Published

Current Power Electronics Topics and Trends: Switching at 2MHz, Devices or Topologies, GaN vs MOSFETs

As I will discuss at Darnell’s Energy Summit 2015: Even a novice to the field of Power Electronics will give you quick and straightforward answer: of course, to reduce dramatically the size, weight and cost of conversion equipment, which go down proportionally to how high is the switching frequency used. Hence, the theory goes all we need to do is to make new switching devices such as GaN devices capable of switching at 2MHz and even higher with following huge advantages:

Part 1: Why Switching at 2MHz?

As I will discuss at Darnell’s Energy Summit 2015: Even a novice to the field of Power Electronics will give you quick and straightforward answer: of course, to reduce dramatically the size, weight and cost of conversion equipment, which go down proportionally to how high is the switching frequency used. Hence, the theory goes all we need to do is to make new switching devices such as GaN devices capable of switching at 2MHz and even higher with following huge advantages:

  1. No need for soft switching as hard switching will do the job!
  2. Once at 10MHz, we can eliminate magnetic ferrite materials with their core saturation and core loss problems and reduce magnetics to just coil windings.

This turns out not only to be a very naïve and simplistic argument but also dead wrong. Before providing the proof for that, we need to answer the additional questions:

  1. Why did we go from 20 kHz (in 1970!) to 2MHz switching now!
  2. What are the benefits or drawbacks of doing that?
  3. Is there any switching frequency beyond which the benefits star to disappear and become actually the hindrance to the Power Electronics progress?

It turns out that answer to all three questions lies in following simultaneous problems:

  1. Classical converter topologies needed a cure in 1970 for its fundamental deficiency of using of PWM DC storage inductors as the sole method of filtering square-wave voltage and square-wave currents generated by the Pulse Width Modulation (PWM) control method of conventional converters.
  2. This resulted in exclusive use of PWM inductors as mandatory filtering components on the output of all basic converters, such as buck, forward, and bridge type converters.

This was and still is considered as an inevitable and led to the following consequences:

  1. The output PWM inductors were not AC inductors, but had to pass the DC current, which in turn led to store all DC energy going through them to converter output.
  2. This is accomplished by inserting into magnetic core an air-gap whose length is directly proportional to the DC current passing though windings so that all DC energy is STORED in the air-gap.
  3. The above “method” resulted in “killing” the filtering inductance value to the point that the magnetic core material did very little (maybe by a factor of 2 to 3) to increase the inductance value above what it would have been if no magnetic core was used but just a coil winding!
  4. At 20kHz switching prevailing at the time in 1970, even a 150W switching DC-DC converter with 5V, 30A DC output would result in such “killing” of inductance value and still very large size and weight of that filtering inductor.

This is when MOSFET switching devices came along to the rescue by providing the switching devices capable of switching relatively efficiently at 200 kHz.

This has therefore enabled the reduction of size of PWM filtering inductors and at the same time increased the DC current levels to say 100A before the magnetic cores become again reduced to become totally ineffective as providing inductance only 2 times higher than possible with just a coil of the same size but without magnetic core and an air-core inductance only.

The inevitable conclusion is that the MOSFET transistors with their higher switching frequencies were simply “hiding” the fundamental deficiency of the widely adopted converter topologies and the related power limitations with consequent size, weight and efficiency) of conventional converters. This has pushed the power limit higher to say 500W as in the above example.

This, in turn, is therefore the origin of pushing the MOSFET to 500 kHz and 1 MHz switching frequencies today. Now come the new devices, the GaN and SiC MOSFETs to take us to the 2MHz and higher switching frequencies.

Are there any problems with that? Yes, because the underlying DC energy storage problem did not go away but was simply pushed to the new levels.

This is also why the paradox: Tesla 200kW AC motor for electric cars operating at 1.5kHz is 100 times smaller and lighter than 20kW DC-DC converter or AC-DC battery charger operating at 10 time higher switching frequency of 150kHz (never mind the huge efficiency differences!)

The beneficial effect of switching frequency increase stopped at about 200 kHz range. While further increase to 1MHz did result in further reduction of size of PWM it had a completely opposite effect on the size and efficiency of the important component: the isolation transformer needed for all applications requiring galvanic isolation and/or need for compact and efficient step-down of the voltage. The transformer was forced to operate at the same 1MHz switching frequency which required specialty high frequency ferrite material. These materials are limited to 20mT AC flux density at I MHz to keep the acceptable core losses per unit volume. Yet, the ferrite materials existed for years which can operate at 100 kHz and 200mT with identical core losses. Consequence: no reduction of the size and weight of transformer at all except for much increased copper losses, and reduced efficiency.

Ten times reduction in operating flux density from 200mT to 20mT does not result in any reduction of transformer size. In addition the transformer has no DC energy storage requirement and leads very naturally to the scaling of power without the current saturation limit of PWM inductors. Clearly output PWM inductor must be eliminated from the successful isolated DC-DC converter topologies.

The good news is that the new storageless switching method already exist which accomplish exactly that!

The corresponding novel converter topologies result in an inherent tenfold reduction of the magnetic flux and corresponding reduction of the magnetic size even at the moderate 100 kHz switching frequencies, while at the same time eliminate output PWM DC storage inductors which cause the problem in the first place. As a result a tenfold reduction of the magnetics size (and converter size!) is made possible despite the modest 100 kHz switching frequency in comparison to 1MHz switching frequency and higher of the conventional DC storage topologies.

The detail explanations of how is this accomplished will be provided by Dr. Cuk in his Plenary presentation at Darnell’s Energy Summit 2015 on September 29, 2015 entitled: Power Electronics for the 22nd Century.

Part 2: Unlimited speed highway myths

  1. No need for soft switching as hard switching will do the job!
  2. Once at 10MHz, we can eliminate magnetic ferrite materials with their core saturation and core loss problems and reduce magnetics to just coil windings.

This turns out not only to be simplistic argument but also dead wrong as it leads to the wrong conclusions propagated by companies developing switching devices and related drives (power management):

  1. Converter topologies (how the switching devices are connected to inductors, transformer and even capacitors (!) is irrelevant and 50 year old topologies (buck, boost, flyback, full bridge, etc.) are just fine!
  2. Magnetic components (transformer and inductors) are stuck with their size and weight limitations imposed by above topologies and nothing can be done about that except going to ever higher switching frequencies.
  3. The AC-DC and DC-AC conversion is forever defined as a higher level power processing requiring the multiple use of the cascaded constituent DC-DC converters (with their DC storage limitations included!) and more switching devices, and proportionally increased losses, size and cost need for such conversion.
  4. All switched-mode power conversion (DC-DC, AC-DC and DC-AC) must use energy storage if required to regulate against input line voltage or load current changes.

The detail explanations of how is this accomplished will be provided by Dr. Cuk in his Plenary presentation at Darnell’s Energy Summit 2015 on September 29, 2015 entitled: Power Electronics for the 22nd Century.

Part 3: Devices or topologies: what comes first?

Clearly there are now two viewpoints which have clearly emerged over the last 40 to 50 years.

Viewpoint that new switching devices come first

All credit for progress of Power Electronics in last 50 years is solely due to advances in new switching devices, both in terms of new types, bipolar, MOSFETs, IGBTs, and now GaN and SiC MOSFET devices. If it were left to system designers (those knowledgeable about converter topologies, magnetics and control) there would have been no progress made at all!

Even brief look at converter used today will find the same converters, such as buck, boost, forward, flyback and bridge type converters. These converters are just fine, they only need to use newer type of switching devices operating at ever higher switching frequencies.

The system designer task is then very simple: when the new devices like GaN devices now, just replace the MOSFETs with GaN devices and you will get a state-of the art-designs. Of course, it is mandatory that the new devices offer even faster switching speeds as that was the only method perceived to lead to the reduction of size and weight of the switching power converters.

They are all fixated on having their new devices enabling higher switching speeds like GaN switching at 2MHz to replace the MOSFET technology operating at 500 kHz to 750 kHz. These efforts are already demonstrated by the very same GaN manufacturers to result in marginal efficiency improvements (1% at best), while not offering any reduction in size and cost!

Opposing view: system technology comes first

This starts with the realization that system technology (topologies, magnetic and control) proposed originally 50 years ago was based on a brute force square-wave (voltage and current) switching method and resulted in simple-minded hard-switching converter topologies laden with very bad properties such as: excessive voltage and current stresses of the switches and large hard-switching losses. The magnetics components such as PWM inductors have fundamental flaw requiring a storage of DC energy and resulting in big obstacle in scaling to higher power and higher efficiencies with smaller size and weight.

So the first objective is to find the sources of such bad traits of present converter topologies and eliminate them in the novel converter topologies operating under new switching principles.

Therefore, novel switching methods and converter topologies come first as they singlehandedly remove all the hugely limiting constraints imposed by conventional topologies on all switching devices no matter what their origins are, be it MOSFET, GaN or SiC MOSFET! It is not devices but topologies and switching methods which matter the most!

Shootout: new switching devices vs. novel topologies

There is an obvious resolution:

  1. Build the prototype based on new converter topologies and new switching methods but using the OLD MOSFETs and 150 kHz switching.
  2. Build the same specification prototype based on the best conventional topologies but using the latest GaN devices operating at 1.5MHz switching.

Compare the two prototypes in terms of efficiency, size and cost and declare the winner.

There is no doubt that the big winner will be new converter topologies based on novel switching methods in all three categories of simultaneously reducing dramatically losses, size and cost.

Ultimate future: best devices with optimum topologies

GaN devices designed to take full advantage of the low voltage and current stresses of the new topologies and much better utilization of magnetics and capacitive components will be ultimate champion. The reason is that GaN devices offer for the first time the properties which CANNOT be achieved by any other past switching device technology such as:

  1. Planar technology which leads naturally to monolithic integration providing all switching POWER devices being interconnected in on a same single chip. Together with its inherent much reduced size (4 times and more) compared to MOSFET and owing to new storageless converter method and related converter topologies promise to result soon in first truly System on Chip (SoC) switching converters.
  2. The GaN device unique feature distinguishing it from all other switching devices which can be in direct AC-DC power conversion with PFC and isolation in a single power processing stage.

The detail explanations of how is this accomplished will be provided by Dr. Cuk in his Plenary presentation at Darnell’s Energy Summit 2015 on September 29, 2015 entitled: Power Electronics for the 22nd Century.

Related Items:

Recommended for you

To Top