I'm currently trying to draw the layout of a device not present in the library of my technology. So, I think to create a symbol to put in the schematic of my circuit and after to draw the complete layout.
When I will finish the layout, what can I do to execute the LVS?
I think the LVS will not recognize the new device. Someone suggested me to add a lot of labels in my circuit near to the non-standard devices.
Is it a good idea?
Are there other more efficient method to execute the LVS of circuit with non-standard devices?
A bit of a vague and open-ended question, unfortunately. It depends what you mean by "non-standard". It may well be that the device extraction in your LVS rules can extract your device, despite the fact that it didn't come from a standard pcell, say. That's because LVS rules would typically extract interacting geometries and recognize the device that way - it doesn't matter what your cell is called, because it looks at the interaction of poly and diffusion (for example, if it's a MOSFET) geometries.
If an existing device recognition does not work, you can either modify the LVS rules to add device extraction statements for your particular device structure, or maybe you can use a "black box" approach to recognize it. Exactly how you do that will depend on which LVS tool you're using (you didn't say).
Sorry, you are right!
I used for my simulation a model of a diode, obtained through TCAD simulations. So, the device doesn't exist in the technology library, but it has been used to design a rectifier!
Now, I have to draw the layout of the circuit, but I think that if I simply draw the layout of the diode and associate a symbol view, the LVS will not work? A solution that someone suggested to me was to insert a lot of labels in the schematic and in the layout, in order to check that the connectivity is correct! Of corse, the LVS will have the errors due to the unknown devices, but at least the connectivity could be checked.
I hope I was more clear...
If the device does not get recognized by the LVS rules, then I think your best bet (if you don't want to code the recognition into the rule deck - which would be worth it if you're using them a lot) would be to use a black box approach.
Relying on labelling nets around the device is risky - for a start, it's hard to guarantee that the mismatch of this device would only cause a localized problem - it could cause a bigger mismatch (this is very dependent on the LVS tool you're using). Also, it makes it much hard to distinguish between real errors and intentional errors such as this.
So I refer you back to the suggestions I made in my previous reply.