Post-layout Results are not correct

[hosseineslahi7]

Dear experts,
I have designed a simple design working as a level-shifter and buffer at 1GHz in GF22nm technology. Then, my colleague and I designed two roughly similar layouts for my design, however, we see completely different post-layout results. We carefully connected the back-gates, DEEP NWELL layers, substrate, gnd and VDD to the correct biases on both layouts. The floorplans are also similar with minor changes in the wiring routs only. So, we could not understand what is going on.

Then I realized that if I change the location of only two PMOS devices (P9 and P8 on schematic) in my layout, as I showed in figures 1 and Figure 2, posy-layout results will change dramatically. You can see the difference in figure 3. It means a small change in layout will shift the result.

Does anyone know how it is possible? Is it because of my design or something is wrong in PDK?

 

[dick_freebird]

Perhaps is it a proximity coupling effect, why don't you run the
two analog extracted netlists side by side and see what you can see?

 

[yardum]

if it's layout dependant effect then one by one switch off the WPE (well related) and STI (stress related) switches and observe the difference. IF they do not make a difference you should look very carefully at your layout.

 

[hosseineslahi7]

if it's layout dependant effect then one by one switch off the WPE (well related) and STI (stress related) switches and observe the difference. IF they do not make a difference you should look very carefully at your layout.

Thanks for your response.
How can I control STI and WPE switches? I have not done that before?

--- Updated Dec 29, 2023 ---

Perhaps is it a proximity coupling effect, why don't you run the
two analog extracted netlists side by side and see what you can see?

Yes, It is a good idea. I had to run simulations with/without the floorplan parasitic effect and see what happens. It is not a logical method but I had to optimize the floorplan to find the best one for layout. I did it by using the Layout "(Parasitics/LDE)" option in ADE XL window. However, I saw weird post-layout results after wiring and I should meet this problem also.

 

[yardum]

well if its calibre LVS for PEX a form appears and maybe the same for PVS for PEX and you'll see WPE and STI options on a list of things that can be set during LVS because for obvious reasons this is where the well and STI dimensions are extracted for use in the bsim model. This also requires an additional license to use. If you cannot see these I'll take a screen shot and post to help you identify what to look for.

I have also fould that if you accidentally leave some unused or wrong layer it might extract properly but simulate wrong. Once I left a np layer square soemwhere and it messed up the simulation. Daft careless mistake I made I know but sometimes these thing happen!

 

[D.A.(Tony)Stewart]

If there is crosstalk from Vinp to Vb that is positive feedback and leads to instability.