Sorry for being absent from this group for a couple of weeks. I have been at APEC for a week, then teaching our workshop in Atlanta.

We've had a lot of discussion about the LLC in another thread. I don't think it can go much further as I discovered at APEC. Many people contacted me privately about their experiences since much of what they know is confidential.

I hoped to get an answer to the real inside of phase shifted bridges and LLC converters, but it's not forthcoming yet. We have seen a posting that one online specification for power supplies explicitly bans the phase-shifted bridge, and other people have voiced their frustration about working with the LLC.

However, I have dug deeper, and found there is a more fundamental question that seems to be being asked in the industry - that is, should bridges even be used at all? There are failure events associated with the antiparallel diode of FETs, and questions about whether they can ever be reliable enough.

I welcome any comments on this.

Does anyone think that paralleling the FET's diode with a Schottky diode would improve the reliability? Would one more step be helpful or even useful that of putting a forward diode in series with the FET and using the anti-parallel diode around the pair insuring no current in the FET's internal diode - this would add a diode drop and reduce efficiency.

Surely the totem pole structure creates a risk of shorting the power supply if noise or any other event turns on the second switch. I was going to look at feeding the bridge from a current source to limit current during any momentary short. My thought was to use an inductor with anti-parallel diode to give the current somewhere to flow, when the bridge supply-to-return path is turned off.

Please give any comments on these ideas.

I agree with you. To me, the totem pole structure is the number 1 cause of bridge failures. This is overcome with gate drive transformers, proper design, careful layout, and exhaustive testing. It is not a short term project.

The second cause of the bridge failures is inadvertant conduction of the antiparallel diode at the wrong time, and its subsequent hard commutation.

I have seen all kinds of schemes devised to avoid this, including the parallel Schottky, but also much more complicated arrangements.

Recent rumblings in the industry suggest that failures in the bridges are a problem, i suspect this is part of the reason.

I suspect his discussion will struggle to get to the bottom of things because those of you on here who are involved aren't allowed to say anything.

What about an AC drive. All over the place, work just fine, do not they? Totem pole.
Resonant circuitry either for soft switching purposes or by the nature of the load itself is the problem... when soft switching ceases.
One ought to make sure that such condition would not occur or, if converter can't tolerate this, design is not up to the task. Some cures are not even better than the disease - I've seen auxiliary circuits larger then the rest of the circuitry...
So are choices - cheap dispensable or nearly indestructible but at what extent
In respect to freewheeling diodes: for instance Cool FETs do not tolerate high dV/dt so don't use 'em in such topologies, period.
It's nice to hear from generation used to work with BJTs. All you need is the inductor in B+ rail, shunted with good fast recovery diode. The whole purpose is to limit dI/dt in case of shoot through, so OVC could turn off the switch safely. Thyristor like control... etc. Well known tricks

Another example: we had to use fast recovery thyristors in grid controlled bidirectional converter. No snubbers would mitigate shoot trough with ordinary devices. It would usually tear apart "dI/dt limiting" reactors...

I built a resonant capacitor charging power supply (had to charge something like 5000uF to 1200V in so many seconds) one time from the AC mains. I tried to use a version of the LLC converter with MOSFETs. The converter was actually a discontinuous mode series resonant converter. ZCS is available for both turn on and turn off of all devices. Theoretically, no snubbers needed if you get the ZCS correct. Anyways, to make a long story short, I initially tried to use ordinary MOSFETs. After quite a few fireworks shows, I put an ultrafast Si diode in anti-parallel with each FET. This got rid of the fireworks. What I ultimately ended up doing was using IGBTs, which ordinarily do not have intrinsic diodes. So, if you use an IGBT with a co-packaged anti-parallel diode, odds are that it is a nice and fast diode that wont cause too many problems.

IXYS makes a neat half-bridge module in an ISOPLUS i-4 package with IGBTs. Its called the FII40-06D-ND. Take a look if interested. They are a little expensive though!

Well I have not had fireworks last 15 years or so... also loosely connected AC distribution bus fireworked at 2kA quite nicely but it was not a converter problem. In respect to body diodes; sure they are not as good as one could expect.

Earlier Ray mentioned that the problem of both switches being turned on at the same time could be controlled with gate drive transformers and a lot of care. However, in some aerospace applications ionizing radiation can turn on both the upper and lower switch in the totem pole so for those specific applications passive current limiting plus a lot of care is necessary and I would say not a desired topology (push-pull, half bridge, or any kind of full bridge). I do hope to make a full bridge work and survive those conditions.

Here is what I think is going on: in aerospace and with older generation engineers, the due diligence was done on the bridge design and testing. This was followed by proper pre-production, more testing, and redesign to catch the flaws.

Nowdays, there is no time to do it right, engineers are less well trained because there is no or little mentoring, and there is an illusion that these things are easy. And, of course, they are not. They never were.

So products are making it to market without proper development and they are failing. Hence the move to ban one technology or another without real reason.