stability analysis RC coupled Vs RC decoupled

Not as far as I know (and I can't really see how it could have an issue). You should contact customer support, ideally with a testcase that shows the problem.

Andrew

Dear Pirate King,

> I have a regulator giving a reasonable
> RCdecoupled phase margin but when
> RCcoupled is chosen then the PM drops
> below zero.

My ignorance will be showing,but my humility makes me ask the questions:

When you write "RCcoupled" and "RCdecoupled" are you referring to the netlist containing capacitors that are either between nodes or all referred to ground (i.e. with no intranodal capacitances)?

If my understanding is correct for your definition, and without knowing the specifics of your regulator design, is it seems possible the use of the netlist containing intranodal capacitors is impacting your compensation? Miller capacitance effects will not be present in a de-coupled netlist and compensation networks often take advantage of Miller capacitance. The second issue that came to mind is if the use of intranodal capacitors has introduced a second feedback loop into the design. This would also have a significant impact on a phase margin computation.

Once again, I apologize for my lack of understanding of your RCcoupled and RCdecoupled terms - which may make my comments totally irrelevant!

Shawn

Dear Pirate King,

> I have a regulator giving a reasonable
> RCdecoupled phase margin but when
> RCcoupled is chosen then the PM drops
> below zero.

My ignorance will be showing,but my humility makes me ask the questions:

When you write "RCcoupled" and "RCdecoupled" are you referring to the netlist containing capacitors that are either between nodes or all referred to ground (i.e. with no intranodal capacitances)?

If my understanding is correct for your definition, and without knowing the specifics of your regulator design, is it seems possible the use of the netlist containing intranodal capacitors is impacting your compensation? Miller capacitance effects will not be present in a de-coupled netlist and compensation networks often take advantage of Miller capacitance. The second issue that came to mind is if the use of intranodal capacitors has introduced a second feedback loop into the design. This would also have a significant impact on a phase margin computation.

Once again, I apologize for my lack of understanding of your RCcoupled and RCdecoupled terms - which may make my comments totally irrelevant!

Shawn

Thank you Shawn, yes we are suspecting there is a parasitic loop that is not being opened along with the intended loop for stb analysis. Transient sims on the bad corners are showing no oscillation. Thank you for the warning of parasitic miller effects, there is an intranodal parasitic C around the large pass device so perhaps this is giving a parasitic miller effect which is not seen with RCdecoupled.