Low ESR Tantalum Failure due to Power On- Design Fix?

Put inductance in series on power line. I don't have numbers nor experience with this problem. But series inductance (and not on ground that is shared by both power and signals) would blunt the initial power inrush. Ferrite bead would be no where near sufficient inductance.

Meanwhile, a typically design used a large aluminum electrolytic and a smaller low ESR tantalum in parallel.

Manganese dioxide tantalums are literally bombs. A bit of dv/dt pushes current into them, a tiny spec of anode gets hot, and the MnO2 is the oxidizer and the tantalum is the fuel. They don't just explode from available electrical energy, they chemically ignite.

These caps are *not* reliable as supply bypasses. Replace them with ceramics, polymer tantalums, or regular or polymer aluminum caps. I think some of the new niobium caps may be ignition-free, but I have no experience there.

Slow voltage rampup would help, but is a hassle. Derating the tantalums by at least 2:1 on voltage may help some, but isn't a guaranteed fix.

This is unpredictable between manufacturers and even batches.

John

So .............. don't use tantalums!

Anything that can't take dv/dt isnt really a capacitor in my book.

It's not like there are no alternatives-- good low ESR aluminum electrolytics are cheap and not prone to shorting. You might need a small ceramic in parallel to match the tantalum's HF performance. If you need long life, there are high temp electrolytics with more hours possible.

Could be worse, there were those old wet-slug tantalums where if the seal failed, you got some really nasty acid spilling out all over.

We've had similar problems. Here are the possible causes (the last one should really piss you off):

A recent batch of AVX 68uF/20V caps were subjected to only 100mA of startup current (I=CdV/dt), small amounts of load-induced ripple current, and had a max applied voltage of 12.3V (no overshoot or undershoot). We had about a

We had convinced ourselves that it was a moisture problem after reading several documents -- including a study from NASA. There's one website that show a SOT223 part blown of its pads due to outgassing from a moisture-rich cap, and they did an experiment where they covered a board with some flour (next to a cap) and then put the board through reflow. There was a jet trail showing the path of the outgassing cap. Upon close inspection they were able to see the bulge and crack where the gas came from.

Our failing parts were built in early 2006 but reflowed only a few weeks ago so it seemed likely that could have had time to absorb enough moisture to cause a problem. AVX told us (before seeing a sample from the "bad" reels) that they now offer caps packed with a desiccant. They wouldn't directly admit that there is a potential moisture problem but why else would they offer them with desiccants now?

However, when we finally did send AVX some parts from the failing reels they responded (in writing), "We did not manufacture these capacitors." Our jaws dropped. The parts were purchased from a broker by our contract manufacturer. That will never happen again.

Anyway, I'm convinced that if you don't exceed the rated ripple current, max rated voltage, and have dry non-counterfeit caps that these tantalums should be fine for supply bypassing -- both on the input and output of switchers -- with none of this bullcrap derating by 50% to 70% in voltage.

Bob

I posted on here some time ago about a similar problem we were having with around 2% of Arcotronics SFE solid tantalums failing short circuit on an aircraft equipment. (Since there were 9 in parallel, this equated to around

Our solution was to replace the tants with multi-layer polymers and we ended up using Evox Rifa's MDC part. We were only able to accommodate about a quarter of the capacitance we had with the tants, but the greatly reduced esr meant that the MDCs were actually better in our application.

After a fire started by a tantalum capacitor did some serious damage, the insurance and legal people convinced Tektronix (over 25 years ago) to stop using tantalum capacitors in most circumstances. Many others did the same.

Like John L. said, these are incendiary devices. Maybe you can come up with a cute turn-on circuit to limit inrush, but it would be better if you have such control over the design to eliminate them before the boards are assembled.

-frank

That's for sure. A client sent me his breadboard to evaluate, but said there's a short somewhere on the VCC line.

So I power it up with current limit set to 100mA, while I ohm the trace drops.

With my face right in there... kaboom! A Tantalum had been installed backwards :-(

Thankfully I was wearing magnifiers which protected my eyes.

...Jim Thompson

If you insist on using tantalums, I think your best option will be to use a combination of (4) and something like tantalums with built-in fuses. One of the problems with tantalums is that they can short to a low, but not zero, resistance and then get really hot if there's enough available current. I understand some companies have a policy that you will NOT design in a tantalum unless there is some value of resistance between it and the supply. You're lucky that your supply is only 18 or so volts; you can get OSCON caps which are very good (albeit a bit pricey) that will handle that. I've worked on some boards that are supplied with 24V, and would not try to put a 25V cap on the line as a bypass.

You could consider multilayer ceramics these days, too. You might not want to spend the money on enough ceramics to get the required bulk capacitance, but some ceramics in parallel with aluminums or regular esr tantalums might give you the performance you are looking for.

My experience is that tantalums are good for low voltage lines, but are dangerous to try to use on high voltages, anything over about 10 volts. Supposedly you'll get better reliability with tantalums that are lower capacitance for a given case size; reliability of cutting-edge capacitance-per-unit-volume is questionable. I've also seen some emperical evidence that low ESR units are also more prone to failure than standard ones.

I'm pretty sure CALCE has some data on tantalum failures, but I'm not sure how much is available without a membership.

Hope this empirical stuff is helpful. Please post a followup sometime when you're pretty sure you have the problem solved. That would be a useful contribution to the group.

Cheers, Tom

Thanks for all your input. Here are the possible fixes you'll gave:

Here is my response: First off I don't have the caps in backwards. I am confident that my suppliers have given us the manufactured caps... I have seen mechanical issues and possible failure to moisture problems. But I don't think this problem is related to these issues. I don't think derating will help because the large the Tantalum the more prone it is to problems...

I found the following precaution on Sanyo OS-CON caps:

I did look to much into these OS-CON caps, but basically the same issue would happen with any low ESR cap...

Basically the inrush current with a low ESR cap is very large. In the example they give 222 amps! Basically you can take the input voltage and divide it by ESR value. In my case 18.5 volts / 2 ohms = 9.25 amps! I tested with a large series inductor (220 uH) and found that it defiantly removes the large dv/dt jumps.

There are about another 30 low ESR Tantalums on the same electronic setup that have no problems. They are not directly hot plugged and run on lower voltages. I don't think that all low ESR Tantalums are bad, but defiantly they should never be put on an unprotected input. On existing electronics three of the Tantalums are used to regulate line in and two more are isolated by 10 uH inductor (ferrite bead).

For a rework of existing boards: I plan to remove the three Tantalums and replace with probably a long life (or high temp) aluminum in parallel with a ceramic. On the other two caps I plan to increase the inductance to about 100 uH to 200 uH.

In the future I plan to redesign the board with a large inductor (100uH-200uH) in series with input power. I will probably also add the fly back diode for safety.

At least that is the plan for now...

Thanks again for all the input,

Matthew

Consider using a parallel resistor in place of a flyback diode across the inductor to damp the overshoot. It will change the ramp up to a step plus a ramp, but won't allow an instantaneous capacitor dump through an input short.

Hot s*it. Really good evidence that our worst nightmare is fully here. Cheap Chinese knock-offs that look enough to fool anything short of knowledgeable Destructive Physical Analysis (DPA). We need to get the message out, to start raising some hell in DC.

Correctly made solid tantalums should not outgas during wave soldering.

Firstly, I would NEVER use solid or dipped tantalums in any power supply circuitry. The only reliable tantalums to use in these situations are hermetic sealed types. Of course these too are prone to failure with excess voltage peaks but not nearly so badly as the solid types.

My suggestion is that the tantalums (if they must be used for low ESR) should be hermetic types and should always be permananently connected at the output of the front end linear supply, not on the DC input of the DC - DC stage following. The DC - DC stage following the linear stage should have only low ESR aluminium electros on the input.