Connect with us

Tantalum cap failures

Discussion in 'Electronic Design' started by [email protected], May 9, 2005.

Scroll to continue with content
  1. Guest

    We have a design in which we are using a parallel bank of 10 Arcotronic
    SFE solid tantalum capacitors, 47uF/63V, (case size 'D'), as a
    reservoir/filter for a 3-phase motor PWM supply opperating at nominally
    80kHz. Normal opperating supply voltage in our application is 28V.

    We have had problems with some capacitors becoming short circuit during
    high voltage testing, sometimes in conjunction with operation at limits
    of temperature, (all voltages and temperatures within the specified
    rating for the components). Sometimes the short circuit disappears on
    removal of the applied voltage, sometimes it results in disintegration
    of the offending component.

    These failures occur in around 2% of the capacitors, which means it
    occurs in around 20% of assembled units and this rate is too high.

    We have reduced the maximum voltage at which we test from 56V to 44V
    and also reduced the rate of change of supply voltage during
    transitions to around 1ms, to reduce the inrush current, but with no
    significant reduction in the failure rate.

    The bank of caps is there to sink/source short-term PWM current
    requirements, so it is not practical to introduce a series resistor on
    each capacitor, but Arcotronic specifically state that this measure is
    not required with these parts. Total peak ripple current is almost 20A
    and lasts for 10 to 20ms, with a maximum duty cycle of around 1%.
    Total steady-state ripple current is around 0.5A.

    Rated ripple current for the caps is 2.7A rms @20C,derating to about
    1.75A at our upper temperature. Given that the caps cannot be
    guaranteed to share equaly, we are exceeding the caps ripple rating,
    particularly at high temperature. However, since the issue of ripple
    current rating is purely one of capacitor heating, this is at such a
    low duty cycle, with such a low background level that this was not
    expected to be a problem. This view seams to be endorsed by the fact
    that failures never occur during the high currnet transients, but
    always occur during increased supply voltage.

    Arcotronics have been unable to throw any light on the problem and have
    stated that they consider our application to be reasonably benign, in
    terms of the stress placed on the capacitors.

    I wondered if anyone on here has any experience of this or similar
    problems and is able to offer a solution, or alternatively suggest an
    alternative part that would be more reliable whilst still meeting our
    basic requirements in terms of ESR and CV product?

    TIA for any help.

    Ted Wilson

    edward.wilsonatbaesystems.com
     
  2. On 9 May 2005 07:13:05 -0700, the renowned
    In similar situations, some makers recommend using capacitors derated
    to 1/3 or so of rated working voltage, so in your case about 100V.
    That may not be possible for you. Perhaps you have to look at using a
    different type of cap.


    Best regards,
    Spehro Pefhany
     
  3. John Miles

    John Miles Guest

    I don't understand how voltage ratings for tantalum caps are derived.
    Do they just ramp them up to 25V over 5 seconds, observe an absence of
    smoke for 5 more seconds, and stamp them "25V"?

    Why isn't this considered a big problem by the manufacturers, if not the
    users, of these parts? It's getting to be the most-F of all FAQs: "why
    are my tantalum caps dying?"

    -- jm
     
  4. Guest

    This particular range of caps only goes up to 63V. Also space is at a
    premium, so increasing rated voltage would probably be at the cost of
    capacitance.

    There is a 20 year shelf-life requirement on the product, so we can't
    use electrolytics. One way or another, our choices seem pretty
    limited.

    Ted
     
  5. Guest

    I take your point - the extent of Arcotronic's response has pretty much
    been simply to confirm that returned caps have indeed gone s/c.

    Ted
     
  6. John Larkin

    John Larkin Guest

    Tantalums hate AC ripple current. It causes local heating of tiny bits
    of the sintered pellet. If the contact media is MnO2, which standard
    solid tants are, the tantalum becomes fuel and the MnO2 becomes
    oxidizer and it explodes. It's an instant-ignition effect, not slow
    heating. It's very erratic from mfr to mfr, or even between lots.

    Replace them with aluminums, or niobiums, or with the new polymer
    tantalums, none of which has the same detonation mechanism. The
    polymer tants have very low esr, too.

    I'm selling VME modules to BAE, for a critical radar application, and
    we used all aluminum caps! I only use tants in slow RC-type circuits,
    never as power bypasses; I've learned that lesson!

    John
     
  7. On 9 May 2005 08:03:29 -0700, the renowned
    Ceramics are getting close. You can get 22uF/50V X5R from TDK, though
    they are expensive. 10uF/50V is more reasonably priced, but you'd need
    a pack of them to get the same capacitance.


    Best regards,
    Spehro Pefhany
     
  8. Paul Mathews

    Paul Mathews Guest

    http://www.speff.com

    Tantalum failure modes and mechanisms are well-known and
    well-documented. Among the root causes are mechanical stresses during
    manufacturing, handling, soldering, and handling after installation.
    You can reduce the probability of failures by specifying 'surge tested'
    parts. Some datasheets and some customers specify a minimum series
    resistance between low impedance power and Ta caps. Most designers who
    have experienced the failure problems design them out.
    Unfortunately, high value ceramic caps are also subject to spectacular
    failures, especially if working voltages are high enough to sustain
    arcs. The surface mount types are especially susceptible to cracking
    due to mechanical stresses in soldering and board handling. As others
    have suggested, other cap technologies are available, and it's possible
    that you can get away with less capacitance, particularly if you
    minimize inductance in the connection of the parts.
    Paul Mathews
     
  9. Joerg

    Joerg Guest

    Hello John,
    The recommended derating alone can make tantalums unattractive. See #5
    about low impedance apps which is in essence a high ripple situation:

    http://www.ncc-matsuo.co.jp/pdf_e/app_tan.pdf
    Same here. Another hard lesson many of us had to stomach was similar to
    the energy crisis. Suddenly many of the tants were "on allocation".
    Distributor-speak for "can't touch this", maybe unless you were the big
    shot executive of a fortune-10 company.

    Regards, Joerg
     
  10. Joerg

    Joerg Guest

    Hello Edward,
    Thing is, if you don't properly derate tantalums that are exposed to
    lots of ripple they'll keep failing. Sometimes they do that with
    pyrotechnic gusto.
    Check out whether the Vishay TVA ATOM series (formerly Sprague?)
    suffices. You will need to contact them as the data sheets are kind of
    skimpy. There may also be suitable mil spec versions as this stuff has
    to have long shelf life. Possibly John could suggest some.

    From my lab stock of quality electrolytics I have yet to see one that
    has aged out. When I restored an old Astor Radio from 1959 I checked the
    electrolytics for leakage and capacitance. As good as new, plays like a
    champ. I did see electrolytics that passed away after 30-40 years but
    mostly because some engineer had placed them right next to the rectifier
    tube.

    Regards, Joerg
     
  11. Joerg

    Joerg Guest

    Hello John,
    Unless this is confidential or they cost an arm and a leg: Which
    aluminum cap series did you use?

    Regards, Joerg
     
  12. Pooh Bear

    Pooh Bear Guest

    < snip >

    The issue of tantalum caps regularly appears here. Even used for supply
    decoupling they have an alarming tendency to expire ( usually with some
    combustion ).

    Derating the voltage seems to help.

    Have you checked the ripple current rating ? Remember that ripple current
    is higher than the DC load current.

    Have you checked the dV/dt - dI/dt capability of the caps ?

    I wouldn't touch tantalums with a bargepole personally. Is there any reason
    you don't want to use low ESR aluminium electros ?

    Graham
     
  13. Graham Holloway
    WPS/Accuphon Electronics
    (Tel/Fax 0(044)1233 662599)
    Solid Aluminium?

    SIC Safco or Rifa (wet) aluminium have some parts that are rated for a very
    long life.


    Graham Holloway
    WPS/Accuphon Electronics
     
  14. Guest

    We're using a 63V part at a nominal supply voltage of 28V, but would
    have used a 100V part, as suggested by Spehro, had they been available.

    The ripple current rating of these caps appears marginal, (2.7A rms per
    cap @ 20C), when assessed in terms of the peak ripple requirement of
    20A shared between the 10 caps, but the quiescent ripple current is so
    low, (less than 0.5A total), and the duty cycle of the peak requirement
    again is very low - less than1% maximum. Given that the issue with
    ripple current is capacitor heating, it was not anticipated that this
    would be a problem and none of the failures has ocurred during high
    ripple current events, but always in association with testing at higher
    voltages.

    So far as dV/dt and dI/dt ratings are concerned, I can find no figures
    for either on the data sheet, other than reference to charge/discharge
    at rated voltage through 0.5 ohms for 1E6 cycles. We have however
    modified our test regime, so that changes in supply voltage occur more
    slowly, (~1ms time constant), than was originally the case, It's
    worth noting however that some failures have occured when winding the
    supply voltage up to the new value on a bench power supply, so dV/dt
    was not an issue in these instances and all evidence points to the
    problem being voltage related, albeit possibly brought about by some
    other mechanism.

    We have traditionally been prevented from using electrolytics in the
    military environment, due partly to our 10 year shelf-life shelf-life
    requirement, (not 20 years as I stated earlier in the thread). I will
    have to check with the powers that be regarding the possibility of
    using electrolytics. Clearly, the current level of failures is
    unacceptable and we are going to have to do something to address it.

    Ted
     
  15. Mike Monett

    Mike Monett Guest

    wrote:

    [...]

    Have you looked at solid polymer? They seem to have much lower ESR than
    tantalum and better life than ordinary electrolytics. They also fail open
    circuit instead of short. Here's a Cornell Dubilier article from 2001
    that describes the construction and failure mechanisms:

    http://www.cornell-dubilier.com/tech/spa2.pdf

    All the major cap manufacturers supply them now, so there should be
    plenty of data on reliability and product lifetime.

    Mike Monett
     
  16. Have you considered a prescreen voltage test that blows the bad units
    up before they are attached to the boards? Perhaps you can design
    this test in cooperation with the maker, and get replacements for all
    the units that fail, since you are not exceeding their rated limits
    and would just be doing their quality control for them.
     
  17. John  Larkin

    John Larkin Guest

    Panasonic ECE series, mostly. Since aluminums tend to fail by drying
    out and slowly losing capacitance, it's easy in most cases to just
    grossly overkill on capacitance and voltage rating and buy reliability
    that way. VME modules allow about 0.6" of component height, so a
    surface-mount aluminum can use the height and provide a lot more CV
    product than a flat tantalum.

    We've had a number of products fail from tantalum shorts - they are
    the least reliable part we've ever used - but none from surface-mount
    aluminum elec failures.

    John
     
  18. Mark

    Mark Guest

    can you store the energy at a higher voltage using alum cap then use a
    regulator to create your desired output voltage.

    the engery is 1/2 C V^2 so you get better space efficeincy at higher
    voltages.

    In other words at a higher voltage you need less CV and CV is prop to
    space

    Mark
     
  19. Joerg

    Joerg Guest

    Hello John,
    Thanks for the info. We have used Panasonic caps a lot as well. No
    problems. I also like their ECH and ECHU series when film caps are
    called for. Oversizing in capacitance is indeed the thing to do.
    The only significant failures I ever saw from aluminum electrolytics was
    when people placed them too close to hot stuff, like processors,
    regulators or heat sinks. Even then they usually go gracefully, not 4th
    of July style like some tantalums.

    Regards, Joerg
     
  20. Guest

    Hello Mike

    I've had a look at solid polimers and, at first glance, they look very
    promising. However, their expected life is only 10 years or so, and we
    have a contractual requirement for not less than 25 years. (How
    realistic that really is is another matter, but that is what we are
    obliged to aim for).

    Ted
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-