Is the bad reputation of tantalums justified?

[AndyC_772]

I read 'inexperienced' rather than 'awesome'. I thought tantalum caps were perfectly reliable too, right up until the point I came across a design in which they simply weren't.

[KJDS]

I shipped about 1000 units out that all tested fine. They were integrated by the customer and had a few days of testing there. 35V devices on an 8V rail. Unfortunately they were in backwards, and because they were so heavily derated took a couple of weeks to fail, by then they were spread over about 300 sites across Europe. Oops.

Mostly they're fine, but it's common to open up an old instrument and find one that's changed colour on a board with dozens on. Easy enough to spot so they're an easy repair. Never known one die spectacularly unless abused.

[DimitriP]

What I find interesting is on Digikey's web site in the parametric search for capacitors, Tantalum capacitors have selection for reliability and another for failure rate .

Looking at a list of 10uF 50V prices range from  $3.75 to  $50.68 for qty 1 !

[G0HZU]

...So I speak for over 100 engineers across maybe 30 years of designing high end comms products. The bottom line is that tants are high performance components with excellent reliability over time and temperature etc provided they are used correctly
...
In a bad design they can fail spectacularly the first time the unit is switched on.

Okay, so how do you use them correctly, or was the purpose of your post - like Pjotr's earlier - merely to tell us how awesome you are?

The purpose of my post was merely to provide a real world example where engineers like myself have been successfully designing with tants for decades and we still use them in our designs today.  We would stop using them in new designs if we thought the technical risk was too great.

Okay, so how do you use them correctly

If you are in doubt I would recommend reading the datasheet carefully and/or speaking to reps from the manufacturer to discuss specific design requirements. eg to see if tants are suitable for a particular circuit design.

[tom66]

Little reason to use them now.

- Ceramics have high capacitance figures.
- Polymer electrolytics for where ceramics lack.
- Ordinary electrolytics when treated well last a long time.
- Modern LDOs are perfectly stable with zero ESR, as are many switching regs (and ceramics can offer better transient response with near zero ESR**)
- No need to worry about voltage rating for ceramics as they will be perfectly tolerant of overvoltage. Only concern is capacitance rolloff.

** D-CAP2 and other such hysteretic switchers excepted: due to how the internal ripple reference is generated lower output ESR causes worse transient response. ask me how I know.

I'm sure for some applications e.g. high temperature ultra high reliability they may still be desired but in the electronics I work on, I haven't seen a tantalum used for 10 years.

[Pjotr]

Agree with that Tom. But high value ceramics have also their peculiarities to take care of. Their value fades over time so you have to count that in for their lifetime. And they are prone to mechanical stress, especially the >= 1808 sizes can have issues with that and thermal cycling of the boards. More rugged ones with a polymer strain relief barrier are also more expensive.

[coppice]

Tantaliums by themselves are very reliable and long lasting, provided they are properly used. Engineers not understanding and/or neglecting their properties are to be blamed if they fail.

Or, really? Then how would you explain the ~0.1% failure rate I observed in tantalum caps coming right off the reel, *before* soldering? Can't really blame *me* for that, can you?

Be careful before you call others incompetent.

How long had those been in storage before coming off the reel? 0.1% sounds a high failure rate for freshly made parts. For parts that have been in storage for a while, whiskers shorting the plates might account for 0.1% failures.

[mrpackethead]

I dont' like using tantalums because So much tantalum is mined by children under very dangerous circumstances.

[wraper]

Tantaliums by themselves are very reliable and long lasting, provided they are properly used. Engineers not understanding and/or neglecting their properties are to be blamed if they fail.

Or, really? Then how would you explain the ~0.1% failure rate I observed in tantalum caps coming right off the reel, *before* soldering? Can't really blame *me* for that, can you?

Be careful before you call others incompetent.

How long had those been in storage before coming off the reel? 0.1% sounds a high failure rate for freshly made parts. For parts that have been in storage for a while, whiskers shorting the plates might account for 0.1% failures.

Or ordered from some shady supplier, so might be counterfeits as well.

[coppice]

I dont' like using tantalums because So much tantalum is mined by children under very dangerous circumstances.

Will their lives somehow be improved by trashing the market value of the commodity which appears to be their best choice for an income right now?

[MagicSmoker]

I read 'inexperienced' rather than 'awesome'. I thought tantalum caps were perfectly reliable too, right up until the point I came across a design in which they simply weren't.

Yes, I'm beginning to suspect the same thing - that those who have not had any problems with tantalums have simply not come anywhere close to any of their datasheet limits. Don't get me wrong - I always use components at less than datasheet-claimed specifications, but at some point derating goes from "prudent" to "absurd". I mean, if I have to derate working voltage, ripple current and temperature on tantalum caps by >50% to keep them from blowing up (and worry about the rate of change of current and voltage, to boot) then why the hell should I select them in the first place? I can suffer the same voltage/temperature restrictions with a high-K MLCC (though with entirely different consequences) and still don't have to worry about too much ripple current (within reason, of course)!

[AndyC_772]

The design in question was, IIRC, using a "35V" capacitor on a 20V rail.

I wonder whether any cap manufacturer has ever been successfully sued over their misleading ratings? The idea that a design engineer should have to implicitly know and understand that the data sheet specs are garbage, doesn't sit well with me at all. The manufacturer is in a much better position to know what voltage a component will really withstand, over time, temperature and a large population of parts. Not me.

[John Coloccia]

Read the datasheet. 

[wraper]

I mean, if I have to derate working voltage, ripple current and temperature on tantalum caps by >50% to keep them from blowing up (and worry about the rate of change of current and voltage, to boot) then why the hell should I select them in the first place?

That exactly what manufacturers suggest depending on the conditions.
http://www.kemet.com/Lists/filestore/Derating%20Guidelings%20for%20Tantalum%202011%20(3).pdf
http://web.arrownac.com/sites/default/files/pdfs/voltage%20derating%20rules%20.pdf
http://www.kemet.com/Lists/TechnicalArticles/Attachments/164/2004%20CARTS-Europe%20-%20Derating%20Differences.pdf

So, why do I need a 50V rated capacitor for my 24V power rail? The
answer is that you can indeed use a surface mount tantalum capacitor
at or near its rated voltage with the understanding that the failure rate at
this voltage condition will typically be between 0.1 and 1.0% per one
thousand hours of operation. In addition, the user can expect to see
increases in initial power on failure rates following board mounting.
Most applications cannot accept this rate of failure however. So, to
improve the reliability of the device, KEMET recommends designers
follow the derating guidelines outlined in this module

For the voltage, because the exponential factor is 17, the failure rate
multiplier increases exponentially as the voltage exceeds 60%. It still
has a near unity factor at 50% rated voltage, but the failure rate doubles
at 60%, jumping to a multiplier of 14.7 at 70%, 134 at 80%, and close to
1000 at 90%. For this capacitor type, exceeding the recommended
derating voltage can lead to very high failure rates.

[MagicSmoker]

...
Or ordered from some shady supplier, so might be counterfeits as well.

If by shady supplier you mean DigiKey, well, guess I'm guilty.

I mean, if I have to derate working voltage, ripple current and temperature on tantalum caps by >50% to keep them from blowing up (and worry about the rate of change of current and voltage, to boot) then why the hell should I select them in the first place?

That exactly what manufacturers suggest depending on the conditions.

Right... so I guess you didn't pick up that I was basically saying that once you do all the necessary derating to ensure a tantalum cap can survive such a stressful application as, say, board-level supply bypassing  that tantalum is no longer even remotely competitive with al. electrolytic and/or MLCC types?

[wraper]

Right... so I guess you didn't pick up that I was basically saying that once you do all the necessary derating to ensure a tantalum cap can survive such a stressful application as, say, board-level supply bypassing  that tantalum is no longer even remotely competitive with al. electrolytic and/or MLCC types?

Well, if you use MLCC of comparable capacitance without significant voltage derating, you'll find that there is not so much of the rated capacitance left. MLCC capacitors are prone to become shorted too. So, your reasoning about being completely not competitive is quiet questionable. Also if you need some thin boards, electrolytic capacitors are not any good and have own downsides.

[AndyC_772]

Here's quite a good presentation comparing different capacitor types for use in power applications:

http://www.kemet.com/Lists/TechnicalArticles/Attachments/5/Avnet2012PowerForum_CapacitorsSelection.pdf

[MagicSmoker]

Well, if you use MLCC of comparable capacitance without significant voltage derating, you'll find that there is not so much of the rated capacitance left.

This really only applies to Y5V and Z5U dielectrics, which can lose as much as 60% of their capacitance when 100% of rated DC voltage is applied; the capacitance of X7R (and similar dielectrics) only declines by about 10% under the same operating conditions.

Now one problem with Y5V and Z5U MLCC capacitors that you didn't point out is their distressingly high piezoelectric effect but that, again, mainly applies to the highest-k Y and Z dielectrics, is much less of a problem with the X series dielectrics, and even less so with NP0/C0G types.

MLCC capacitors are prone to become shorted too.

One common cause for an MLCC to fail short is flexure, especially if lead-free solder is used (something I thankfully don't need to use). Aside from using leaded solder, there are usually "soft-termination" versions of MLCCs which are much more tolerant of flexure, and there are even safety-rated versions that are guaranteed to open like a fuse if they fail short for any reason. Conversely, the only choice you have with tantalum failure modes is whether it just blackens the board from overheating or it also turns into a little flamethrower.

Now, all that said, I mainly design stuff for industrial and transportation applications, so a very hostile operating environment with plenty of temperature swings, vibration and impact shock, etc., yet I have never had a single MLCC fail on me and by now I've probably used about a million of them.

So, your reasoning about being completely not competitive is quiet questionable.

Question my reasoning all you like; matters not to me.

Also if you need some thin boards, electrolytic capacitors are not any good and have own downsides.

Well, I'll grant you that one advantage for tantalum: they do come in much lower height packages for an equivalent CV rating. They do tend to take up a lot more board area, however, to achieve a usable confluence of C, V and I_rms.

[ez24]

Here's quite a good presentation comparing different capacitor types for use in power applications:

http://www.kemet.com/Lists/TechnicalArticles/Attachments/5/Avnet2012PowerForum_CapacitorsSelection.pdf

Added this to

https://www.eevblog.com/forum/beginners/electronics-primers-course-material-and-books/

thanks

[splin]

Well, if you use MLCC of comparable capacitance without significant voltage derating, you'll find that there is not so much of the rated capacitance left.

This really only applies to Y5V and Z5U dielectrics, which can lose as much as 60% of their capacitance when 100% of rated DC voltage is applied; the capacitance of X7R (and similar dielectrics) only declines by about 10% under the same operating conditions.

X5R & X7R can also have terrible voltage coefficients - there are plenty of graphs showing 70% loss at rated voltage and more than 50% of their capacitance at 50% rated voltage. It depends on the film thickness - the smaller the part is physically for a given rated capacitance the worse - for examples see page 5 of:

http://www.yageo.com/exep/pages/download/literatures/High%20Capacitance%20MLCCs_2012.pdf

You can reduce the problem by choosing larger parts with larger case sizes or paralleling lower capacity parts with the same size but it doesn't come for free - the larger parts are more expensive.

Worse, the loss of capacitance increases with time so it might lose another 5 to 15% over the course of a few years at room temperature and considerably worse at higher temperatures. The ageing rate is also dependant on the DC bias voltage. Interestingly Y5V is so bad that Kemet state:

"This points out why the Y5V should not be used in systems that are intended for long life (>3 years) applications".

Some information on ageing can be found here:

https://books.google.co.uk/books?id=tdx7lS4fO9wC&pg=PA418&lpg=PA418&dq=x7r+ageing+dc+bias&source=bl&ots=OZm-1hMU3W&sig=GbPCbe62Hq4GUsKuZpcpkFaHBGs&hl=en&sa=X&redir_esc=y#v=onepage&q=x7r%20ageing%20dc%20bias&f=false

[MagicSmoker]

....
X5R & X7R can also have terrible voltage coefficients - there are plenty of graphs showing 70% loss at rated voltage and more than 50% of their capacitance at 50% rated voltage. It depends on the film thickness - the smaller the part is physically for a given rated capacitance the worse - for examples see page 5 of:

http://www.yageo.com/exep/pages/download/literatures/High%20Capacitance%20MLCCs_2012.pdf
...

Well, I stand corrected. I don't use Yageo parts, nor do I have the pressing need to cram 10uF into a 0603 package, but all that said it does appear that there is quite a wide range of DC bias characteristics vs. package size and capacitance value for the X7R dielectric from one of my preferred suppliers, Murata (another supplier, Kemet, doesn't even bother to quantify this behavior for their commercial and automotive parts, only military).

[wraper]

....
X5R & X7R can also have terrible voltage coefficients - there are plenty of graphs showing 70% loss at rated voltage and more than 50% of their capacitance at 50% rated voltage. It depends on the film thickness - the smaller the part is physically for a given rated capacitance the worse - for examples see page 5 of:

http://www.yageo.com/exep/pages/download/literatures/High%20Capacitance%20MLCCs_2012.pdf
...

Well, I stand corrected. I don't use Yageo parts, nor do I have the pressing need to cram 10uF into a 0603 package, but all that said it does appear that there is quite a wide range of DC bias characteristics vs. package size and capacitance value for the X7R dielectric from one of my preferred suppliers, Murata (another supplier, Kemet, doesn't even bother to quantify this behavior for their commercial and automotive parts, only military).

Capacitors from all vendors experience this behavior. You usually can find data in the datasheet for particular part number, or using calcucators from manufacturers where you can select particular series but this DC bias stuff usuallyis even less obvious than derating requirements from tantalum capacitors. Murata has this data in their datasheets too http://psearch.en.murata.com/capacitor/product/GRM21BR61A106KE19%23.pdf
For this particular X5R, 10 uF, 10V, 0805 capacitor, only 25% of the rated capacitance is left at the rated voltage.
If you consider the capacitance dropping more than by half being insignificant, my condolences to your customers. Yet again, I wrote:

Well, if you use MLCC of comparable capacitance without significant voltage derating, you'll find that there is not so much of the rated capacitance left.

Because you were arguing:

that tantalum is no longer even remotely competitive with al. electrolytic and/or MLCC types

I guess they are supposed to compete in similar usage scenario?

[wraper]

Or this Murata X5R 10uF, 25V, 1206 cap http://psearch.en.murata.com/capacitor/product/GRM31CR61E106KA12%23.pdf
At 10V, only half of the capacitance left, at 25V only about 18%

[derGoldstein]

Slightly off-topic -- I've worked with rockets and fireworks hobbyists on all sorts of ignition circuits, and a very cheap and reliable way to ignite a charge is using a 90's era dipped tantalum capacitor. The ones I've used are 10uF, 10v ones.
When you hook them up in reverse and pump around 100 milliamps at 50V into them they blow up instantly and very impressively for their size. I don't think I've ever seen them fail to ignite black powder, and sometimes they're hot enough to ignite candy rockets with no additional chemicals (though you'd probably want to sprinkle a tiny amount of magnesium powder to increase the temperature of the ignition).
Their advantage over heating wire (nichrome) is that you can hook them up using extremely thin wire-wrapping wire, much thinner than you'd usually need to use, so they're great for pencil-scale rockets and tiny remote-activation circuits.

After you use them in that way for a while, you get used to *always* do a reverse bias test every time you start using a new model of tantalum capacitor. And I would never use an older dipped tantalum capacitor in any circuit, no matter how cheap I manage to get them. They're a huge fire hazard.

[MagicSmoker]

...(asshat bits deleted)...
Yet again, I wrote:

Well, if you use MLCC of comparable capacitance without significant voltage derating, you'll find that there is not so much of the rated capacitance left.

Because you were arguing:

that tantalum is no longer even remotely competitive with al. electrolytic and/or MLCC types

I guess they are supposed to compete in similar usage scenario?

Firstly, X7R MLCC capacitors are mainly used for bypass/decoupling, and I really can't think of a case where a bypass capacitor won't do its job even if its capacitance has dropped by 50%. I mean, do you seriously think that, say, a uC is going to start glitching or an op-amp start oscillating because a 1uF cap bypassing the supply rail is really only 0.5uF?

Secondly, there is usually no penalty to specifying a much higher voltage rating for the MLCC than is necessary. For bypassing I mainly use 50V rated X7R caps in values of 1uF and 100nF, depending on how much current the particular IC uses and how much dI/dt might be expected. Even for a 15V rail this results in precious little reduction in capacitance from DC biass

Thirdly, if you want to claim that any given tantalum is equivalent to a MLCC then lets not forget about case size and cost. I have not exhaustively searched DigiKey et al., but a cursory search shows that, e.g., a tantalum 6.8uF/10V/0805 capacitor will cost 2x as much as the equivalent X7R MLCC.

But hey, if you want to use tantalum and I don't then what's the big deal? Frankly, I am sorry I ever commented to this thread and will definitely think twice about relaying my experience here in the future. I really have spent way more time here on this than is warranted. Good day.