🚀 Let’s start the new week with some in-depth considerations on a 130W LLC PSU from ST devices (AN3106) that can be used as a LED Power Supply. The main focus of this week’s Newsletter is the simulation of the Transformer employed in this design. What else, otherwise? 😜
This time, while checking the Application Note, I could find many details about the core and the windings construction. So, spoiler alert: my simulation closely matches the performance of the original design!
📑 Specs
Taking a look at Table 1, you can see the most relevant specs of the Transformer and the Converter.
We've been lucky this time, as we had enough information to start the simulation of our design. The first step was to use Frenetic’s SUZUKA Circuit SimulatorTM to create the excitation waveforms for this Transformer design, as you can see in Figure 1.
⚙️ Simulation time
Figure 1. Simulating the LLC design of ST app note AN3106
Since we had thermal images of the Transformer in the App Note, where the Converter is working under 130W/100kHz and at nominal input voltage 430V, that’s exactly the steady state condition we wanted to target in this simulation, to make a few comparisons at the end of the Newsletter.
Let's export the Transformer voltages/currents to Frenetic Online and start the design process!
🔎 Core Specs
While checking the Application Note carefully, I could find a Magnetica custom Transformer in the BOM table, and some further details about the core and the windings, like the fact that the core used was an ETD34-PC44. I plotted in the Core OptimizerTMall ETD cores setups that for a certain number of primary turns satisfy the requirement to be below the peak flux density of 250mT, and I quickly found out that ETD34/14/11 was used. Moreover, the material PC44 behaves quite similarly to Ferroxcube 3C94, anf I decided to pick that one, as shown in Figure 2.
Figure 2. Core Selection
Core losses are approximately 1.21W for 47 turns, and the peak flux density is just below 145mT, which is indeed a good operating point for the core.
➿ The windings
As you can see back in Table 1, the number of turns is known for all windings (47:9:9:3), and given that the fact we need 175uH of leakage inductance, a two-chamber approach is the only way to achieve this without the need of an external shim inductor.
Figure 3. Windings
💡 By using a 5mm distance window between primary and secondary windings, we can get close to the 175uH leakage inductance needed. Notice that the two secondaries are grouped together and are wound in a bifilar approach. I couldn’t find out if these windings were grouped together or not directly, except for the fact that the measured DC resistance was identical at 22mΩ, and the only way to achieve this is by using exactly the same wire length. This can be done only with grouping, as you can see in Figure 3.
Figure 4 shows the BOM used for this Transformer:
Figure 4. Bill of Materials
📊 Performance Comparisons
In Figure 5 we compare the simulation against a thermal picture of the app note for the same operating conditions, which were 230VAC input, 130W load, 100kHz switching frequency, 27°C ambient temperature and natural convection for the setup under test.
Figure 5. Performance of the simulated transformer VS app note thermal image
🚀 There is basically no error between the simulation and the reality for the same operating conditions!
We don’t claim that zero error is true for all designs. Depending on the actual cooling conditions, the results may deviate. In this design the core, the windings and the cooling type were all known, so the accuracy of the temperature and power losses model is of course high!
Figure 6. Transformer 1860.0013 created for STEVAL-ILL053V1 - AN3106
✅ Take a deep look to the gap between the two chambers in the thermal picture, and you’ll notice that the temperature there is close 66-73°C, which is really close the temperature of the center leg (68°C predicted in FO). That is a fair assumption, given that the 3D printed bobbin looks like Figure 6.
The leakage inductance with 5mm gap between chambers (as shown in Figure 6.) is predicted to be ~180uH, as needed in the first place.
🚀 And if you didn't have enough of me, then join me in the next Frenetic Webinar!
On April 25th at 17:00 CET I'll hold live lecture on "PFC - Magnetic Design Considerations", analysing the basics of Power Factor Correction and all the necessary steps for its design process.
