# MOSFET Selection criteria considering conduction losses, switching losses and working frequency

I am trying to choose a suitable MOSFET for my full-bridge resonant converter for wireless power transfer. The circuit topology is something like the following. Specifications: rated power $$\-~500 W\$$, DC supply $$\-~50 V\$$, switching frequency $$\-~200~kHz\$$

I need some help to determine the best-suited MOSFET for high-efficiency considering conduction and switching losses.

Here is my approach so far: I have been following MOSFETS Selection guide by Infineon

1. Decide the range of MOSFETS based on $$\V_{DS}\$$ and $$\I_{D}\$$ (selection $$\V_{DS}=100V\$$ and $$\I_{D}@T_A=25^0C>20A\$$)
2. Select the package type to have sufficiently low lead inductance at switching frequency.

Question 1 - What are the suitable package types at $$\200~kHz\$$ switching frequency? For example, Is $$\{\rm D^2PAK7pin}\$$ good enough, or should I go with something like TO-Leadless/QFN? What is the rule of thumb for package selection at different frequencies? see below picture.

1. Next, I look at the specifications such as $$\R_{DS-on}\$$, $$\Q_{g}\$$, $$\C_{oss}\$$, $$\FOM=R_{DS-on}\times Q_{g}\$$, etc. For example, the following is the comparison of six different $$\{\rm Infeneon-OptiMOS^{TM}5}\$$ MOSFETS (This list can be very long!).

This is where I am lost.

Question 2 - Can I just chose only based on the above $$\FOM\$$? What are the other criteria for setting up the boundaries? For example, What will be the maximum allowed $$\C_{oss}\$$ for these specifications (i.e., at 200 kHz)? I can only find generalized claims about these relationships in the literature.

PS. I have been trying to simulate my circuit with LTSpice, but the simulation was not successful with the available spice models - It takes a very long time to simulate and return an error. I am working on getting the simulations done, anyway I am interested in a faster method for such comparison. Therefore, I want to rule-out the answers "just simulate it and compare the losses"

• Post the exact circuit. Use the simulation schematic that takes too long. Your first picture isn’t a resonant converter. Jan 14, 2021 at 18:56
• The FOM is a good pointing. You should calculate it according to your gate driver circuit. Let say 15V on, 0 V off. Take the value of Q. Then what about thermal characteristics? Those products are nice, but you also need a good PCB manufacturer, they won't make you filled thermal vias, you need lots of tiny holes as thermal vias, at least 4 layer PCB, so on...The best is merely subject to PCB production cost. Jan 14, 2021 at 21:02
• @andy aka it is a resonant. With secondary matched mostky used in wireless power transfer. Jan 15, 2021 at 1:51
• Yes, this is for wireless power application
– Pojj
Jan 15, 2021 at 5:59
• @LarsHankeln, I am working on detailed loss calculations.
– Pojj
Jan 15, 2021 at 6:00

Since you have a resonant converter, you should focus more on the Rdson. The output capacitance also plays a role, during the dead time the output capacitance needs to be discharged. Be aware that the equivalent capacitance is the transformer, output caps of the FETs? as well as the reflected capacitance of the bridge rectifier

So, basically during the dead time you can consider the current through the inductance as constant. This is the current that needs to discharge the equivalent capacitances. If the output capacitances are large and dominant, then ic=c*dv/dt tells us that for a given C either the current needs to be large(large magnetizing current) or the dead time needs to be large. So yes, you want to have both a low Rdson and Output capacitance for resonant topologies, but in my experience the Rdson is far important.

Btw: are you sure about the secondary series cap? This is only required for wireless power transfer. The presence of that cap makes the control complex.

What are the suitable package types at 200 kHz switching frequency? For example, Is D2PAK7pin good enough, or should I go with something like TO-Leadless/QFN? What is the rule of thumb for package selection at different frequencies?

Generally speaking, the shown packages can be used at that frequency. Most critical regarding the additional inductance would be the TO-247 and TO-220 packages. The different surface mount components should perform similarly. Often the same die is available in different packages. Check the datasheet if it makes any meaningful difference regarding e.g. rise and fall time.

For package selection it is also important that you already have an idea about the cooling strategy and the available space for a heat sink. The TO-247 and TO-220 can be attached to a heat sink directly, while the surface mount components need to be cooled through the pcb with thermal vias and probably a heat sink on the other side of the pcb. Depending on the losses, the thermal performance can be critical. For example 10 W of power dissipation can already be difficult to get rid of for SMDs. Smaller packages save space, but make sufficient cooling more difficult.

Can I just chose only based on the above FOM? What are the other criteria for setting up the boundaries? For example, What will be the maximum allowed Coss for these specifications (i.e., at 200 kHz)?

While the FOM can be a guideline for selection, you should not select the transistor solely based on it. An approximate loss calculation as suggested in MOSFET Power Losses Calculation Using the Data-Sheet Parameters can be done relatively quickly and gives you a better idea on what parameters to look out for in the datasheet, when comparing MOSFETs. In addition to this, make sure that you understand the benefits of your topology and how to achieve zero voltage switching (see Navaro's answer). If ZVS is achieved, you could neglect turn-on switching losses in your calculation.

After your first loss calculations you will have an idea, if the conduction losses or the switching losses are more dominant. Then you can look for other parts with either lower $$\R_{ds,on}\$$ or lower capacitances and lower $$\Q_g\$$. Unless you have a specific reason, you shouldn't limit yourself to one manufacturer. Search for MOSFETs using a distributor's search filters.

Nevertheless it is important, that you also get a simulation model going. This helps in understanding the topology and is useful not only for determinining losses, but for driver selection and design, resonant tank design and so on.