ANSYS Maxwell: DC-DC Resonant LLC Split Secondary Converter

Hello everyone, David Giglio here with Ozen Engineering. In this video, I show you how to use the ANSYS Maxwell magnetic transient solver to model and analyze a DC-DC LLC resonant converter.

Here is a PQ50 core, with the primary turns of the primary winding highlighted, the turns of the secondary winding (a split winding) highlighted, the top half of the split secondary winding highlighted, and the bottom half of this split secondary winding highlighted.

Here is the Maxwell circuit used to excite this converter, with two pulses used in series (one pulse positive, the other pulse negative) and no pulses active at the same time. There is a decay constant (a negative constant) and a decay constant time between the pulses.

Here is the resonant capacitor, computed using the excitation frequency and the leakage inductance of the primary winding (which can be computed using the short-circuit test, as demonstrated in my previous video on computing inductances in transformers).

On the secondary side, we have the split transformer windings (basically two windings in series with a ground connected in between them) and the rectifier diodes, output capacitor, and load.

I have used variables for the excitation and timing of the pulses, which I will explain the meaning and use of.

The resonant frequency versus the switching frequency is provided, with no control loop integration because of this, we can work with multiple voltages outside of a regular life cycle (for example, if we have a range of 200 volts).

The output orientation gives us one of two outputs, and this one is being realized here. The current maggie sort for this candidate DELLA efficinator is 29.5 microhenries (the leakage inductance of the primary). The formula to compute the resonant capacitor is shown here.

This converter is designed to deliver a thousand watts (one kilowatt). I use a duty cycle, shown here, and specify the width of the pulse, shown here (where $ represents a global variable, so whenever we use global variables, we use $).

The time rise or time fall of the pulse is specified, and I equal this to a time step (so it's one time step to rise, one time step to fall).

There is a delay time, which I compute with this formula (the total period of the square wave voltage minus four time rises and falls minus two pulse widths, times all this by half).

I specify by how much to delay the first pulse and the second pulse, and I check to make sure that each time variable (the width, delay, and rise and fall time) are a multiple of the time step.

I take these time variables divided by the time step to make sure I get an integer, because this is a time variable.

If I take this time variable and the time step, this means that these time variables are not a multiple of the time step, which means that when the square wave voltage is running in the simulation, it will not perform as expected because it is not a multiple of the time step.

We must ensure that this is checked. Now, we can use basically any duty cycle, but depending on the duty cycle we use, that will affect how much we need to divide the period by. All of this must be taken into account. Let's look at some results.

We have current, which looks very sinusoidal, but there is a slight current imbalance in the secondary windings, which I need to investigate and figure out why.

The curve in red is primary current, and here is the square wave voltage (as I mentioned, I used two pulses in series, none of the pulses are active at the same time, one pulse is negative, the other pulse is positive, and all of this produces the square wave voltage).

I have avoided using switches for simplicity. In another video, I will use switches, but for now, for simplicity, I just use this method. Here is a plot of magnetic flux density and magnetic field strength (the highest values in this simulation).

This is a ferrite cube core on the C97 material specification, where it says that, for example, at 200 milli tesla, the permeability is about 5000. Here, when magnetic flux density is 200 milli tesla, the permeability is 5000, which corresponds to h being around 32. In my results, we see that b is 0.372 tesla, which corresponds to h being around 59. I am operating around this point in the linear region, and I have chosen to use a simple linear material for this example.

In a future video, I will do an example using a non-linear material. I will use the bh loop here. Let me show you some effects. I just want to show the core and air gap. Here are the fringing effects (the magnetic field lines tend to move outwards near the edges).

I have shown you how to set this up, shown you the windings, the electrical circuit, and how to make the connections. The electrical representation corresponds to the physical representation. We've seen some results, and that is all for this video.

Contact us to learn about our simulation capability and request a demonstration for us to show you how we can help you with your engineering projects.

Ozen Engineering Inc is an Ansys Elite channel partner and we provide training to use Ansys tools, offer consulting services, and sell Ansys software packages. Thank you very much, and take care.