May the Fourth Be With You...,

[JonMarsh]

This is something I've been thinking about for a few days- I'm on a road trip, storage retrieval related, so pretty busy today, so this will get edited and updated as I have time.

My thinking runs along these lines- considering the word play and the basis of this message, "May the Force be with you"- what I envision is a a scene like from the 3rd movie, Return of the Jedi, at the end, where we have some "Force Ghosts" lined up.... and from my perspective, each of us could have our own cast of characters who have had a significant influence, AKA "Force" in our lives, but are no longer with us, some perhaps clearly passing before what should have been their time.

Got the concept?


So first I'm going to enumerate the "cast of characters" for me, and then as I have time today, fill in more details..., such as the relationship to me, and the arc of influence.


Haskell Scott

Haskell was the owner/proprietor of a HiFi shop in Boulder Colorado that I worked at while following up on school, which had their own house brand speakers built literally "in house", that is, in Haskell's basement. Though I worked the store floor part of the time, a lot of time was spent in his basement, mastering the fundamentals of the Radial Arm shop saw, a multi-purpose tool that can do many of the things we expect from a table saw and a miter saw, if not as accurately. Haskell did his designs by ear, which I found left a bit to be desired at time. But I did have a lot of fun building a large TL system for myself, and a "Clone" of the AR-LST.


Ross Godres

Of the many musicians I worked with, working my way through school, Ross was the only one with which we had some personal affinity and regard- he was a guitarist and founder of the local Boulder Band "Family Hand", which I mostly played keyboards for, (B3 and Synth) but also some guitar work and flute. Probably my favorite song with that band was Loggins and Messina "Angry Eyes", because I got to play keyboard, guitar, and flute, all in one song. After the band eventually dissolved, Ross started a music booking business and tried his hand at some other endeavors. we stayed in touch for a while, until I moved to CA with the relocation by Siemens Semicondutors from Broomfield Colorado to Santa Clara, CA. Checking back later, I found he passed away from a heart attack in 2005, leaving a wife Jan (whom I knew) and two children.


William Keith Kennedy

Bill, as I knew him, was the chief designer for a few different speaker companies, Ultraphase being his most recent endeavor before I met him. They produced a three way speaker that was essentially minimum phase, and could reproduce a pulse waveform with some fidelity. But not enough for Bill.



These had relatively thick front panels, but still a conventional cabinet format. I met Bill after I had invested in and become a partner in the Boulder Sound Gallery, a relatively high end oriented HiFi shop. We worked together for several years, from the mid 70's to 1980, when we had some prototype speakers completed, a high end phono preamp and moving coil step up transformer ready, white I'd been running engineering at a small pro sound company called Innersound, which due to some products I developed, the original owners were able to sell to Taco/EV in Seattle. Our development tools including a B&K 4133 pressure zone mic with HP preamp, the White 1/3 octave analyzer, and some fairly critical ears. We had some potential investors lined up the end of 1980, but while I was away at the Winter CES, Bill, who was 53 at the time and a long time smoker, died from a dissecting aortic aneurysm. And that was that, other than the many excellent memories and skills gained.


Charlie Hansen

While I was running engineering at Innersound, my most "interesting" employee was Charlie Hansen, a physics student at CU, and five years younger than I. He was a bicycling fanatic, but also a music lover, and fairly interested in what Bill and I were doing. Charlie wound up being the recipient of one pair of our speaker prototypes, a 4 way, with a much more elaborate mechanical design and, yup, you guessed it, very thick faceted front panels... He was the recipient of a set in 1984 because my wife thought they were ugly in the living room.

Now, Charlie had his heart set on starting a company making his own design bicycles- when that didn't work out, he started Avalon Acoustics over 30 years ago. I visited him a few times after that was up and running, and it was very graceful of him to introduce me as the grandfather of Avalon, though the lineage is more complicated than that.

Charlie sold Avalon to Jeff Rowland, then started Ayre Acoustics. I got to do some consulting design work for Ayre on one of their early preamps, and still have a sample, along with the last notes from Charlie (after production start) on some ways we could improve it with some film bypass caps in critical biasing circuitry.

Unfortunately, while enoying his cycling hobby on Boulder Canyon road, a motorcycle rider in the wrong lane struck him and the accident resulted in severe injuries, especially for his back and neck. He was what is called a partial quad after that, in that he had some use of his arms and hands afterwards, but not with normal control and dexterity. Those kinds of injuries have a life long impact, and Charlie passed away in 2017.

https://www.stereophile.com/content/charley-hansen-1956-2017

Jenö Tihanyi

Jenö worked for Siemens Semiconductors leading the development of power MOS based devices, with many patents, but I'd say his most significant development was the Super Junction MOSFET, which used critically sized alternate columns of p and N material in a n- epitaxial layer, to cut the on-state resistance by a factor of 5 area related, but allowing high blocking voltage. This revolutionized MOSFETs in the 500, 600V and higher ranges. I worked with him and his team in Munich for developing the original launch collateral, and it's from him I got one of my favorite quips, "Slow work takes time..."

Jenö retired at 60, the normal age in Germany, and had just remarried, and had a new child. Very sadly, he died of a heart attack just two years later.


Milan Jovancovic

Milan was a key member of the power development team at Delta Power in VA., and an alumnus of Virginia Tech with close working ties to the CPES group. I met him in my capacity working for Siemens Semiconductors, introducing him to new components like the CoolMOS transistors Jeno developed.

One of the most interesting papers Milan co-authored with the CPS team was the 1995 IEEE APEC paper on a tri-level PFC converter block, which offered the prospect of much lower switching ripple current in both the MOSFETs and PFC power inductors, and pointed to the potential for very high density and efficiency at 99% or above... if only we had the right power transistor switches- nearly ideal perfect switches.

Those did not exist in 1995, and not for quite some time afterwards. But with the development of Silicon Carbide MOSFETs with very low Qoss, and very low loss behavior for the intrinsic diode, especially the new 2nd generation parts just introduced by my former employer Infineon Technologies, (Siemens Power Semiconductors spun off as a stand alone company in 1999), that day is pretty much here, as I'll show with some additional information after I'm back in Idaho.

Jenö was born the year after I was, but passed away in 2018- another loss for his friends and family.

https://cpes.vt.edu/news/news/655

And I'll tie in some of these to relatively recent events, to show the impact of the circle of time.

[JonMarsh]

Back in town after my storage run, and with access to the stuff on my main working Mac...

So, I'm going to follow up on Milan Jovanovic's paper and how modern semiconductors have made it feasible... VERY modern semiconductors!


Fair warning- I'm going to highlight and describe the following:

• The basics of the three level PFC converter, and it's specific advantages with regards to power dissipation and efficiency, especially for higher voltage buss applications- like we have these days with charger systems, and advanced server power supplies
• A published reference design schematic from Toshiba for this type of converter, but illustrating the complexity if trying to use the best modern silicon switch technology, HV Superjunction transistors similar to Infineon CoolMOS.
• Two brief overview of key power semiconductor characteristics for high frequency converters, and power switch issues, including a comparison chart I put together originally for an IEEE Professional Education seminar I prepared in late 2020, with updates since then
• A quick view of a converter assembly for a complete server SMPS being finishing development at my former employer, employing this topology with our newest not yet released 400V SiC MOSFETs, and some efficiency results...

You've been warned...


OK, from the title page of Milan's paper:






The abstract on the left captures the key reasons for this work, and the final paragraph of the introduction highlights the primary benefits.

High frequency switching Ripple current in a power factor correction circuit has to be filtered by the input EMI filter - the more ripple, the larger and more expensive the filter. The next simulation graphs highlight the difference in ripple current for this topology versus the conventional PFC boost converter, using the same size 50uH inductor, and a 100kHz switching frequency, with 185VAC input (low line voltage for 220VAC nominal) and a fairly conventional 400V output bus.




Now, this indirectly highlights a key point- you do not need the same size inductor for the three level circuit, it can be substantially reduced in size for the same ripple current. What this graph above actually shows is that 50uH is far too low for a conventional PFC, and that 50uH is likely more than you need for this Three Level version.

The guidance the authors condensed out of the data goes like this:

"From the above analysis, for the same ripple current, the required inductance of the boast inductor is L for the three-level boost converter, 2x2L for the conventional two phase interleaved boost converter, and 4L for the conventional boost converter. With the assumption of 20% input ripple current (without the EMI input filter) Fig. 9 illustrates the inductor size comparison. The size of the two-interleaved boost converter is slightly higher than that of the three-level boost converter because of high core loss due to larger ripple current in each inductor.



As one of my former colleagues would say, "This is bigly!"


So why hasn't this topology seen widespread use? Well, there's one key issue- it involves clamped recirculating currents like those that exist in motor drives, and requires high performance body diodes in the MOSFETs (Low Vf and low Qrr- reverse recovery charge) and HV Silicon MOSFETs just don't do very well in those regards.

There are ways to get around that, like using LV blocking FETs in series with the HV switching MOSFETs, to block the body diode in reverse mode, and use an external diode part with better characteristics, like an SiC HV diode- those have been around for a while.

This Toshiba schematic illustrates how to do that- but the switching control is complicated, and necessitates a microcontroller and custom code.




Note, this example shows the use of a totem pole half bridge instead of a full bridge input, which cuts one diode out of the input voltage drop losses, and is common in modern PFC circuits.


Now, just to give you a bit more to think about, here's a foil I prepped for several internal training presentations as well as for one of those progressional education workshops:



Qoss is the actual switching charge, from 0V to 400V in this case, for a variety of technology devices; here, I'm plotting it as a function of voltage, so you can see how the epi structure of SJ transistors increases the output capacitance at low voltage. This actually lowers turn-off switching losses, but increases turn-on losses, and is a big problem if you are trying to create low loss resonant switching converters.






Wide bandgap components made with either 4C SiC crystal or GaN have some clear advantages, as they don't have the column n+p structures of HV SJ transistors.

[JonMarsh]

One of the premises I came up with for comparing the usability of switching FETs is to use something other than Ron or ID as the common selection basis- especially for high frequency hard switching converters, and for resonant converters, it makes more sense to me to select parts with similar Qoss, and then compare how the other technology characteristics look in order to get a handle on the overall device technology development and performance.

The following chart is an evolution of one prepared in 2020 for the APEC Professional Education workshop- evolved, in that it includes parts from other manufacturers and the newest 2nd Generation SiC MOSFET technology from Infineon, as well as Infineon's current 650V GaN HEMT transistor. It was prepared in March as part of an advisory presentation from someone who will remain unnamed but wanted my most up to date inputs. Again, the key selection parameter is Qoss, or Coss at 400V if the manufacturer doesn't spec Qoss. (some don't, I guess they don't have a function to evaluate it on their calculator...)




The data is all public information from published data sheets- nothing confidential here. The IMBG65R040M2 is from the new Gen2 series just introduced this last February. The target Qoss range for selection was ~50nC, the GaN part on the far right was the closest. The SiC MOSFETs are all 650V parts. Note, Gen2 includes some 750V parts, and some as yet unreleased 400V parts, which I'm salivating over for potential use in Class D amplifiers.

I'm not going to go into a deep comparison here of the differences in the physical device fabrication and how it relates to these performance differences, but I can if so desired- be careful what you ask for!

A few comments:

• Current ratings derive from the on state resistance AND the thermal impedance junction to case- the lower that last number is, the power power can be dissipated, for a give Ron, and the higher the current rating.
• Note the variation of 25C typical on state resistance- for wide bandgap MOSETs, much of the total device resistance is not in the vertical drift region, but a function of the channel control region, and it's design, more like low voltage silicon MOSFETs.
• From the normalized Tempo of Ron for Tj 25C to 150C, you can see that GaN does not like high temperatures; in point of fact, it's normally prudent to not use it above 100C, and due to other characteristics, the peak current handling is low and the operating bus voltage for long term reliability is low (generally ~400V) compared to SiC or even Silicon SJ. This is why you see GaN a lot in chargers, but not so much in high power applications like Server SMPS.
• While overall gate charge is often used as a FOM for switching speed capability, it should be noted that Qgd charge is the charge actually required to achieve the voltage transition interval in switching. So, pay attention to the Gate to drain switching charge. Many SiC MOSFETs are well suited to lower switching frequencies in applications like motor drives, and you'll see them in electric vehicle appositions, where the higher gate to drain switching charge slows things down and makes EMI easier to deal with. But for power conversion, you want a low number there. Big difference between 5.3 nano-columbs and 25. Also, the dV/dt immunity is directly related to this number, because at high dV/dt, this energy is fed back into the gate, and could result in re-turn on of a MOSFET that is supposed to be turned off.
• ID-Pulse is the drain current peak pulse capability, a function of the transconductance of the MOSFET and also the transient thermal impedance. SiC has very good thermal properties, but if you use split wafers and backside grind them, then this results in a thin wafer construction that has excellent thermal impedance in proportion to the chip size, and chip size is very much related to cost.
• One other interesting data point in relation to design effectiveness I did not include in this chart, is the typical low volume distribution price. Though just introduced, the Gen2 parts are available at my favorite distributors, and you can go and do the research yourself, but I'll give you a hint, the 40mOhm typical part is the least expensive. That's good design for manufacturing.

OK, but back to 3 level PFC! Here's a couple of foils from the Widebandgap Developer Forum recently hosted in Europe last month, which I attended remotely, because of my desire to follow up on developments, and also having shared the Foil above with the VP of the SiC team, whom I've know and worked with for years. He liked the different way of looking at the data, and we followed up with a conference call to discuss this and related topics.


This what I call a teaser foil- they didn't really go into details, didn't even explain the basics of the PFC concept, which are not widely know.






But what really caught my eye was the statement at the bottom of the follow on foil:





And this efficiency graph:





That rise in efficiency at lower load levels reminds me of the best resonant switching CrCM interleaved PFC, where the output capacitance of the switches is largely nullified as a loss mechanism due to resonant switching.


So, I'm really curious to see the preliminary data sheets on the 400V EES (early engineering sample) parts, if my contact follows through...


And to Michael, Ymin, and Milan, I raise a toast and congratulations on how your work from nearly 30 years ago is coming to fruition!


May the Fourth be with you!

[technodanvan]

Hey Jon, I'm sorry these folks passed. Too early, I'm sure. I'd make a post that is similar but frankly I can't think of anyone that had that great an impact on me...partially because I just don't have a great memory, partially because I tend to not get particularly close to people. I also wonder how much other things impacted me, like LEGO and Erector sets, movies like Star Wars and series like Star Trek, to name a few. And then there are all the (mostly fiction) books I read as a teenager, holed up in my room. Nature vs. nurture is a weird thing, since at a certain point you have some internal control over both...but it's probably those events that you have no control over that shape you the most.