[Solved] Ideal diode reliability

[blueskull]

Has anyone used ideal diodes (active OR-ing controller+MOSFET) in critical environments?
Will it fail easily? I'm thinking to use one of them on a battery backup device, but the battery I use does not have BMS, so if the ideal diode fails, the battery will be overcharged and explode.
The battery charger is designed with triple safety feature that practically, will never fail. The only possible battery overcharge path is if the active OR-ing fails.
Has anyone used field effect rectifiers? They also look promising -- low Ir compared to low Vf schottkys, and low If compared to normal schottkys.
So, any suggestions on will it fail? I'm looking at LTC4358, integrated 5A ideal diode, or FERD15S50, 50V 15A FERD.

[danadak]

Reliability Calculations would have to be done based on
vendor characterization data.

Google "reliability calculations", lots of info available.

Regards, Dana.

[digsys]

We now sell 100's P/Supplies / OR units based on these type devices. Up to 30A. Never ever had an issue, and I've run destructive tests.
One of my favourite jiggers of the decade :-) The PTC behaviour is also a bonus. Good lo' BravoV put me on to these a while ago.

[Miyuki]

Look for example at commercial power supplies with ICE2HS01G it drive primary side and synchronous rectifier and in Corsair ATX PSU it have 10 years warranty. For how long it must be designed if they can offer so long warranty?

[Miyuki]

  I read wrong what you need 

[Siwastaja]

If they fail, you get a toasted PSU. If a battery OR-ing fails, it causes a fire.

I think you are oversimplifying the battery.

As you see, electronics can never be 100% reliable. This is why the li-ion industry has developed protection features on many levels (chemistry, construction, electrical). Just don't use random Alibaba cells, as you shouldn't anyway; get proper brands.

In a lab LCO cell, thermal runaway probably happens and flames emit when charged above 4.5V. A high-quality production cell is different.

Many proper cells even have overcharge conditions specified in the datasheet: such as many hours connected directly to a strong 12V supply. The cell of course dies and heats up a lot and may vent out some pressure, but no destructive fire should be seen.

I have tested a Sony 18650 myself connected to a 40V 5A supply for at least an hour. The cell didn't even vent; it become a thermostat heater, the integrated PTC has steep curve and large hysteresis and was controlling the cell on and off. The green plastic heat shrink wrapping around the cell didn't even melt or anything, it was a boring test.

This being said, you should never abuse li-ion cells. It's kind of risky. But it's already a somewhat managed risk. Battery electronics and BMS makers all take the risk and place some amount of trust on the cell level protections.

Use a plain old fuse. It's hugely reliable. If the diode fails and applies way more voltage than the open-circuit voltage of the cell is at that moment, a current high enough to blow the fuse will flow. What is your voltage? You could actually test applying 5V from a powerful lab supply to a fully charged cell and see how much current it takes. It will probably exceed your normal operating currents so that you can design in a fuse.

Designing the electronics to be super reliable is of course a good additional plus, to give one extra layer of protection. My view on the cell internal protection is that you should never design anything to rely on them, but they will 99% surely save the day if your design fails.

Proper protection only exists in quality cells, i.e., Sony, Panasonic, Sanyo, Samsung, LG and maybe a few others.

High power going into the cell is actually the most probable cause of thermal runaway and fire, not out-of-range voltage per se. And you can fuse against high currents. All proper cells are already fused, but I don't like the fact that it's self-resetting in a 18650, so add a normal fuse.

[Gyro]

Maybe you should change the thread title to 'Active diode' or similar to help with future searches. An Ideal diode would, in traditional terms, have zero drop, zero leakage etc. ....and be totally reliable. 

[digsys]

Maybe you should change the thread title to 'Active diode' or similar to help with future searches. An Ideal diode would, in traditional terms, have zero drop, zero leakage etc. ....and be totally reliable.     

Except that all the manufacturers refer to them as "ideal" diodes or "ideal" bridges. Irrespective of what the term "should (or shouldn't)" be, that name is now fixed,
and any personal changes will greatly reduce search results. As FET RDSon reduce with time, they will approach "ideal"

[Gyro]

Those damned Marketing people again! 

[digsys]

.. Those damned Marketing people again! ..

There is also the interpretation of the word "ideal". Ideal does NOT necessarily mean "PERFECT" !! and any technology / system would have a different "Ideal".
As an experiment - suggest to your wife / gf that you have a set of criteria that will make her "Ideal" !! (Set up an exit strategy before-hand though) :-)

[Gyro]

Are you kidding? I'm pedantic, not Mad! 

[CatalinaWOW]

Just checked the manufacturers reliability data for the first part on your list, and they suggest that the MTBF for the part is over 4000 years.  While all reliability numbers need to be taken with a grain of salt it doesn't seem that you should be overly concerned.  Their numbers are based on tests of thousands of devices under extreme operating conditions in which they observed zero failures.  The large sample sizes show that random weaknesses actually are fairly low probability.  The relatively short actual test times are extended using the activation energy methodology which can be problematic, but their assumptions are pretty conservative.

Interestingly, I found it difficult to find reliability info on FRED and TMBS diodes.  The only reference I found was an MTBF based on the activation energy methodology with no supporting data.  It was much lower than the number referenced above. 

The complexity of the ideal diode parts is actually relatively low.  Probably the total transistor count is measured in the low hundreds at most, probably much less so the die area is dominated by the FETs needs.  While it is more complex than a simple diode I would think that manufacturing methods would dominate complexity in evaluating the reliability of the part.  I am not surprised at a finding that the ideal diodes have a higher predicted MTBF than simpler parts.

The one area where the ideal diodes might be worse is in failure mode.  The available data makes no representation of how failures are allocated between modes.  Simple diodes may be more likely to fail open, but this is only a guess and might not be true in your application.

In summary, I am not sure you are meaningfully reducing risk by avoiding the ideal diodes.  If the power savings or other features of the part are attractive for your application I would say go for it.