What causes this 360 phase shift at ~20khz?

[kevinkr]

There is no inherent reason in this circuit design or the model itself for this abrupt phase shift, it is an artifact of the simulation (software) itself.

I don't know what simulation software you are using but it is likely there is a configuration issue or math error.

Is Vee referenced to (spice) signal ground?

[Fusion916]

Originally Posted by kevinkr

There is no inherent reason in this circuit design or the model itself for this abrupt phase shift, it is an artifact of the simulation (software) itself.

I don't know what simulation software you are using but it is likely there is a configuration issue or math error.

Is Vee referenced to (spice) signal ground?

I'm simulator i'm using is ngspice and yes vee is referenced to ground.

[kevinkr]

If you're using a PC running windows or linux (wine) try downloading and installing LTSpice IV which is free and very good. This should give you the expected result. I'm not familiar with ngspice so I can't advise as to what settings might be causing this result. What is the generator source you are using in the simulation? (sine source?)

[Fusion916]

Originally Posted by kevinkr

If you're using a PC running windows or linux (wine) try downloading and installing LTSpice IV which is free and very good. This should give you the expected result. I'm not familiar with ngspice so I can't advise as to what settings might be causing this result. What is the generator source you are using in the simulation? (sine source?)

I'm running linux with linux compiled ngspice:

Ngspice circuit simulator - NEWS

In general this simulator is very good and I suspect it's something with either the model or some spice settings. Source is just a voltage source DC 0 AC 1. No time domain sinusoidal component.

[DF96]

When software outputs an angle it almost certainly comes from an arctan calculation, or something similar. Assuming you are dealing with normal angles, where 360 degrees = 0 degrees, then +180 and -180 are exactly the same thing. There is no sudden phase shift, just a smooth transition between 179 and 181 (=-179) or vice versa.

So this is not an artifact of the display or the calculation; it is simply a reflection of reality. That is how angles work! Some software can hide this reality from the user, by pretending that an angle greater than 360 is possible - but then the hiding is an artifact.

[Fusion916]

Originally Posted by DF96

When software outputs an angle it almost certainly comes from an arctan calculation, or something similar. Assuming you are dealing with normal angles, where 360 degrees = 0 degrees, then +180 and -180 are exactly the same thing. There is no sudden phase shift, just a smooth transition between 179 and 181 (=-179) or vice versa.

So this is not an artifact of the display or the calculation; it is simply a reflection of reality. That is how angles work! Some software can hide this reality from the user, by pretending that an angle greater than 360 is possible - but then the hiding is an artifact.

In in the arctan function, there is some input in the arctan calculation that causes the output of arcan to go from -179.99 to 180.01 or something similar? Wouldn't that apply that there is a 3rd pole somewhere near 20khz? If so, why doesnt this show up in the gain plot? It's perfectly flat all the way out to 10MHz or 100MHz.

[MagicBox]

You won't have a full 360 degrees phase shift timewise. The line is totally vertical, there's no 'instantanious' phase change. The grapher just changes around the sign and continues using the positive view on the phase.

[DF96]

To get from -179.9 to +179.9 is a phase shift of 0.2 degrees. A rolloff many decades away in frequency can easily do this.

To get an arctan to switch output like that you need the input to change by about 0.0035.

[wojtek5001]

In AC analysis Spice opearates internaly with complex numbers (a+jb, where j^2=-1, see Complex number - Wikipedia, the free encyclopedia if You are not familiar with that). When You display magnitude or phase, then is computed from that complex numbers. Phase is computed using arctan function: phase=arctan(b/a) which always gives You value between -180deg and 180deg.

And also this is AC analysis, so is in frequency domain. There is no time delay, because there is no time. So there is no difference if phase is 365 deg or just 5deg. If You wanna see delay, use transient analysis. If there is really >360 phase shift (at some high frequency), You can see it(although it may be not so easy).
-------edit-------
In some commercial software phase is computed in way that it may exceed -180 - 180 boundaries - I don't know how it is done, but there is probably some simple trick. Mathematically there is no difference, just looks better on graphs.

[Enzo]

When your clock on the wall moves from 11:59 to 12:01, there is an instantaneous 24-hour change. The time line instantly switches from the positive end of one day to the exact opposite end of the next day. This is the same idea.