Understanding the phase behaivor of two diplexers

[e.sotillet]

Hi Folks!!
 I've two different diplexers (same architecture HP & LP, order 7 but different component values), they have almost the same amplitude response but the phase response is quite different. In fact, one of those (with the large phase change) cause an oscillation in a posterior amplifier.


The diplexer is part of a bidirectional amplifier. My hypothesis is correct? The abrupt phase change may cause the oscillation?

Best regards

[fourfathom]

That top one isn't an abrupt phase change.  Instead, you are seeing the phase transition through 180 degrees, so the display snaps from -180 to +180, which is actually the same value.  It's a display thing, not a phase behavior thing. If you were to offset the phase by about 90 degrees the two displays would look fairly identical. [edit: no they wouldn't]
It's still possible that associated impedance change of the diplexer is causing your amplifier oscillation, but again, this is not an abrupt phase change.

[e.sotillet]

So, the phase response of both diplexers are ok? 

Yes, I will try using a broadband load (attenuator) to reduce the impedance variations in the band of interest.


Best regards

[fourfathom]

Sorry, I was wrong about the two plots being similar.  The top plot shows a phase shift change totaling about 675 degrees (somewhat monotonic constant phase change vs frequency over the range shown), while the lower one shows a non-monotonic phase change ranging about 100 degrees or so.  What is probably more important is the impedance presented to your amplifier, and from the charts shown, these two designs are likely too have significantly different impedances.

But what I did get right is that the top chart does not have an abrupt phase change.

[coppercone2]

I noticed that phase thing on VNA before, first time I saw it I thought WTF.

Usually you deal with phase on a polar plot, which is why its so weird.

[e.sotillet]

Here is the smith chart of the main port of diplexer 1 and diplexer 2. Both has good behavior in the interest band, RL>20dB.

One thing I notice is that the oscillation occurs near the "phase transition" thing, that's why my curiosy jumps to this theme. In my regular basis, I dont check the phase of the blocks (filters, amplifiers, etc).

Best regards

[fourfathom]

Again, that's not a phase transition, it's just where the phase goes through the +/- 180 degree display range.  In the one filter it passes through 180 degrees a couple of times, and at some point in the frequency range this can turn a negative feedback condition to a positive feedback (oscillatory) one.  I don't know the details of your amplifier (and I'm far from being an expert in this sort of analysis), but perhaps someone else here can help.

[Joel_Dunsmore]

Phase on S11 is a bit tricky to evaluate and I will note that as the return loss gets better (closer to 50 ohms) the phase values becomes more and more meaningless; in the limit, at exactly 50 ohms, the phase value is undefined (at 50 ohms the phase is the limit (atan2(imag/real) as imag and real tend to zero).  At low return loss values you only really care about the magnitude, at high return loss values, you care a lot about phase as it tells you what kind of component (inductive or capacitive) you need to use to match the impedance to 50 ohms. the key difference in the plots is the difference in group delay to the reflections.  Likely you are "seeing through" the duplexer to the load on the other end, and one has more electrical length (delay) than the other so it has steeper phase slope (delay being the difference in phase over the difference in 360*frequency).   As stated before, the veritcal transitions are just where the phase loops from -180 to +180 and you can use unwrapped phase to eliminate them if you don't like the transitions.