Hi @User, sorry to bother you on holiday. When I try to generate an inductor device (from Devices 1 menu) I get the following error:

No import routine for sky130 library cell sky130_fd_pr__rf_test_coil1!
No import routine for sky130 library cell sky130_fd_pr__rf_test_coil1!
    while executing
"error "No import routine for ${library} library cell ${gencell_type}!""
    (procedure "magic::gencell" line 45)
    invoked from within
"magic::gencell sky130::sky130_fd_pr__rf_test_coil1 "
    invoked from within
".layout1.pane.top.titlebar.mbuttons.pdk1.toolmenu invoke active"
    ("uplevel" body line 1)
    invoked from within
"uplevel #0 [list $w invoke active]"
    (procedure "tk::MenuInvoke" line 50)
    invoked from within
"tk::MenuInvoke .layout1.pane.top.titlebar.mbuttons.pdk1.toolmenu 1"
    (command bound to event)

(I'm using magic 8.3 rev 245) Also, I think these coils are just very rough test because I can't even find a spice model in the entire pdk

Holiday? What's a holiday? Thank you for bringing this to my attention. I discovered that the routine names for the inductor devices in sky130A.tcl were incorrect, causing the error. I just corrected this in open_pdks:

commit e4b453599ad0b112e90c89220145c8219dda3cb0

on opencircuitdesign.com, open_pdks version 1.0.262. It is true, however, that there is a huge disconnect in the PDK with respect to inductors. There are layouts for the ones in the parameterized cell menu in magic called

test_coil1

,

test_coil2

, and

test_coil3

. There are SPICE models for three inductors called

ind_03_90

,

ind_05_125

, and

ind_05_220

. I have not been able to validate if the three models correspond to the three test coils. It appers that the values

xx

and

yy

in the name

ind_xx_yy

refer to the number of turns and the radius (?). The number of turns seems to correspond to the test coil layouts, and the

yy

value is proportional to the radii of those layouts, but not equal. So I have held off on equating the layouts and models with each other in the PDK installation. The other issue is that there was one specific variant of the older s8 process that was said to support inductors and had a 3-metal stack with a thick metal3. So there is the additional uncertainty as to whether the test coil models are even appropriate to sky130. Clearly, one would prefer to create coils on the topmost metal, so for sky130, that would mean metal5 inductors. I have one layout of a metal5 inductor from SkyWater called

xind4_01

that comes from an LVS verification layout but has no model anywhere, making it somewhat useless except maybe as a reference for how to do an inductor layout.

Thanks you for the fast and useful reply! I was trying to implement a basic LC oscillator for rf applications but as you said inductor design in the pdk isn't straightforward. I've checked the

ind_xx_yy

spice models and they seems to be for a very specific coils. They needs some (probably a lot!) work to be general or widely usable. Then, reading here I found two extra feature named Inductor-TLM and Cu Inductor of which I don't know the exact purpose. Also, you have raised another doubt: why is the inductor made out of met3 instead of the top layer? I was thinking to run some EM simulation to extract some useful spice parameters but i see it more like a desperate attempt because I'm not sure it will led me to a working "model". Honestly, if you were me, would you try to make this topology work or just let it go and focus on something else (like ring oscillators and so on)? Have a nice day!

There are some inductor designs that were on the MPW runs, so eventually we should get some data for devices with a known layout. However, it never hurts to do more if you have the room on your chip for them (and especially if you have an S-parameter analyzer on hand). Then, it would be nice to run some geometries through a 3D field equation solver like FastHenry to see if it gives meaningful results.

I've set up a 3D EM simulation of

test_coil1

and

test_coil2

from which I extracted some spice parameters (pi-model, nothing exotic). About the first one, I've found

l=1.54e-9 @ DC

and

1.43e-9 @ 1GHz

which seems to be consistent with

ind_03_90

spice model. Also, for the bigger coil I've found

l=5.64e-9 @ 1GHz

which is similar with

ind_05_125

spice model. I'm going to run more simulation and see if I still get consistent result