Electric Automation Forum
Forum » Power Supply » Anti-parallel diode in resonant power supplies
Topics: Anti-parallel diode in resonant power supplies on Power Supply
#1
Start by
Marty
09-09-2013 05:09 AM

Anti-parallel diode in resonant power supplies

The anti-parallel diode is now the weakest link in resonant power supplies.

The MOSFET's anti-parallel diodes are not the foremost importance on a device designers mind. Actually, most anti-parallel diodes suck. They can have up to 1.4 V forward drop and be quite snappy. This stress within a resonant supply can be an unforeseen device stress and cause failure.
Your thoughts?
09-09-2013 05:11 AM
Top #2
Ray
09-09-2013 05:11 AM
Actually, the anti-parallel diode can be the weakest link in PWM bridges, and the phase shifted bridge as well. I remember my first full bridge failing because of this.

Always, you have to make sure you don't force-commutate the diode because of its poor recovery, and this can happen with the nonresonant topologies too.

For the phase-shifted bridge, the failure mode can occur at light loads which is always a surprise, and this has bitten a lot of companies in their production.

Perhaps I might rephrase your statement - "The anti-parallel diode is now the weakest link in well-designed resonant power supplies"

that assumes the designers have taken the product past all the other failure modes of cross-conduction in the bridge, EMI, poor layout, transformer, saturation, etc. It is always surprising to me how often some of the big failure mechanisms still exist in production designs.
09-09-2013 05:12 AM
Top #3
Andrew
09-09-2013 05:12 AM
Hi Marty,

I agree the general points that yourself and Ray have made.

However now in fast developing world of Silicon Carbide (See products by GeneSiC Semiconductor) the Vf can be smaller than the parasitic diode in the MOSFET, thus enabling trouble free operation.

Another design consideration is to examine carefully the characteristics of the MOSFET parasitic diode and by playing with parameters the problem can be avoided.

Keeping circulating VAs low in transient conditions is also important.
09-09-2013 05:13 AM
Top #4
Mario
09-09-2013 05:13 AM
Maybe can help the devices specified for motor control application.
09-09-2013 05:13 AM
Top #5
Marty
09-09-2013 05:13 AM
Mario, Motors have always stressed the antiparallel diodes by their commutation current. The typical antiparallel diodes within MOSFETs really do stink. The Si designer's goal is to minimize Rds(on) and not the third quadrant characteristics (diode).
The diode, though, is rated at the same current, but has a much higher Vf (which can be as high as 1.4V), which can lead to over dissipation problems. The currents are very high, even though the conduction time period may be short.

Diverting the resonant (commutation) current around the, typically lousy, antiparallel diode would spread out the dissipation and increase the reliability of the system. But it is a trial and error situation. Si manufacturers have devices that work well for the motor industry (not my strong suit). Depending upon the application, IGBTs are typically used for off-line motor applications.
09-09-2013 05:14 AM
Top #6
Robert
09-09-2013 05:14 AM
I thought that the antiparallel diode was an integral part of every mosfet package and could not be isolated. I have worked very closely with a major Hexfet manufacturer and they usually suggested an additional higher speed device to work in parallel with the package.
09-09-2013 05:15 AM
Top #7
Ilya
09-09-2013 05:15 AM
Snappy intrinsic diodes if used properly can and actually do help to improve the efficiency. No shoot throughs are allowed if one knows how to organize soft switching stuff. 700kHz, 8kW, was not a problem and is still in production when proper soft switching technique is used. Let me know if somebody needs a help in resolving problems like that. "Do not fight the flaws, use them"
09-09-2013 05:16 AM
Top #8
Ray
09-09-2013 05:16 AM
The problem arises when they get out of sequence and conduct at the wrong time. This can be caused by EMI, bad layout, current sensing noise, timing problems, transformer imbalance, etc.

As pointed out above, if you can avoid these second and third-order effects by proper design of the analog power circuit, the diodes can be used effectively for good efficiency.

When they fail, it is a difficult event to pin down, because they don't all fail under the same conditions. It can be a low statistical failure and hard to isolate and get to the root cause.
09-09-2013 05:16 AM
Top #9
Charlie
09-09-2013 05:16 AM
Marty / Ray - OK we know the body diode reverse characteristic of a MOSFET isn't good but can it really fail to recover properly when hard switched? Barring a few esoteric motor drives, all of them are hard switched and there are bucket loads of motor drives using MOSFETs out there that work day in and day out!! I have read about the ZVT failure mechanism at light load (APT and IR have papers on it) but isn't that due to the lack of voltage appearing across the diode to sweep out the stored charge properly in time?
09-09-2013 05:18 AM
Top #10
Jay
09-09-2013 05:18 AM
IGBTs do not have an intrinsic body diode like a mosfet. So, if an IGBT ever comes with a diode, it is always optimized for a nice recovery transient. I try and use IGBTs whenever I can if the body diodes have to conduct current for the application I have.

I built a ZCS series resonant converter to charge 5000uF to 1.2kV initially using MOSFETs. Even though the current was sinusoidal, I was getting MOSFET failures. I switched to IGBTs with co-packaged soft recovery diodes, and the failures vanished. The circuit worked like a charm...no snubbers needed.

IGBTs perform great in resonant converters, especially ZCS. Plus, they can be much cheaper (and smaller) than FETs when the current levels start to rise.
09-09-2013 05:19 AM
Top #11
Ilya
09-09-2013 05:19 AM
Jay, IGBTs are great and can be ran up to 180kHz in similar applications (at least in my cap chargers), but... one has to look at the system as whole. The payment is RMS currents in everything else - transformer, resonant capacitors etc. and IGBTs loose big time if miniaturization is required and switching frequency is high. However in some specific applications if there is no need in high performance PS and price is the most important parameter - IGBTs are championsof the world.
09-09-2013 05:19 AM
Top #12
Jay
09-09-2013 05:19 AM
ILYA: I saw your other comment mentioning 8kW at 700kHz....I immediately thought "Wow." Anyways, yes, you are right about IGBTs. They still loose out at high frequencies when compared to FETs. But yes, there are some applications where they out-perform FETs.

I remember seeing a MOSFET with a co-packaged series diode to disable the intrinsic diode, and an added anti-parallel diode all in the same package. IXYS makes the IXKF 40N60SCD1 COOLMOS POWER MOSFET. Not sure of the availability of this device, but maybe IXYS still makes them.
09-09-2013 05:20 AM
Top #13
Ron
09-09-2013 05:20 AM
For low voltage applications Fairchild makes "S" devices (like FDMC7660S) that have integral Schottky diodes rather than PN junction diodes. Even at these low voltages (14V) I have had standard upper MOSFETs in sync rectifiers fail when the circuit goes into "boost mode" (power is being drawn from the load and pushed back into the bulk supply) due to the parasitic diode. The Schottky devices fix the problem and EMC issues disappear.

In years past we have had sync rectifiers doing Vout ranges of 5 to 200V. We have to place Schottky diodes in series with the drain on the high side and an anti-parallel ultra fast recovery diode that around the series combination of MOSFET and drain Schottky diode.
09-09-2013 05:21 AM
Top #14
Jose
09-09-2013 05:21 AM
I witnessed this problem in ZVS phase-shifted full bridge. and the problem is really worst at high temperature as the reverse recovery time double or triple. It is better to have mosfet with fast recovery diode. Also, we had to lower the switching frequency enough to get away from this problem. The diode has to fully recover during the zvt transition time, otherwise, it will get hit with a very high shoothrough current.

In hard switching, it does not matter and the diode never conducts.
09-09-2013 05:21 AM
Top #15
Ray
09-09-2013 05:21 AM
I have seen it happen in hard-switched bridges too. If you don't put enough primary snubbing in place, the ringing waveform can go all the way up to the input rail, bounce back to the other rail, and if the next half cycle begins at just the right (wrong?) time, it can force commutate the diode, leading to failures.

It is most likely to happen at full load, or under overcurrent conditions, and more likely at low line.
Reply to Thread