
and lo and behold, what looks like sadness.
So:
1. how bad is this
2. will it be caught?
3. will it/can it be fixed?

TBH, you're barely 1.5x the limits, doesn't seem like a big deal.

It's the tt_um_psychogenic_wowa design on tt06, original repo at <https://github.com/psychogenic/tt06-analog-wowa>
but... phew!  Ok, well am coordinating with Matt to get some content in the course on that.  And I'm thinking everyone who submitted analog in that course beta group should probably take a peek.

<@U070BEJAXRC>:   The bottom line is that antenna rules are largely written for high volume yield, and so the values are quite conservative for making a test chip.  I don't have a great rule of thumb about this other than that I regularly ignore antenna violations that are less than about 3X the rule limit.  I try to resolve all issues above 4X.  In your case, it looks like you don't need to worry about it.

In addition, for things like Tiny Tapeout or the Caravel user project, magic does not have any information about what is connecting to the project from outside, so without running an antenna check on the entire finished chip (which would take a long while to run), magic will see all the wires to the edge of the project space as hanging antennas, meaning that it will flag a number of violations that are probably not violations if it had global knowledge about the entire chip.  This could be solved (with a bit of coding work) to save antenna gate and diffusion area values in a LEF file, so that the values outside the design area could be folded into the antenna check calculations.  As usual, such solutions must wait until somebody (usually me) has the bandwidth to take care of them.

Ah, ok, this is all quite edifying--thanks!

FWIW in tiny tapeout the analog signals all come from a fairly close large diffusion (the analog switch).  The input signal to TT mux all have diodes on them. And they're also all connected on `met4`  to/from the user modules (with no segments allowed on `met5`).