Ouch. The Arduino GSM shield has a pretty serious design flaw, with its capacitors.

Tantalum is a really misunderstood capacitor. Well, all capacitors are misunderstood, but that’s a subject for another post. I ran across this post on the Arduino forums on the Arduino GSM shield. In the post, ddewaele, reports that the shield blew up, catching fire. At first some might think it was due to abuse by the user. While it is possible that reversing the polarity or applying over-voltage could cause a catastrophic failure, it is also possible that the user doing nothing wrong could result in the same failure mode!

Wait, what? So what gives? Well, there’s two things to understand. First, Tantalum doesn’t explode. It takes almost 2000°C before Tantalum metal will ignite. Okay, so if Tantalum doesn’t explode what is ddawaele seeing? It’s the cathode material, Manganese Dioxide, (MnO2) that is exploding…

Why don’t capacitor manufacturers fix the problem?

If you buy capacitors from any Tier 1 supplier (aka, not some no-name in Asia), they are testing every single capacitor they ship multiple times. So how do MnO2s that eventually fail escape? Because the damage isn’t done until the parts go through the reflow oven.

Tantalum Capacitor Construction

Even though terms like anode and cathode aren’t typically associated with capacitors, they are useful here. A tantalum has 3 major components: the anode (Tantalum), the dielectric (Tantalum Pentoxide), and the cathode (Manganese Dioxide). The anode is composed of a bunch of very fine tantalum particles sintered together. After being fired in an oven at about 1500°C, the anode pellet is dipped into an electrolyte solution and a formation voltage is applied. That voltage grows the dielectric layer on top of the pure tantalum. It is important to understand that the dielectric layer is very thin and very fragile. Next, the oxidized anode pellet is dipped into an manganese nitrite solution then fired in an oven at temperatures over 200°C which forms layers of Manganese Dioxide (MnO2) to complete the cathode layer. So now you have the dielectric layer between two conductive “plates.”

Ideally the dielectric layer would be a perfect insulator. However, if weaknesses form in the dielectric layer, relatively large amounts of current will flow through it. The cool thing is that when current flows through these weaknesses, it causes the MnO2 to heat up changing it into Mn2O3, which isn’t very conductive. In other words, leakage sites in the dielectric are “healed” by the localized areas of the cathode becoming non-conductive.

Capacitor manufacturers work very hard to keep the dielectric layer defect-free during production, but some defects develop. So, all tantalum capacitors go through various screenings and aging (sometimes beyond 24 hours) to heal-up the dielectric layer. So that’s good right? Parts ship on the reel all healed up and ready to go.

Why do they still fail?

Unfortunately, when a part goes through a reflow oven, the part is subjected to extreme temperature changes. As such the very rigid metal tantalum and the very solid MnO2 put a ton of mechanical stress of the very fragile Tantalum Pentoxide between them, causing it to sheer and form cracks. Those cracks allow leakage currents to flow through the dielectric.

Oh. So when they get soldered to the board, they get damaged again.

This previously mentioned healing property comes with a downside. If too much current flows through the dielectric that means excessive current is flowing through the MnOcathode layer.  It becomes so hot, it isn’t able to transition into Mn2O3 fast enough. So the MnO2 gets really hot and ignites the surrounding oxygen. Boom. You get the “exploding” failure associated with “Tantalums.”

BUT… notice the subtle point.  It isn’t the Tantalum’s fault!

How do you avoid this failure?

There’s really three ways to avoid this kind of failure:

  1. Reduce the applied voltage. Common belief is that this failure is cause by surge currents. This isn’t really the case. The failure is caused by voltage activating a weakness in the dielectric. Those weakness are formed during reflow. Capacitor manufacturers strike balance between growing the dielectric layer thick enough to make it robust while still keeping it thin enough to get the desired capacitance. They grow the dielectric knowing that these weaknesses will form during reflow. By applying less than the rated voltage, it is less likely you’ll activate a weakened area. So surge current doesn’t cause the problem, however, it does contribute once a breakdown occurs.
  2. Allow the dielectric to heal by limiting the current, so that catastrophic breakdown can’t occur.  It is actually possible to “heal” or “proof” the dielectric after reflow. Unfortunately it isn’t always practical to power up a capacitor while limiting the current, to allow the capacitor to heal.
  3. Don’t use Manganese Dioxide (MnO2) While #1 and #2 might sound good on paper, in the real world, they might not really help. Since MnO2 contributes to the destruction of the dielectric during soldering, one option is:  don’t use it!

What if you didn’t use MnO2?

Manganese Dioxide-based Tantalum capacitors are about a 50 year old technology. Around the year 2000, a conductive polymer material was introduced to the material science community. This organic polymer (PEDOT) can be used in place of the Manganese Dioxide as the cathode material. Polymer capacitors have 3 distinct advantages over MnO2:

  1. Polymer cathode layers put less stress on the dielectric layer during reflow, meaning less chance for a failure. In fact, if you derate the applied voltage on a Polymer by 10% it is still less likely to fail than a MnO2 derated by 50%!
  2. The polymer actually consumes oxygen when the localized heating occurs. You still get the healing property we saw with MnO2, however when the Polymer gets hot, it oxidizes. It actually consumes available oxygen as it becomes non-conductive. So it isn’t possible for the Polymer to Ignite!
  3. The Polymer material is highly conductive, compared to Manganese Dioxide which is relatively resistive. So it is possible to get orders of magnitude less ESR with a Tantalum-Polymer that has the same Voltage/Capacitance/Case Size as a MnO2.

An important note here is that the phrase “Polymer Capacitors” get used a lot in the capacitor industry.  Remember that the dielectric is still Tantalum Pentoxide.  It’s the cathode material I am talking about in this post.

Coming back to the GSM Shield

So here’s the design flaw of the Arduino GSM shield, which catches fire:  6.3V rated MnO2 was used on a 5V rail. It is well known in the capacitor industry that this will result in failures. Unfortunately for the Arduino Team, and their customers, these failures will be explosive (and not just in the number of blog posts about it!) There’s really only two options for the board:

  1. Replace the 6.3V Rated Part with a 10V Rated part. This may not be possible given the capacitance value.
  2. Replace the part with a Polymer. Again, limited by the cap value (the largest 6.3V Polymer I know of is 1500µF).
In the mean time, if you buy a GSM shield, “burn in it” a few times. Cycle power on it a few times, to stress the cap out a little bit. If it doesn’t fail, it’ll heal-up and be more robust. On the other hand, keep anything flammable away from it just in case it hasn’t completely healed yet.

Disclaimer: James is a Senior Technical Expert for Technology and Applications at KEMET Electronics, a capacitor manufacturer. The content of this post are his and in no way reflects opinions of his employer.

Long comments, URLs, and code tend to get flagged for spam moderation. No need to resubmit.

ALL comments submitted with fake or throw-away services are deleted, regardless of content.

Don't be a dweeb.

Leave a comment

27 thoughts on “Ouch. The Arduino GSM shield has a pretty serious design flaw, with its capacitors.

  1. This is a great explanation, but I’m unclear on one point. If the cap is being used on a rail that is at a lower voltage than its rating, why is this a common cause of failure? And why would using a cap with an even higher rating on the same rail prevent the problem?

    • Most of the article is dedicated to explaining this aspect. In summary, it’s related to the thickness of the dielectric. Higher rated voltages have thicker dielectrics. So when cracks for during reflow, it takes a higher applied voltage to cause a breakdown.

      • I understand that. And I understand the part of the article that states, under “How do you avoid this failure?”: “Reduce the applied voltage…By applying less than the rated voltage, it is less likely you’ll activate a weakened area.”

        By using a 6.3V-rated cap on a 5V rail, it seems the designers did take care to apply less than the rated voltage. Is the problem that it’s not “less enough”, and that a 10V-rated cap will provide a sufficient safety margin that a 6.3V-rated cap does not?

        Thanks for your response.

        • Yes, it is the case it wasn’t “less enough.” The confusion is rooted in the definition of “rated voltage.” All components fail eventually due to two reasons: A) Infant Mortality (Design or Manufacturing Flaws) or B) Wear-out (material fatigue over time).

          Rated voltage is based on the wear-out mechanisms and actually throw out infant mortality causes from their results. So “rated voltage” is determined using an accelerated life test, intended to wear-out the component. For Tantalum Pentoxide, this wear-out is on the order of hundreds if not thousands of years.

          Since power-on failures, like the one seen in this shield, are considered infant mortality failures. (Note that infant mortality is *not* time related.) So a 6.3V Tantalum-MnO2 cap on a 5V rail will have a very long wear-out time, but it may result in an infant mortality failure.

          Lastly, it is very important to note that different capacitor types benefit from voltage de-rating for different reasons.