Electric Automation Forum
Forum » Power Supply » High Efficiency Converters - Next Generation
Topics: High Efficiency Converters - Next Generation on Power Supply
Start by
Ray Ridley
09-13-2013 04:17 AM

High Efficiency Converters - Next Generation

I want to use as a starting point for this discussion the article A35 in the design center. This article shows how the size of converter shrinks, and its efficiency goes up, if you eliminate the need to regulate. The converter only isolates. Storage elements virtually disappear, and resonant transition switching can be naturally achieved.

Vicor led the way on this with their converters which give 95% efficiency and amazing density. Synquor also did their isolation with a simple push-pull, no exotic resonant conversion, and IR had a demo kit with a half bridge, also achieving 95%. Again, no resonance, just soft transitions.

Now, I see publications coming out with promised efficiencies of 99%. Remember, we are talking high voltage input converters, not little point of load bucks. I'm not sure I believe it, but there again we never believed Vicor could do his converter either, 5 years before it came out.

What is the truth? And what is the role of the LLC converter moving forward? It is claimed the LLC is part of the picture to get to even higher levels, but is that true?

When I was at Virginia Tech, we found there was no magic to resonant converters. The switching loss of regular converters was replaced with losses in the added passive components, and it was a struggle to come out ahead. Resonant inductors were especially challenging with proximity losses causing a lot of problems.

So, two questions:

What are the real numbers that people are seeing for next generation efficiencies? Real-world products, please, not simulations or calculations.

And, is LLC needed to get there?
09-13-2013 06:23 AM
Top #2
Nelson Garcia
09-13-2013 06:23 AM
From the magnetics perspective we have helped a variety of customers in the last few years as they attempt to design LLC with IC's from Fairchild and ST and at the end most customers decide to either cancel the project completely or change to a simpler topology. The LLC seems to be promising but it is clear it is a challenging topology to implement. We have learned to advise the novice engineer to switch to a different toplogy from the beginning,
09-13-2013 08:58 AM
Top #3
Bruno Torremans
09-13-2013 08:58 AM
The team of J.W. Kolar at the ETH Zurich - Power Electronic Systems Lab seem to be designing and building prototypes of some very high efficiency converters (98 - 99%) :

For example :

also commercial products exist at 96% efficiency, more specifically for telecom aplications :

For datacenters the efficiency improvement will mainly come from the 400Vdc power distribution architecture. This architecture removes some conversion cycles between the AC mains and the load (CPU).
09-13-2013 11:43 AM
Top #4
Ray Ridley
09-13-2013 11:43 AM
Fantastic - you answered all the questions on this group in one posting!

First - no LLC needed, the phase shifted bridge wins. That's really what I was hoping to learn and its what Vicor is doing (or close to)

Second, 99% is being approached - with zero line range.

Third, they are using the three-phase buck that come up the other thread about three phase PFC. Good if you don't need the isolation.

Good work Bruno!
09-13-2013 01:55 PM
Top #5
Sanjaya Maniktala
09-13-2013 01:55 PM
Nelson, you said it right.... no one (fully) understands LLC to the point needed to have a simple design procedure that guarantees operation over wide line and load. That is why I wanted to talk to you in April when you visit. Yes the phase shifted full-bridge is good for high powers...but LLC can win at low power levels.

The good news is I believe I have fully mastered the design methodology of LLC...I sat over Christmas break with my family in India...20 days...and focused day and night on LLC...pairing off Mathcad with hand calculations with Simplis...I figured it all out I believe and at Microsemi we are now spearheading LLC --- even in PoE applications. My Second Edition Optimization book, due this year, has one 70 page chapter on resonant topologies and LLC design procedure...that will clear the mystery I think. But you were spot on when you talked about customers trying to get into LLC and moving on...we are here to stay (Microsemi). So join us in building the next LLC converters. The time is ripe
09-13-2013 04:05 PM
Top #6
Ray Ridley
09-13-2013 04:05 PM
Oh well, I thought a thread on a group would end for once, but it seems like the LLC might live on!

I think the question I would have is how well both of the topologies handle some line range variation. The 100% phase-shifted bridge doesn't like it much as you lose the ZVS when you have to tri-state the output and you have to put the inductor back in (see Article A35 again). The LLC, from what I hear from those that know it, drops off in efficiency as well once you start to open up the range.

Maybe it will be a tie....
09-13-2013 06:19 PM
Top #7
Sanjaya Maniktala
09-13-2013 06:19 PM
Dr Ridley, my design procedure is AND-ed with ZVS...so I think I can get wide efficiency...buy we will see as we build of course...paper designs always work well
09-13-2013 08:38 PM
Top #8
Gregory M.
09-13-2013 08:38 PM
Actually,efficiency can be talked about only in relation to operating power of a designed power supply. In my experience a soft-switching mode allows for achieving maximum power within the widest load range. I am an author of a patent on a soft-switching power supply extending soft-switching operation down to very low loads (below 5%).

As to J.W. Kolar and his co-workers' achievements: they have been working on Swiss rectifiers (analog to Vienna rectifiers) for many years but I have a doubt about achieving 99% efficiency by their devices just because these have "abnormal" amount of diodes, and components of these rectifiers (IGBTs or MOSFETs) are subject to excessive stress. You know, it is impossible to achieve a very efficient operation from a device under stress.
09-13-2013 10:50 PM
Top #9
09-13-2013 10:50 PM
Well, they are very clever guys - ultrafast switches and rather low operating frequency. Most importantly the whole system was optimized for a particular parameter. Price wise it's just spectacular, not mentioned unobtanium components like SiCED.
09-14-2013 01:32 AM
Top #10
Harm-Jelle Zwier
09-14-2013 01:32 AM
We have been working with LLC converters for a few years now. Efficiency in practical situations seems to be very high compared to topologies we used before that.
The efficiency we achieve on a 400W power supply (24Vdc out) has an efficiency of over 95% at half load. At full load (17A) this drops to around 94% because of conduction losses in sense resistors and PCB tracks. From 50W upwards the efficiency if over 91%.

Compared to what we did just a few years ago this is truely a leap forward !

I must admit this is all with a fixed 400Vdc input from a PFC. I have no experience with this topology with variable input voltage. To us that normally doesn't matter, because we have to comply with EN61000-3-2 anyway. And using a PFC makes designs so much more easy for wide input supplies.

The best results with an LLC IMHO are achieved with the resonance inductance incorporated in the transformer. The transformers YuJing (Taiwan) supply are possibly the best and easiest to use. (No, I have no commercial interest in that company, just like to share my experiences with you all).

I don't imagine we will make much more progress in the near future efficiency-wise. But that's exactly what I thought 10 years ago when we were proud to reach 80% overall ;-)
09-14-2013 03:48 AM
Top #11
Denis Fariseo
09-14-2013 03:48 AM
The fundamental flaw of an LLC is that it does not handle input voltage variations well. In the medical power industry EN6100-4-11 dip compliance is a must and getting long holdup times (even at reduced loads) is challenging. If you cut the turns ratio to get holdup time you sacrifice light load performance because the converter wants to go to
excessive frequency or goes into pulse skipping mode.

Unfortunately, it seems the LLC is here to stay. It’s like the Fly back of today. No need for output storage choke, need larger power transformer then other topology’s, lots of output capacitance required, low component cost…

Personally, I would like to work with other types of converters. I have only worked with one Phase Shifted Bridge and only to a limited capacity. I remember it drawing too much power at light loads. Given the comments, I think I will revisit this topology.

09-14-2013 06:27 AM
Top #12
Runo Nielsen
09-14-2013 06:27 AM
About resonance converters:
It’s a long story but I will try to make it (too) short.

I believe the statement “no-one (fully) understands LLC …” is correct, even though lots of articles and application notes try to help us. The only known analysis method is the “FHA” which is a rough approximation, and besides, I do not find it straight forward to use.

In my repertoire, the resonance converter has become extremely popular, and it always wins over push-pull, dual switch forward, half bridge and some others, regarding efficiency and noise – and also regarding ease of design! There are two reasons: 1) I have developed my own time domain analysis tool in MathCAD, which is accurate and easy to use. 2) The strongly non-linear and incalculable AC transfer function of an LLC stage can be converted into a pole-zero-free function which is nearly linear all the way from zero load to peak resonance load, so the converter becomes a controlled power source. I got the idea in 2003, unfortunately (for me and the world) it was patented by my customer, but I got a free license to use it. With these tools in hand, it is a piece of cake to optimize and predict the behaviour of a resonance converter and I have designed dozens of them from 40W to > 1kW.

My experience is that indeed the LLC can make an excellent power supply with full load efficiency of 96 – 97% if sync. rectification is used. Sync. rectification is easy because output diodes always turn off with low dI/dt. But I found that the LLC is not ideal for wide range input (or output). If input (or output) must cover a range of 2:1 or more, the LCC converter normally wins the battle. The LCC is probaly unknown to most of you. It’s very strange to me why nearly nobody seems to have discovered the LCC. If you ask Google for LCC, it answers with LLC.
Of course there is a price to pay for a wide range LCC design: A 3:1 LCC has 1.5 times higher rms current than a 2:1 LCC, i.e. 2.25 times higher fet losses if the same fets are used. But it works nicely and is easy to control.
LCC is also the ideal candidate for high output voltage because parasitic output capacitance is absorbed in the 2nd “C”.

Both LLC and LCC run with zero voltage switching from no load to peak load, and can and should do it even with a shorted output. Just be sure to keep them above the resonance point (which is moving!!). No dissipative snubbers, no overshoot, clean waveforms with RF noise an order of magnitude lower than pwm converters.

Today I make approx. 50% LLC and 50% LCC (+ flyback etc). The control idea (we called it Charge Mode Control) works in both. I have math tools for both. For more information, look behind the links in http://www.runonielsen.dk/engelsk/resonans.htm . I should update the worksheet examples soon.

Mr. Ridley’s group is a good opportunity for me to get this knowledge out. Thank you Ray for inviting me :o)
09-14-2013 09:06 AM
Top #13
Ray Ridley
09-14-2013 09:06 AM
Back in my VA tech days, we looked at both the LCC and LLC converters. For some reason, neither of them really took off in industry at the time. I think probably because everyone was trying to do a relatively wide range and it was not suitable for that.

Regarding the AC analysis: I only know one person who could do that, and that is Vatche Vorperian. He did the original SRC and PRC converter for his dissertation at Caltech. When he presented this to us on coming to Virginia Tech to teach, he gave a seminar about his work. Quite honestly, I couldn't even get my head around the first equation that he put on the board, and I doubt anyone else there did either. And that was a lot of smart people.

If you are not familiar with the work, it requires the kind of thing they do in telecommunications, with sideband analysis and other specialist techniques. The SRC and PRC have a fascinating property that they have an LC filter characteristic which pops up at the difference between the switching frequency, and the resonant frequency. So it moves, quite a lot, with the operating point.

I don't know if he did LLC. I suspect it falls in the same class of analysis, and is indeed incalculable to anyone but Dr. Vorperian. I don't know if anyone else out there used the same techniques and did it either. I've invited Dr. Vorperian to join the group, I hope he does so and answers these questions.
09-14-2013 11:57 AM
Top #14
Sanjay Havanur
09-14-2013 11:57 AM
Efficiency and incomprehensibility aside LLC offers one benefit to designers unmatched by other topologies - simplified magnetics. 1+1 winding on two separate bobbin segments, can't get any simpler. Anyone who has struggled with sandwiched windings, three layers of insulation, margin tapes and shields will find the LLC transformer irresistible IMO.
09-14-2013 02:22 PM
Top #15
Ramesh Kankanala
09-14-2013 02:22 PM
In the recent past we have developed and tested digitally controlled Interleaved LLC resonant converter for a 360-400 Vdc input and 24V output DC DC Converter. Prior to that we have also developed a math model using Extended Describing Function for a standalone ACMC LLC resonant Converter.
09-14-2013 04:28 PM
Top #16
ron vinsant
09-14-2013 04:28 PM
And then there is Teslaco (Dr. Cuk) with a "hybrid switching" topology. See (page16),


AC to DC efficiency at 300 watts is 98% but no output specs.
09-17-2014 04:03 PM
Top #17
Charles Potter
09-17-2014 04:03 PM
Is this thread to be continued?  Some interesting remarks therein.  I would like to know of some good application notes for the LLC, SRC and PRC structures.  Is there a chart of topologies like that for PWM converters for the LLC converters?  This would help me to sort out all the confusion with these converters.  Heck, there are so many topologies to consider.
Reply to Thread