Discussion regarding high switching frequency leading to smaller magnetics in Power designs

Slobodan Ćuk has been on social media asserting that “APEC and other Power Electronics conferences and journal publications have been for years making widely disseminated wrong assumption that an increase of switching frequency will proportionally reduce the size of the magnetic cores needed for transformer! Not only that this is NOT the case……”

I commented to Mr. Ćuk that we should take this discussion to our worldwide, vast technical audience of electronics experts on Planet Analog for a true and fair discussion. I have asked Efficient Power Conversion (EPC) to start us off with their view on the subject, since GaN has increased the switching speeds of the power element like never before and then I ask our audience to make their comments and observations from their experience regarding this issue.

Here is Efficient Power Conversion’s Michael de Rooij’s comment to Dr. Cuk’s commentary:

Ćuk converters are certainly an extraordinary development in power conversion. One of several advantages of the Ćuk topology is that it is the only one that can offer continuous current for both the input and output simultaneously when operating in continuous conduction mode (CCM). This reduces ripple in the input and output capacitors but comes at a price; a) it uses two inductors that contribute to losses, b) the output voltage is inverted, c) it induces high ripple current in the interim capacitor, and d) yields poor magnetic material utilization in both inductors.

One of the barriers to wide adoption of the Ćuk topology may be because it competes head on with the standard Buck/Boost topology which only uses one inductor. The Ćuk topology can be combined with the SEPIC topology to yield a single FET split voltage output (±). One commercial product where this technique has been successful in is a low power (1-3kW) wind turbine inverter, where it converts the rectified generator voltage to a high voltage split supply for single half-bridge AC conversion around the neutral. The voltage stress on the FET is input plus output voltage, and it is challenging to make it synchronously switched, given the inverted diode.

GaN FETs bring further advantages to the Ćuk topology in a way similar to that for soft switched topologies; wider duty cycle control. This allows the converter to operate with larger voltage ratios, similar to the 48 V to 1 V Buck converter. For this reason, there appears to be little reason to operate at higher frequency as it buys little in size reduction given the need for two inductors. If the Ćuk topology is operated at higher frequency one could increase control bandwidth, but due to the rectifier limitations, it cannot be fully realized as it cannot sink output current effectively when using a diode.

The adoption of any technology requires certain elements that include its ease of use, enabling something useful, demonstrates cost effectiveness and proves reliable. This is true for power electronic converters. In some market sectors there is a demand for specific characteristics from these power electronic converters; such as low profile, compact size, or new solutions to old problems. These are areas where the Ćuk converter may not be able to penetrate as well.

The characteristics of GaN FETs, such as lower capacitance, lower switching losses and no reverse recovery, yield benefits to the entire range of power electronic converters, including Ćuk converters. The effect of these benefits can be divided into two groups; 1) improvements to existing products and 2) enabling new products.

An example of existing products benefitting from GaN is the AC/DC “brick” power supplies for portable computers where most people would agree smaller and lighter is better. Another example is the trend to have thinner laptops. These trends drive the need for high performance devices that, in turn, push an increase in the operating frequency. This increase in operating frequency is one of the few adjustments engineers can make to their designs to address reducing size, and the Ćuk topology is less physically scalable and therefore cannot meet the challenging size requirements.

As technology evolves, so do the products that can be imagined. Innovation drives emerging markets. The advent of GaN FETs over the last few years has already lead to a variety of emerging products, such a lidar drivers for 3-D real-time mapping, efficient highly resonant wireless power transfer, and envelope tracking power supplies for efficient broadband communications just to mention a few specific examples. These emerging end-use products have no choice but to operate at higher frequencies and they rely on it to succeed. This is no different from when semiconductors first became available and suitable enough for use in power electronic conversion that have evolved over the last 40 years leading to the products we all rely on today. By the same argument back then we would still be using mercury arc rectifiers instead of semiconductors.

At the end of the day this is not a discussion about frequency so much as it is a discussion about market demand and how technologies advance to meet it.

Designers can learn more about simplifying power designs. You can register for the free webinar, “Simplify Power Designs with Micromodules Products” sponsored by Analog Devices

14 comments on “Discussion regarding high switching frequency leading to smaller magnetics in Power designs

  1. D Feucht
    April 7, 2018

    The reduction of core volume with frequency is often true for converters that store input energy each cycle in the core, then transfer it to the output, such as common-inductor and flyback converters. Cores are limited by magnetic field ripple and saturation. The field ripple is related to the field-referred current of the core and the saturation to the static current operating-point. Their product is the core per-cycle energy, and with full core utilization is a maximum. Then the power transfer is proportional to the frequency at which that maximum per-cycle energy is transferred through the coupled inductor. The relationship between flux ripple amplitude and saturation is expressed through the current waveform ripple factor .

    For transformers, the situation is different because little of the magnetizing energy is involved in power transfer. Cuk converters operate at the boundary between transformer and coupled-inductor behavior. Their CCM waveforms at both power ports indicates that both energy storage in the inductor and transformer-type transfer is involved. I discuss this in more detail in Power Magnetics Design Optimization found at innovatia dot com.

    Transfer-power density is maximized in transformers when they are operated at the frequency that allows for maximum power transfer through them, and this depends on the core material. The aspect ratio of the winding window can affect the maximum window current density, and there is a nontrivial interplay between eddy-current losses and winding current density. Generally, simple statements made about magnetic component optimization are oversimplifications.

  2. Steve Taranovich
    April 7, 2018

    Thanks for you excellent  tech insights Dennis

  3. sby
    April 11, 2018

    Michael de Rooij's, The orginal Cuk converter is not the converter under discussion but rather a new one. See:”  

    Further, this is adiscussion about magnetics: 

    “Discussion regarding high switching frequency leading to smaller magnetics in Power design”

     so what is your opinion?

    You say “An example of existing products benefitting from GaN is the AC/DC “brick” power supplies for portable computers where most people would agree smaller and lighter is better.” Why would they be smaller and lighter, just becuase the Si MOSFETs are replaced by GaNs? I think not. 

    Here is mine analysis on the issue posed:


  4. Steve Taranovich
    April 11, 2018

    Hi Sam,

    Thanks for your inputs.

    If you read this article by Gene Sheridan at Navitas, it might explain what we are trying to say

    There is no mystery regarding GaN's high power capabilities at higher speeds than Si. Of course, any good power designer who wants to optimize their design for efficiency and size will take into account the magnetics. My good friends in the power business, as power application engineers, design their own magnetics in order to optimize a power design. Of course, there is no magic that will occur in simply replacing a Silicon MOSFET power element with a GaN MOSFET device—a good amount of design optimization and trade-offs must happen before the final design has achieved the desired optimal performance.


  5. sby
    April 12, 2018

    Hi Steve, Thanks for referencing the article:

    “High-Frequency Magnetics: Lower cost, smaller size and higher efficiency? 

    On the one hand, the title suggests that “none of the above” but then the author writes:

    “Furthermore, the size and cost of the magnetic core goes down as the frequency increases. All of these cost, efficiency and density benefits have been demonstrated by Navitas and many of their partners in the frequency range of 500 kHz to 1.5 MHz.”

    which refutes Cuk's claims.


  6. D Feucht
    April 12, 2018

    A few comments in response to

    “Furthermore, the size and cost of the magnetic core goes down as the frequency increases. All of these cost, efficiency and density benefits have been demonstrated by Navitas and many of their partners in the frequency range of 500 kHz to 1.5 MHz.”

    which refutes Cuk's claims.

    As the magnetic frequency is increased, the ΔB driving the core must be reduced because power-loss density in the core, which has a maximum set by thermal design formulas, would otherwise increase. Yet the power transfer rises in proportion to the frequency. The frequency at which ΔB decreases more than frequency increases is the maximum design frequency for the core. A further increase in frequency causes ΔB to decrease superlinearly and power transfer actually decreases. It is above this fMAX that Cuk is right. However, below it, power transfer increases with frequency. (The actual derivations for fMAX are in my book, PMDO in the section “Frequency Optimization”, p. 255.) So Cuk is right above fMAX and presumably wrong below it.

    If one asks what the typical fMAX is for higher-performing core materials, it is in the GaN, SiC range given by Navitas, et. al. , and the faster switching speed can be put to good use increasing transfer-power density in converters.

  7. sby
    April 13, 2018

    The moral of Feucht's comment, to which I totally agree, is simple: You need to know how to read Ferrite material datasheets. When applied correctly, higher frequency Ferrites will lead to samller size magnetics.


  8. Steve Taranovich
    April 13, 2018

    Just for our audience information 'sby' is an esteemed member of our electronics industry:


    Prof. Sam Ben-Yaakov is the Chief Scientist Officer, Director, and co-founder of Green Power Technologies Ltd. He brings to the Company his entrepreneurial experience in theoretical and practical Power Electronics and analog electronic circuits and systems. Prof. Sam Ben-Yaakov serves as a professor and a researcher in the Department of Electrical and Computer Engineering, Ben-Gurion University, Israel. For several years Prof. Ben-Yaakov served as the Chairman and Head of Institute of Electronics of the Ben-Gurion University. In addition he is a well-known consultant to many companies worldwide.

  9. sby
    April 14, 2018

    Hi Steve, 

    Thanks for the introduction. However, this short bio is very old. A more recent one can be found at:

    AS for Green Power Technologies , in a nutshell:

    Great idea-> patents-> startup-> investments > bad decisions by investors->



  10. Steve Taranovich
    April 14, 2018

    Hi Sam,

    Thank you for that latest bio update—I am sorry about the older bio, but it was the latest one I could find online. I did not think about Linkedin because so many people do not update their bio there

  11. Steve Taranovich
    April 14, 2018

    Hi Sam,

    I am sorry to hear about Green Power Technologies—we need good technology efforts in the renewable sector—especially when waste can be part of the solution.

  12. MrPWM
    April 14, 2018

    This was excellent Sam. A very good tutorial for sizing the transformer for an LLC converter.

  13. sby
    April 15, 2018

    Thanks Steve and MrPWM.

    No question that in the race for higher and higher operating frequency of power processing systems, the magntics is faltering behind. Furthermore, magnetic material manufacturers are slow in catching up even with the materials they already have. For example: core losses of an inductor are a function of not only the magnetic flux deviation but also of the  DC operating point. And yet, vendors provide loss data only for the zero DC case. I wonder how we, the customers, can give the core manufactures a wakeup call. 

  14. RituGupta
    July 11, 2018

    Sometimes not every single perception is concrete but it could be with further probe on its foundation. There needs to be a baseline to fall back on to see if what is being laid out can be relied upon or even trusted to begin with. Without much info, the entire idea needs to be dropped eventually to prevent further misunderstanding on the topic to an even bigger group of enthusiasts.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.