Phase margin of the OPAMP

[noor84]

Hi all,

Why does the phase margin on the graph start from 0, sometimes from -180, sometimes +180? what is the reason? what is the starting value of the Y-axis in the Bode plot of the phase margin? any explanation, please?

Thanks.

 

[danadakk]

If you have a inverting amp then at DC, 0 Hz, the phase of the output
is inverted, 180 degrees.

If you have a non-inverting amp then at DC, 0 Hz, the phase of the output
is not inverted, 0 degrees.


Regards, Dana.

 

[noor84]

If you have a inverting amp then at DC, 0 Hz, the phase of the output
is inverted, 180 degrees.

If you have a non-inverting amp then at DC, 0 Hz, the phase of the output
is not inverted, 0 degrees.


Regards, Dana.

Hi Dana,
Thanks for the reply,

Do you mean that in inverting AMP the Y-axis will start at 180 for phase margin, and 0 for non-inverting?

What if the Y-axis started with more than 180? as in the previous figure?

Please, how can I measure the value on the graph for phase margin if it started from +180 or 0 or -180? from where to where can I take the value of the phase margin?

Thanks again for your help.

 

[danadakk]

1) Do you mean that in inverting AMP the Y-axis will start at 180 for phase margin, and 0 for non-inverting?

What if the Y-axis started with more than 180? as in the previous figure?

Yes on ques 1 above.

They started > 180 so they could show the 180 actual, so curve did not fall on axis and confuse
people. But we are confused It was personal choice by author.

You measure phase margin and that is at unity G what is phase, and total from in to out
must be < 360 degrees. The ref point used is arbitrary, the < 360 is not arbitrary. Arbitrary
because you can graph including the internal DC phase shift or implied you already know
in your head what DC phase shift is.

Keep in mind internal phase is a f(freq) and is typically 1 or 2 pole, up to 90 or up to 180
due to the poles alone. Then we have the DC static phase to consider as well. 1 pole typical in
low speed op amps 2 or more in high speed.


Regards, Dana.

 

[noor84]

Yes on ques 1 above.

They started > 180 so they could show the 180 actual, so curve did not fall on axis and confuse
people. But we are confused It was personal choice by author.

You measure phase margin and that is at unity G what is phase, and total from in to out
must be < 360 degrees. The ref point used is arbitrary, the < 360 is not arbitrary. Arbitrary
because you can graph including the internal DC phase shift or implied you already know
in your head what DC phase shift is.

Keep in mind internal phase is a f(freq) and is typically 1 or 2 pole, up to 90 or up to 180
due to the poles alone. Then we have the DC static phase to consider as well. 1 pole typical in
low speed op amps 2 or more in high speed.


Regards, Dana.

Thank you Dana for the explantation,

So, when we need to see the value of the phase margin we always start from the 0db (gain=1) Y-axis of the gain? but what about the Y-axis of the phase margin?
As shown in the attached picture, when they measure the phase margin, they start to find the 0db- Y-axis of the gain (gain=1) then take a vertical line from 0dB gain to find the phase margin; now in figure 1, they measured the PM(phase margin) from 0 to almost 77Deg. in Figure 2, they found the 0dB gain then vertical line to the PM and measured the PM from 0Deg to the point of 0dB gain. It is really confusing!

what is the standard rule of thumb to measure the phase margin from the Bode (from which point on the bode to which point) at the gain 0dB?

 

[LvW]

The phase margin is defined based on the loop gain.
Normally, we only have negative feedback - and we open the loop at a suitable node to measure/simulate or calculate the loop gain.
That means: The negative sign within the loop (because of negative feedback) is and must be part of the loop.
For this reason, the phase function of the loop gain (defined as mentioned above) must ALWAYS start at -180 deg.

Therefore, prior to such a measurement you have to ensure yourself if the phase inversion (for negative feedback) is included in the analysis.
I strongly recommend to use only the above definition (including phase inversion) because there are many applications where this phase inversion does not take place at the summing junction of the loop (inverting opamp terminal) but somewhere in the loop. If this sign inversion is not taken into account for loop gain analysis, this can lead to some confusion (see your question).

 

[noor84]

The phase margin is defined based on the loop gain.
Normally, we only have negative feedback - and we open the loop at a suitable node to measure/simulate or calculate the loop gain.
That means: The negative sign within the loop (because of negative feedback) is and must be part of the loop.
For this reason, the phase function of the loop gain (defined as mentioned above) must ALWAYS start at -180 deg.

Therefore, prior to such a measurement you have to ensure yourself if the phase inversion (for negative feedback) is included in the analysis.
I strongly recommend to use only the above definition (including phase inversion) because there are many applications where this phase inversion does not take place at the summing junction of the loop (inverting opamp terminal) but somewhere in the loop. If this sign inversion is not taken into account for loop gain analysis, this can lead to some confusion (see your question).

Hi LvW,
Thank you for the reply,

So, the rule of thumb is: if the negative sign in the loop is taken, then the rule is (180+ phase at gain 0dB), right?

if no, then the PM is from 0dB Y-axis of PM till to 0dB of the gain y-axis? right?

Thanks.

 

[LvW]

Hi LvW,
Thank you for the reply,

So, the rule of thumb is: if the negative sign in the loop is taken, then the rule is (180+ phase at gain 0dB), right?

No . this sounds not clear to me.
The phase of the loop gain must start at -180 deg for negative feedback (when the phase inversion within the loop is taken into account). Thats all.
For finding the phase margin, you have to identify the phase ot 0 dB loop gain.
Whe the phase is 25 deg less than -360 deg - the PM is 25 deg.

 

[noor84]

No . this sounds not clear to me.
The phase of the loop gain must start at -180 deg for negative feedback (when the phase inversion within the loop is taken into account). Thats all.
For finding the phase margin, you have to identify the phase ot 0 dB loop gain.
Whe the phase is 25 deg less than -360 deg - the PM is 25 deg.

Absolutely, the following is understood (The phase of the loop gain must start at -180 deg for negative feedback (when the phase inversion within the loop is taken into account).

But, this is not understood (When the phase is 25 deg less than -360 deg - the PM is 25 deg).

Could you see the previous picture? there is no -360 in the first figure for example, so how can I know the value of PM? ( in case the negative is not taken in the loop)1

 

[LvW]

Could you see the previous picture? there is no -360 in the first figure for example, so how can I know the value of PM? ( in case the negative is not taken in the loop)1

In this case, the other form of the criterion applies, of course.
When the phase inversion for neg. feedback is not considered, the stability limit is, of course, at another -180deg phase shift (and with consideration of the inversion it is at 0 deg resp. at -360 deg).
Does this not sound logical?
By the way: Here is another argument in favour for the 360deg-criterion: Only in this case, the loop gain determination is in accordance with the oscillation criterion formulated by H. Barkhausen (unity loop gain with zero resp. 360 deg phase shift)

 

[noor84]

In this case, the other form of the criterion applies, of course.
When the phase inversion for neg. feedback is not considered, the stability limit is, of course, at another -180deg phase shift (and with consideration of the inversion it is at 0 deg resp. at -360 deg).
Does this not sound logical?
By the way: Here is another argument in favour for the 360deg-criterion: Only in this case, the loop gain determination is in accordance with the oscillation criterion formulated by H. Barkhausen (unity loop gain with zero resp. 360 deg phase shift)

Thank you LvW

 

[dick_freebird]

How you "break the loop" (or pretend you did by post facto
calculator magic) I think causes some of this flakiness. I find the
stb analysis to be unpredictable as some of this looks; using a
1H inductor might lead to the sort of stark low frequency
inflection (maybe you just shouldn't be looking at sub-Hz
stuff to begin with). No idea how you are feeding the inputs,
or for that matter picking off true input difference voltage
for your denominator. Setup does affect measurement.

 

[FvM]

Another comment regarding the bode plots in post #5. None of the plots has constant gain at the left side, you need to guess what the DC gain and phase is, particularly for case 4.
Inverting DC loop gain (phase = 180 or -180°) is an absolute requirement for feedback amplifiers, thus phase at respectively near DC is a simple criterion for correct bode diagram phase annotation.