Cadence Stability Analysis - Local Ground Name

[Khalid_Eissa]

Hi,

I am trying to perform stability analysis on an LDO circuit, everything was great and the results were close to expectations when I was using ideal VDD and GND.
When I added VDD bond-wire (L and R) and also GND bond-wire, the results became very weird: the DC gain has significantly dropped (like 40dB) and the loop phase changed from starting at 180 degree to starting at 0. also I got very wierd high frequency behavior.

When I specified the local grnd name as shown in the following pic to be the internal gnd (before gnd bond-wire and not ideal gnd), the results went back to normal and everything is now OK (same results as before I add the bond-wires to VDD and GND except for some minor differences at very high frequencies, which were expected because of the addition of bond-wires)

So my questions is the following: which solution is the correct solution? before specifying Local Ground Name or after specifying it? and why?

I tried searching online for any material describing how cadence solve and get the loop transfer function with and without specifying local ground name, but I have found nothing unfortunately. Can anyone help please?

Thanks a lot

 

[BigBoss]

Have you ever checked the functionality of the bond-wire stand alone ?

 

[frankrose]

I think might be the issue is the STB probe, inside it there is a current source connected to global gnd, but not sure. 40dB DC drop is very suspicious at all, I wouldn't believe it. I would double check the effect of bond-wire with other method, analysis.

 

[AMS012]

Just do the STB analysis manually. I.e break the loop manually ensuring DC operating point is not affected. Consider the loading at the location of breaking. Now check with and without bondwire the difference. Also, another cleanest solution is to not break any loop, but look at the step response on closed loop with and without bondwire. This should tell you if there's any big problem like you have mentioned.