If someone could take a look at this and point out where I v open-source-silicon.dev #analog-design

If someone could take a look at this and point out...

John Kustin

05/07/2021, 10:19 PM

If someone could take a look at this and point out where I've gone wrong, I'd appreciate it. I'm trying to layout 39 parallel PNP devices and 1 PNP on its own (total of 40). All base and collectors are connected to ground. I run locali through the base & collector zones throughout the array (while not touching the emitter). I run metal1 over the emitters. In principle, is this how parallel PNPs are laid out? When I extract a netlist from this layout, netgen reports 40 individual PNP devices, while I expect 2 total PNPs. I'm going to include pictures in a thread to show the connections I've made. Any help is appreciated 🙂

John Kustin

05/07/2021, 10:20 PM

emitter areas connected together by metal1

John Kustin

05/07/2021, 10:21 PM

base & collectors connected via locali. jump up to metal connection to ground is in top right.

John Kustin

05/07/2021, 10:22 PM

the 40th PNP does not have its emitter connected to the rest of the array.

John Kustin

05/07/2021, 10:23 PM

if sharing the

.mag

files makes helping easier, I'm happy to do that.

Weston Braun

05/07/2021, 11:06 PM

have you checked the netlist it generates after extraction?

John Kustin

05/07/2021, 11:28 PM

yeah. i get a bunch of of instances of the form

Copy code

Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1

John Kustin

05/07/2021, 11:28 PM

If this helps, here's the entire generated netlist. If you look through it, let me know if you find anything off

Copy code

* NGSPICE file created from bandgapCoreCombined.ext - technology: sky130A

.subckt sky130_fd_pr__res_xhigh_po_0p35_BNSLGU VSUBS a_n35_n808# a_n35_376#
X0 a_n35_n808# a_n35_376# VSUBS sky130_fd_pr__res_xhigh_po_0p35 l=3.76e+06u
.ends

.subckt sky130_fd_pr__pnp_05v5_W3p40L3p40 Emitter Collector Base m=1
X0 Emitter Base Collector sky130_fd_pr__pnp_05v0 area=1.156e+13p
.ends

.subckt bandgapCoreCombined Vb Va GND
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_9 GND m1_10400_6300# m1_11600_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[1|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[5|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[6|0] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|0] Va GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[1|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[5|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[6|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|1] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[1|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[5|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[6|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|2] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[1|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[5|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[6|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|3] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[0|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[1|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[5|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[6|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_10 GND m1_10800_6600# m1_12000_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_11 GND m1_12400_6300# m1_11600_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_12 GND m1_12800_6600# m1_12000_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_14 GND m1_12800_6600# GND sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_13 GND m1_12400_6300# GND sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_15 GND m1_12400_6300# GND sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_16 GND m1_12800_6600# GND sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_1 GND Va m1_8400_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_0 GND Vb m1_8000_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_2 GND m1_8800_6300# m1_8000_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_3 GND m1_9200_6600# m1_8400_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_4 GND m1_8800_6300# m1_9600_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_5 GND m1_9200_6600# m1_10000_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_6 GND m1_10400_6300# m1_9600_7800# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_7 GND m1_10800_6600# m1_10000_7500# sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
Xsky130_fd_pr__res_xhigh_po_0p35_BNSLGU_8 GND sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ Vb sky130_fd_pr__res_xhigh_po_0p35_BNSLGU
.ends

John Kustin

05/07/2021, 11:37 PM

I might not be reading the LVS report correctly. If it helps, here it is

comp.out

Weston Braun

05/07/2021, 11:41 PM

whats the schematic for this?

Weston Braun

05/07/2021, 11:41 PM

It looks like netgen is reporting a lot more nets in the extracted netlist than the refrence netlist

Weston Braun

05/07/2021, 11:42 PM

so its not an issue of not merging devices. Looking at the netlist it looks like the devices are not all connected to the same nets

John Kustin

05/07/2021, 11:42 PM

image.png

John Kustin

05/07/2021, 11:42 PM

the other thing I have going on is the way I've laid out the resistors

John Kustin

05/07/2021, 11:43 PM

image.png

John Kustin

05/07/2021, 11:43 PM

so i'd expect some extra nets from these intermediate connections

John Kustin

05/07/2021, 11:43 PM

but this does imply a completely different schematic

John Kustin

05/07/2021, 11:44 PM

i can remedy this by modifying the resistors in the schematic, right? i'm pretty set on this kind of layout for matching concerns

John Kustin

05/07/2021, 11:44 PM

the other thing is that netgen correctly interprets this resistor layout as 3 resistors

Weston Braun

05/07/2021, 11:45 PM

for example:

Copy code

Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[2|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[3|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1
Xsky130_fd_pr__pnp_05v5_W3p40L3p40_0[4|4] sky130_fd_pr__pnp_05v5_W3p40L3p40_0[7|4]/Emitter
+ GND GND sky130_fd_pr__pnp_05v5_W3p40L3p40 m=1

John Kustin

05/07/2021, 11:45 PM

in which case, if netgen recognizes there are 3 resistors, why do the intermediate nets matter?

John Kustin

05/07/2021, 11:46 PM

what do you see in that snippet

Weston Braun

05/07/2021, 11:46 PM

oh wait, no. Thats just instance name

John Kustin

05/07/2021, 11:46 PM

yeah

John Kustin

05/07/2021, 11:46 PM

all the emitters are connected together, which is good

John Kustin

05/07/2021, 11:46 PM

otherwise the base and collectors are all connected to GND

John Kustin

05/07/2021, 11:47 PM

at the bottom of the netlist is where the intermediate nets of the resistor network are

Weston Braun

05/07/2021, 11:48 PM

Something else is up because your number of nets do not match

Weston Braun

05/07/2021, 11:48 PM

but I see that the bjt model is a subcircuit. That may be impacting the merging?

Weston Braun

05/07/2021, 11:50 PM

Copy code

Subcircuit pins:
Circuit 1: sky130_fd_pr__pnp_05v5_W3p40L3p |Circuit 2: sky130_fd_pr__pnp_05v5_W3p40L3p 
-------------------------------------------|-------------------------------------------
(no matching pin)                          |Emitter                                    
(no matching pin)                          |Collector                                  
(no matching pin)                          |Base

Weston Braun

05/07/2021, 11:51 PM

can you post the schematic netlist?

Weston Braun

05/07/2021, 11:56 PM

Its actually unclear why netgen thinks there are so many nets. going through the netlist I do not see that many

Weston Braun

05/07/2021, 11:56 PM

When is the last time you updated netgen?

Can Wang

05/08/2021, 12:12 AM

Have you painted a psubdiff and psubcontact to connect the GND to psub? I can't really see it in the layout. Sometimes it matters. @John Kustin

Tim Edwards

05/08/2021, 12:15 AM

@John Kustin: Can you tarball the example and post it to me? I'm suspicous of the PNP device setup because I just pulled the layout for it but things were scrambled between the two PNP devices and I had to untangle it all in the repo. I'm pretty sure that device has never been run through extraction and LVS before.

Tim Edwards

05/08/2021, 12:25 AM

@John Kustin: Yeah, I'm pretty sure this is caused by the fact that the SkyWater layout for that device has pins labeled "Base", "Emitter", and "Collector", while their SPICE model has pins "c", "b", and "e". It is probably easiest to modify the labels in the GDS file and update open_pdks. That is definitely one issue, probably the main one.

Tim Edwards

05/08/2021, 12:39 AM

@John Kustin: I just pushed an update to open_pdks on opencircuitdesign.com to correct this issue (which I did the opposite way, by changing the SPICE netlist to "Base", "Emitter", and "Collector", which was even easier because I already had a patch file to mess with those lines). The quick way to update is to cd to open_pdks and do

git pull ; cd sky130 ; make tools-a ; make primitive-a ; sudo make install

; it's not necessary to rebuild everything.

Tim Edwards

05/08/2021, 12:46 AM

@John Kustin: Actually that fix will make no difference whatsoever to you, because your schematic-side netlist comes from xschem. The xschem symbol for the PNP needs to be changed to have its pins match the layout in the PDK. Also, it needs to have only three pins. The PNP device cannot have a substrate connection because the collector is the substrate. This is SkyWater's error.

Tim Edwards

05/08/2021, 12:58 AM

@John Kustin: I have asked Stefan Schippers to update the xschem sky130 library, but for now, you can just replace the PNP symbol file with this one. That should cause the schematic netlist output for the PNP to match the layout's.

pnp_05v5.sym

Bradley Minch

05/08/2021, 12:58 AM

A couple of project groups in my class were interested in using the pnp device to make bandgap references. I made a symbol in xschem that maps the three-terminal pnp device extracted from magic to the model in the pdk. See the tutorial video (

Magic / Xschem Sky130 Vertical PNP Transistor Layout / Simulation / LVS Tutorial▾

) for an explanation of the set-up. The required files are in the attached zip archive.

sky130_fd_pr__pnp_stuff.zip

Tim Edwards

05/08/2021, 1:00 AM

@Bradley Minch: How does that work for LVS without the pin names matching the layout?

Tim Edwards

05/08/2021, 1:08 AM

@Bradley Minch: And where does the confusion between 05v0 and 05v5 come from?

Bradley Minch

05/08/2021, 1:14 AM

The subcircuit wrapper name came from the device name that magic used for extraction. The pin ordering came from the pin ordering that magic was using for extraction. This was meant only as a stop-gap measure. I had not intended to make it public, but from this thread, it seems like others may benefit from it.

Tim Edwards

05/08/2021, 1:20 AM

@Bradley Minch: Aaaugh, the wrong name comes from my magic techfile!---Not that it matters all that much, because there is no such model, as you noticed since you made your own netlist to convert back to the SkyWater device model name. To get the extracted PNP to work out-of-the box, I will have to use the same tactic that I did for the poly resistors, with a device identifer for the specific device. I will need to work on that. Thanks for the files.

Tim Edwards

05/08/2021, 1:25 AM

@John Kustin: Well, that explains even more. . . the device name extracted from magic is wrong so it doesn't match the device in the netgen setup and so everything about it will be wrong.

John Kustin

05/08/2021, 2:25 AM

Huh! Well thanks everyone for pitching in. @Tim Edwards the device name extracted from magic is the one found in this subckt definition?

Copy code

.subckt sky130_fd_pr__pnp_05v5_W3p40L3p40 Emitter Collector Base m=1
X0 Emitter Base Collector sky130_fd_pr__pnp_05v0 area=1.156e+13p
.ends

John Kustin

05/08/2021, 2:28 AM

I guess I can edit the

model=

field in the new

pnp_05v5.sym

file you sent to get the size of PNP I want then? @Tim Edwards

John Kustin

05/08/2021, 2:30 AM

This is making sense now 🙂 Thanks @Bradley Minch for sharing those files

Tim Edwards

05/08/2021, 2:32 AM

It's a bit of a hack and probably will continue to be until I make a proper parameterized device generator for the bipolar in magic.

John Kustin

05/08/2021, 2:55 AM

Fine by me. Thank you!

2 Views

Open in Slack

Previous Next