Stability Analysis of Opamp

[nahid99]

Hi, i am a newbie in Analog IC design. I perform a stability analysis of my Two Stage OTA. As from the textbook the Gain Margin calculated to the -180 phase along with the (0 - Calculated Gain). But in Cadence from the Direct Plot Grom, I get the GM= 7.67dB, when i cross check with the Gain and Phase Plot, the GM calculated in 0dB. Why its calculated in 0 degree rather than -180 degree. Bellow the figure

 

[LvW]

The stability criterion requires that at 0dB loop gain the phase must be still slightly negative (because for negative feedback the phase starts at -180deg).
So it must not cross the 0 deg line for loop gains still above 0db.
When the stability margins are found by simulation, the calculation always takes into account the phase inversion (180 deg) at the summing junction. This method is in full agrrement with the loop gain definition.
However, for some theoretical analyses (calculation by hand) this phase inversion is not automatically taken into acount - that is the reason for the -180deg criterion.

 

[nahid99]

The stability criterion requires that at 0dB loop gain the phase must be still slightly negative (because for negative feedback the phase starts at -180deg).
So it must not cross the 0 deg line for loop gains still above 0db.
When the stability margins are found by simulation, the calculation always takes into account the phase inversion (180 deg) at the summing junction. This method is in full agrrement with the loop gain definition.
However, for some theoretical analyses (calculation by hand) this phase inversion is not automatically taken into acount - that is the reason for the -180deg criterion.

Thanks, LvW for your replies. Here I attached a figure where the Gain Margin calculated at -180-degree phase plot. I don't understand why cadence calculate it 0 degree.

 

[LvW]

Let me repeat again: The keypoint for the two alternatives for the stability criterion is the question: Is the necessary phase inversion (by design !) for negative feedback inluded in the loop gain- yes or no?

1,) In the avove figure, the loop gain phase starts at 0 deg. If this would be the phase of the total loop gain, we would have 1ßß% positive feedback at low frequencies: No stable operational point - the circuit could not work underclosed-loop conditions.

2.)However, assuming that the design is OK, the above phase response tells me that the shown loop phase did NOT take into account the phase inversion at the summing junction. Unfortunately, some books/authors define the loop gain simply as a product of all the elements within the loop - WITHOUT the mentioned phase inversion . In this case, the critical phase shift is, of course, at -180 deg (because the -180 deg from the supressed phase inversion would give the 0 deg criterion - in agreement with the stability limit at 0db and 0 deg).

3.) During loop gain simulation, the program will, of course, consider ALL the phase inversions in the calculation - and, thus, will consider the 0 deg criterion.

Any further question?

 

[nahid99]

Let me repeat again: The keypoint for the two alternatives for the stability criterion is the question: Is the necessary phase inversion (by design !) for negative feedback inluded in the loop gain- yes or no?

1,) In the avove figure, the loop gain phase starts at 0 deg. If this would be the phase of the total loop gain, we would have 1ßß% positive feedback at low frequencies: No stable operational point - the circuit could not work underclosed-loop conditions.

2.)However, assuming that the design is OK, the above phase response tells me that the shown loop phase did NOT take into account the phase inversion at the summing junction. Unfortunately, some books/authors define the loop gain simply as a product of all the elements within the loop - WITHOUT the mentioned phase inversion . In this case, the critical phase shift is, of course, at -180 deg (because the -180 deg from the supressed phase inversion would give the 0 deg criterion - in agreement with the stability limit at 0db and 0 deg).

3.) During loop gain simulation, the program will, of course, consider ALL the phase inversions in the calculation - and, thus, will consider the 0 deg criterion.

Any further question?

Yes, Thanks for your answer

 

[LvW]

........we would have 1ßß% positive feedback at low frequencies: No stable operational point - the circuit could not work underclosed-loop conditions.

Sorry for the typing error (1ßß%) - must be: 100% of course.

 

[dick_freebird]

If you are putting your trust in results from "stb" analysis,
don't.

"Seventy percent of the time, it works every time!".

I will only use a A=1 vcvs to "instrument" the input difference,
and the direct input and output voltages (mag & phase) for
plotting and calculating. Too much hidden (and sometimes
off in the weeds) "stuff" that's done to make things easy for
you (a trivial bit) at the cost of having to go debug bad data
(and you can't - it does what it wants, deliver bad results
every so often when an old school method comes right side
up 100%).