3
\$\begingroup\$

I'm on my first large board design, and it's only begun to dawn on me how infuriating capacitors can be to design with.

Exhibit A: I need 100uF for a simple circuit to keep my FPGA on for 1ms after power is cut. This is hard to find in a cost-effective ceramic, but because it is just being used for energy storage I can deal with tantalum. So I start looking for a 100uF tantalum that's good to 25V, should be easy right?

It probably is, but I've noticed that all tantalums quote endurances between 2000 and 10000 hours. This seems terrible, that's barely a year of continuous operation. These numbers are quoted at Tmax, so how do I derate for a lower temperature? For example, at 55 °C I imagine the lifetime is more reasonable, but I can't find any resources on this. Is there any way to know? Is it a crapshoot? If so, isn't designing with tantalum caps incredibly risky in something that expects to be used on the order of years? How do you deal with this conundrum?


Edit I've since come across two resources that helped my comfort level with tantalum lifetimes, which I'll add here to help anyone else who comes across this question. AVX has a useful primer on tantalum derating that goes over the MTBF calculations:

Voltage Derating Rules for Solid Tantalum and Niobium Capacitors

Better still, Vishay has a very simple "FIT calculator" that does easy MTBF calculations for all of their tantamlum families. Playing around with this was super useful for framing typical MTBF values in my mind:

Vishay FIT Calculator

Bizarrely, Kemet only has such a calculator for aluminums, and while I can find references online to an AVX calculator I can't find it on their site as of Sept 2017.

\$\endgroup\$
  • 2
    \$\begingroup\$ read Arrhenius Law then apply MTBF factor assuming defect free \$\endgroup\$ – Sunnyskyguy EE75 Aug 2 '17 at 20:09
  • 1
    \$\begingroup\$ For wet electrolytics the endurance is also specified at a high temperature (e.g. 2000h @ 105°C), and as a rule of thumb can be derated to twice the endurance at every 10°C lower (e.g. 64 000h @ 55°C). Maybe a similar principle holds for tantalums? \$\endgroup\$ – marcelm Aug 2 '17 at 20:09
  • 1
    \$\begingroup\$ Things will only get better at lower temperatures. How fast they get better depends on the activation energies for the particular wearout processes. See if you can turn up some research into tantalum lifetimes against temperature, as even a small difference from the energies for other capacitors could make a significant difference. But 2x per 10C will be a good start. \$\endgroup\$ – Neil_UK Aug 2 '17 at 20:14
  • \$\begingroup\$ How did you arrive at C value? C=Ic 5ms * 6 (for 10% sag? or * 3 for 20% or * 1 for 60% sag =T) Also what ESR*C=T did you assume >100us? (g.p.) or < 10us?? (low ESR) \$\endgroup\$ – Sunnyskyguy EE75 Aug 2 '17 at 20:17
  • \$\begingroup\$ @TonyStewart.EEsince'75, the capacitor is at the input of a very robust dual-channel switching regulator that provides both 1v2 and 3v3 to the FPGA and can provide Vout up to Vin (LTC3622, can't recommend it enough). Between LTSpice simulation and C ~= i * delT / delV I arrived at 100uF to buy >1ms. \$\endgroup\$ – jalalipop Aug 3 '17 at 14:29
5
\$\begingroup\$

How do I derate for a lower temperature? For example, at 55 °C I imagine the lifetime is more reasonable, but I can't find any resources on this. Is there any way to know? Is it a crapshoot?

If the manufacturer doesn't provide the information then you could test it yourself or email the manufacturer and see if they could provide more info. There are two reasons why they didn't:

  • One is the lifetime for a cap at room temperature is significantly longer than a high temperature capacitor and the testing time becomes prohibitively long.

  • The second reason is the degradation happens faster at high temperatures so they provide the spec at the high temperatures to provide and upper bound and to show that the lifetime will be degraded if you run them at +50C.

Another thing to note is a manufacturer can never provide an exact time of failure only a mean time between failure (MBTF), so if your application is critical, testing and buying high reliability parts will be necessary

I've never had a problem with tantalum capacitors running in products at 40C for years. And the specifications probably degrade slightly from the original specs, but in my application I use them as power filter capacitors and my design is tolerant to slight degradation of a few percents.

If so, isn't designing with tantalum caps incredibly risky in something that expects to be used on the order of years? How do you deal with this conundrum?

1) Don't use the capacitors for applications that are dependent on the capacitors value.
2) Don't run them at high temperature.
3) Use the right kind of tanalum capacitor, the electrolyte makes the difference.

4) Know what the faiure modes are and mitigate them:

The life time, service life, load life or useful life of tantalum electrolytic capacitors depends entirely on the electrolyte used:

  • Those using liquid electrolytes do not have a life time specification. (When hermetically sealed)
  • Those using manganese dioxide electrolytes do not have a life time specification.
  • Those using polymer electrolytes do have a life time specification.

The polymer electrolyte have a small deterioration of conductivity by a thermal degradation mechanism of the conductive polymer. The electrical conductivity decreased, as a function of time, in agreement with a granular metal type structure, in which aging is due to the shrinking of the conductive polymer grains.[62] The life time of polymer electrolytic capacitors is specified in similar terms to the non-solid electrolytic caps, but its life time calculation follows other rules which lead to much longer operational life times.

The extremely thin oxide film of a tantalum electrolytic capacitor, the dielectric layer, must be formed in an amorphous structure. Changing the amorphous structure into a crystallized structure is reported to increase the conductivity by 1000 times, combined with an enlargement of the oxide volume.[12] The field crystallization followed by a dielectric breakdown is characterized by a sudden rise in leakage current within a few milliseconds, from nanoamp magnitude to amp magnitude in low-impedance circuits. Increasing current flow can accelerate in an "avalanche effect" and rapidly spread through the metal/oxide. This can result in various degrees of destruction from rather small, burned areas on the oxide to zigzag burned streaks covering large areas of the pellet or complete oxidation of the metal.[7] If the current source is unlimited a field crystallization may cause a capacitor short circuit. In this circumstance, the failure can be catastrophic if there is nothing to limit the available current, as the series resistance of the capacitor can become very low. If the current is limited in tantalum electrolytic capacitors with solid MnO2 electrolyte, a self-healing process can take place, reducing MnO2 into insulating Mn2O3

Impurities, tiny mechanical damages, or imperfections in the dielectric can affect the structure, changing it from amorphous to crystalline structure and thus lowering the dielectric strength. The purity of the tantalum powder is one of the most important parameters for defining its risk of crystallization. Since the mid-1980s, manufactured tantalum powders have exhibited an increase in purity.

Surge currents after soldering-induced stresses may start crystallization, leading to insulation breakdown.[69] The only way to avoid catastrophic failures is to limit the current which can flow from the source in order to reduce the breakdown to a limited area. Current flowing through the crystallized area causes heating in the manganese dioxide cathode near the fault. At increased temperatures a chemical reaction then reduces the surrounding conductive manganese dioxide to the insulating manganese(III) oxide (Mn2O3) and insulates the crystallized oxide in the tantalum oxide layer, stopping local current flow.[7][66] Failure avoidance

Solid tantalum capacitors with crystallization are most likely to fail at power-on.[70] It is believed that the voltage across the dielectric layer is the trigger mechanism for the breakdown and that the switch-on current pushes the collapse to a catastrophic failure. To prevent such sudden failures, manufacturers recommend:[12][66][71]

50% application voltage derating against rated voltage
using a series resistance of 3 Ω/V or
using of circuits with slow power-up modes (soft-start circuits).

Source Wikipedia Tantalum Capacitors

\$\endgroup\$
  • \$\begingroup\$ Okay, looks like I still have reading to do. Much of this I'm fine with: A. The exact cap value is unimportant and extremely derated for the application. B. The application is continuous operation, so at least the capacitor would only be subject to a couple power-on cycles. C. Temps are around 55 °C worst-case. Definitely need to research tantalums further though, thank you! \$\endgroup\$ – jalalipop Aug 3 '17 at 14:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.