https://open-source-silicon.dev logo
#osu
Title
# osu
t

Teo Ene

07/21/2020, 3:26 PM
We have not checked them, no. When we started, we were not fully aware of the SkyWater cells, and by the time we found them we had already established our own design choices. We have welltaps inside every std cell, yes. We made them fit so that they do not take much additional space at all, though perhaps our power rails are on the thick side as a result. Thick power rails is not bad though!
a

Adrian Freed

08/06/2020, 11:58 PM
@Teo Ene I was encouraged to hear the OSU cells will have thick power and welltaps in very std cell. This helps for analog applications of those cells. Another thing that helps is a clkinv cell with approximately 3/1 or 4/1 p/n width ratio. Will there be one of those? Also The Sky130PDK has 1x 2x 4x 8x 16x drive versions of many cells. Will there be different drive versions of clkinv or any of the other cells? Thanks.
t

Teo Ene

08/07/2020, 12:03 AM
• Thanks! • Right now, our CLKINV is 3/1 p/n width ratio, as according to the spice models this makes it balanced. • There will be different drive versions of all the cells. Right now, all or nearly all cells have X1 and XL (low-drive). Most cells also have X2 and X4, and a few of the simple ones have X3, X4, X6, X8, X10, X12, X16. It's just a matter of man-hours to create higher drive-strength versions of other cells as well.
a

Adrian Freed

08/07/2020, 12:31 AM
@Teo Ene One of the designs I am looking at annoyingly would require an X5… The 3/1 ratio makes sense. I haven’t figured out why the Sky130 libraries use 4/1 which is what I also see commonly in papers that build analog circuits from inverters. They only use something near 3/1 in the high-density low-leakage library.
t

Teo Ene

08/07/2020, 1:41 AM
X5 should be no big deal for an inverter
I haven't looked at the SkW foundry std cells, but I'm not sure why they'd use 4/1. I've always learned that somewhere around 3/1 / 2/1 is where you should start off, and then fine-tune it based on the tech.