Hello everyone, I have a question about the layou...
# magic
Hello everyone, I have a question about the layout of a ring oscillator whose frequency is controlled by varactors. When performing the LVS, I get a message indicating that the netlists don't match. Reviewing the file that is generated (comp.out), I have noticed that there seems to be no match in the bulks of the MOSFETs. Does anyone know if this is true or if I'm misinterpreting? I appreciate help. I'm attaching images of the layout, the circuit in Xschem, the message displayed in the console and the comp.out file.
@Antony Brayan Sanabria Calderón You’re interpretation of the results is correct. Do you have Nwell and Psubstrate taps in the layout? The Nwell for the pfets should be connected to vdd and the psubstrate tied to vss. Could you also post your extracted netlist and the netlist generated from the schematic. I’m not used to seeing labeled terminals on the mosfets. That may be causing problems too. Also, there was a discussion on slack a few days ago about the merits of connecting both diffusion terminals of the varactors for better performance.
@Antony Brayan Sanabria Calderón: Circuit 2 (the schematic) is reporting pins "gate source drain bulk" which means that the schematic is using "M" as the SPICE prefix for the device. The devices are all modeled as subcircuits, so Magic will extract them with the "X" prefix. The schematic has to declare each of the devices to be a subcircuit, as well.
@Tim Edwards Do you know if exporting mosfets as subcircuits is an option in xschem? I thought it was the default.
@Mitch Bailey: Like everything in xschem, if you want to change it, you just edit the text in the properties menu. The xschem library should be set up to write them as subcircuits by default, but I think commercial tools regularly convert to the base SPICE types for LVS, so people may be used to doing that, not realizing that netgen is set up to run LVS on the same netlist used for simulation.
👍 1
@Mitch Bailey the sky130 MOS components are netlisted as subcircuits because spice models describe these devices as subcircuits. Xschem adds a
to the instance name (M12 --> XM12). This can be changed by disabling menu `Simulation -> Use spiceprefix attribute`'
Hello everyone, thank you very much for your help. It helped me a lot to understand the problem. After reading your comments, I decided to go into the generated .spice and found that nets names and device names were different. In a simpler circuit, I verified that the difference in the nets names doesn't generate any problems. I then decided to manually change the device names and, when performing the LVS, the console message mentions that there is a spice match. I don't know how to get Magic to generate a spice with the correct device names. Because it generate names like (X#). I attach an image of the console message when doing the LVS.
Below is an imagen of the before and after comparison, thanks a lot.
@Antony Brayan Sanabria Calderón: The device names starting with "X" (i.e., subcircuits) are correct for this technology. You need to change the option in xschem where you are disabling the SPICE prefix (see Stefan's comment above), so that xschem will also produce device names beginning with "X".