SMPS LT Spice Simulation - Gain/Phase Margins

[Smillsey]

Hi all

I hope you can help, I have been trying to get my head around gain/phase margins in LT spice and I have been following an Intersil App note (https://www.intersil.com/content/dam/Intersil/documents/tb41/tb417.pdf)

I have been able to design and simulate the output filter stage correctly and I achieve the same bode plots as expected when comparing to the app note and my own mathematics. (The double pole and ESR zero behaves as expected, dipping down near to -180deg and then the ESR zero pulls the phase back around closer to -90deg)

Yet when I come to design the compensator (Type 3) I am getting really confused with the phase response of the system, specifically the compensator phase behaviour.

I tried to create the compensator on its own in LT spice and ran a bode plot on it, I am just using an ideal op amp model for now. This gives the expected zero's and pole's for the gain but I am mystified why I cannot get the same "-90deg to +90deg then to -90deg" phase response as shown in the Intersil app note.

Instead, my bode plots for the compensator start at 90, rise to around 210 and fall off. The overall shape is correct but I am not "starting" at the correct phase.

I have attached an image of the simple circuit, I have arrange the circuit as an inverting op amp as required.



I am pulling my hair out with it, have I misunderstood the phase response of this circuit?

Why doesn't it resemble the phase response in the image below?



Any help will be greatly appreciated.

Feeling pretty dumb right now :thumbsup:

 

[FvM]

The problem is a sloppy representation of the compensator phase characteristic in the Intersil paper. It's normalized to 0 degree phase, means the phase inversion of the inverting amplifier is eliminated. But LTspice shows the actual phase of the inverting amplifier which is the correct answer.

In case of doubt, review a OP circuit design text book.

 

[crutschow]

The plots' frequency limits are also different.

 

[Smillsey]

Thank you FvM,

I though I was going insane.

Crutschow - that doesn't matter the plot was clearly different.

Thanks

 

[Smillsey]

Hi guys

I am continuing on my quest on compensating this loop, I am a bit of a novice when it comes to being comfortable with the phase margins and I fear that my phase margin around the "double pole" of the output filter is too low. It dips down to around 40deg before recovering for a while.

At 0dB crossover I achieve a phase margin of 57deg which is great. But what about that 42deg margin at 3.1kHz (pictured below), should I play with the compensator in order to achieve the rule of thumb 45 degrees. It is worth it for an extra 3 degrees! ?

I need to work on my gain margin as that is only 12dB.

Any pointers? Experience is everything in this game and I am light on that right now!

I am looking forward to building this regulator and seeing it explode into a million pieces

thanks

 

[Smillsey]

Oops

See plot below:

 

[Smillsey]

Can anybody help?

I know I may be asking a silly question but strictly speaking 42 degrees is below the "rule of thumb" 45 degree phase margin.

It is located near the double pole of the output filter.

My understanding is that this is ok and providing the phase does not cross zero at this point we will have a stable loop.

I also have read that it is the 0dB area of the plot which is much more susceptible to environmental variations causing changes in delay and that is why we are only concerned with phase margin at 0dB crossover.

Have I got it right?

Many thanks

smillsey

 

[FvM]

The phase margin stability criterion is valid for loop gain characteristics with single crossover, as the present one is. In case of multiple crossover, you have to check the more general Nyquist criterion.

Phase margin is by definition measured at unity loop gain ("0 db crossover"). There are different rules in use, for a second order system (double pole), you get e.g. no overshoot in step response with > 57° phase margin. If you have multiple poles and compensating zeros as in the present case, it's true that the phase characteristic outside the unity loop gain zone also matters for the system behaviour. In case of doubt, you should simulate the actual step response.

 

[Smillsey]

Ok thank you FvM.

I still feel uneasy but I suppose a "feel" for what is good and what is bad will come with time.

I learning this on my own, which is great for in depth understanding but painful at times!

I need to look into nyquist criterion as I have neglected it so far. I wanted to take one step at a time.

Before I move on to anything else I just want to get this buck converter working.

Thanks