https://open-source-silicon.dev logo
Join Slack
Powered by
Hello everyone, I hope you're doing well. I am doi...
# magic
k
Hello everyone, I hope you're doing well. I am doing parasitics extraction from a temperature sensor. During this process, I encountered two errors while attempting to extract two nodes, which correspond to the emitters of the BJTs. It's worth noting that the result of the LVS indicates that the netlists match uniquely. I attach the files created up to the moment of the error and the .mag file. Thanks in advance.
Temperature_Sensor.sim
Temperature_Sensor.nodesTemperature_Sensor.extTemperature_Sensor.res.extTemperature_Sensor.mag
m
@Karla Julieth Camacho Mercado LVS matches before flattening, correct? IIRC, BJT extraction is related to labels on the BJT devices. Maybe something is happening to these when you flatten the layout. @Tim Edwards can advise, I think.
t
Ping me again later if I haven't responded. I am attending the OSDA workshop today.
k
Hi @Mitch Bailey. I type in tckon the commands shown in the image and the LVS match.
m
There might be a misunderstanding here. This is just netlist extraction. LVS is not being run. The option 
ext2spice lvs
, sets certain options (those helpful for netlists used for lvs), but doesn’t actually run comparison. LVS is performed with netgen. If you are running netgen, can you share your output log?
k
When I run netgen, it creates this file...
comp.out
m
Is this the LVS result when using the setup file?
k
When I use the setup file the LVS result is this; however this result is not so different from the previous one. And node errors still occur when extracting parasitics.
comp.out
Hello @Tim Edwards. I have tried to fix the error but I have not succeeded. Could you help me?
t
@Karla Julieth Camacho Mercado: As far as the LVS, you can consider it clean. I was not aware of the 
NE
property in the schematic and I'm not sure if that's a real property or an error in the schematic or symbol. The setup file can be modified to ignore any parameter that is not recognized by the layout extraction so that you can get a completely clean LVS, but it's not necessary.
As far as the error messages when doing full R-C extraction: I will need to look into that. In the worst case, when 
extresist
fails to find a device terminal, the only effect is that there is no resistance calculated at the terminal, so your full R-C netlist will be missing some resistors and will be correspondingly less accurate. However, for bipolar circuits, it's usually the parasitic resistance at the emitter that is important to simulate, so I will try to make sure that is working correctly.
k
Ok @Tim Edwards. Thanks for your support! I'll be aware.
t
@Karla Julieth Camacho Mercado: Can you please do me a favor, though, and post a tarball of all the primitive devices in your layout (e.g., 
sky130_fd_pr__pfet_01v8_lvt_HLJ7ZV.mag
) (there are 7 of them total).
k
Hi @Tim Edwards. Of course, below I attach a tar file containing the devices (I found 2 mag files for the BJTs inside the open_pdks folder, I don't know which of the two you need so I attached both).
devices.tar.gz
t
@Karla Julieth Camacho Mercado: I didn't get any errors when extracting this layout. What versions of the PDK (magic tech file version) and what version of magic are you using? The PNP device in question does not have a substrate---the collector is tied directly to the substrate, so there is no independent substrate connection as there would be for an NPN. That should be reflected in the device model, though, which should only expect to see three terminals and no substrate connection.
k
@Tim Edwards thanks for test it. I have this version of Magic: Version 8.3 revision 326. Compiled on dom 25 sep 2022 131117 -05. If the version of the pdk is the one I found in the attached file, then the version is 1.0.376-0-gf97df97.
pdk_sky130A_libstech_magic.tar.gz
t
@Karla Julieth Camacho Mercado: Your version of magic is most likely the cause of the problem. The current release of magic is version 
8.3.392
. (Version 8.3.326 is from September, which may seem pretty recent, but I have been doing a lot of development specifically of the "extresist" code in the past half year.)
The PDK version is from December and should be fine.
k
Hello @Tim Edwards. Ok, I will update Magic to solve the problem. Thanks for your support!
4 Views
Open in Slack
PreviousNext
Powered by