Tantalum caps

[asgard]

I have an existing design in work, that includes a modest boost/buck dc-dc conversion stage.  The datasheets (and my previous practice) always used a 10uF and 22uF electrolytic caps as the "safest" bet in the post-flyback filtering to reduce the switching noise.  I know that electrolytics are "easier" to get in the larger capacitances, but lately tantalum caps are showing up "everywhere", especially in the SMD packaging.  I am wondering just how "safe" replacing these polarized caps with tantalums.  I knew that dipped caps of the previous day was notoriously unreliable, but how about now?  For an application that had a 14x10mm cap with 50V rating, how small can one get away with with the same ratings in a tantalum chip cap?
J.R.

[Dr. Frank]

I have an existing design in work, that includes a modest boost/buck dc-dc conversion stage.  The datasheets (and my previous practice) always used a 10uF and 22uF electrolytic caps as the "safest" bet in the post-flyback filtering to reduce the switching noise.  I know that electrolytics are "easier" to get in the larger capacitances, but lately tantalum caps are showing up "everywhere", especially in the SMD packaging.  I am wondering just how "safe" replacing these polarized caps with tantalums.  I knew that dipped caps of the previous day was notoriously unreliable, but how about now?  For an application that had a 14x10mm cap with 50V rating, how small can one get away with with the same ratings in a tantalum chip cap?
J.R.

Hi J-R.,

the small case size is the biggest adavantage of tantalums, definitely, but that's all.
Reliability still is not good, that's given intrinsically by its chemistry, and by the fabrication process.

Ta / TaO is a sensisitive system and will burn, if there's a leakage in the isolator TaO (sort of chain reaction), or if the charging / discharging current is not limited. Also, Tantalum is getting more & more expensive, as it's rare already.

Our development  (automotive mass production) banned Ta caps for those reasons.

Alternatively, there are Ta - polymer caps available, which do not burn down any more, but we don't use it, because instead CN is emitted, perhaps, and they are even more expensive.

Instead, ceramic caps (MLCC) are used, because in the last years, high cap value in small packages are available.
Disadvantage might be, that they might crack under rough environment ( temperature shock, vibration, PCB bending) and will build a short. So in critical places of your circuitry, you might use two CerCaps in serial, or use special types with flexible connection caps, or non-shorting-types.


For power supplies, switch mode or linear, a low ESR is most important in certain places to avoid oscillation. That's given by the regulators datasheet. Therefore, you have to chose  your case size and capacitance value following that parameter. The technology (Ta, CerCap, wet electrolytic) is chosen  following your space, safety and budget requirements.

Frank

[Fraser]

Frank,

Thank you for this very informative post. I had been told of Tantalums habit of burning but had never seen it or known exactly why. Now I know and I will not be using them in any of my projects that do not absolutely require their use. Many thanks to the OP for raising this question and for your answer.

Fraser

[Niklas]

There are a couple of things that you need to check first. Some things that come in mind are:
- Does the converter datasheet specify a minimum or maximum ESR of the filtering cap?
The regulation might become unstable if the ESR of the filtering cap is too high or too low. I have seen appnotes where the recommended cap options have included a small series resistor. Mostly in cases witth ceramic capacitors on older switchers, but anyhow, check beforehand.

- Ripple current?

- Max output voltage including overshoot?
I have not seen it in any datasheet, but there seems to be an old rule of thumb that you should have at least twice the max operating voltage as voltage rating for tantalums. About 5 years ago, before I knew, I encountered some problems with burning tantalum SMDs. Local energy buffer, no high current or temp, but with 13-14 Volts over a 16 Volt rated tantalum.

Size and cost?
Do you really need a smaller capacitor alternative and at what cost? Tantalums are smaller, but they are usually also more expensive.

[quantumvolt]

This is the first time I have seen automotive safety and reliability applied to hobby electronics

So just for those who don't know: Putting two discrete caps in series reduces the capacitance/footprint area ratio (by around 4), but it may prevent your circuit (car) electronics going dead:

Flexisafe components are also structured with a cascade electrode design (which is effectively two capacitors in series within one MLCC) that protects the ceramic capacitor from low insulation resistance failure that may result from thermal stress, repetitive strike ESD and placement damage, said the company. These devices are suitable for military/aerospace COTS+ applications, automotive terminal 30 applications to replace the two capacitor in series requirement, automotive terminal 15 applications to replace the two capacitor requirement, and other safety-critical designs.
http://www.eetasia.com/ART_8800442411_765245_NP_291eef8d.HTM

I myself use tantalums without much thought. Yesterdays caps lasted 5 or 50 years - no reason to think todays won't...

[amyk]

Wet tantalums are still sold for high-rel applications (and the bigger values are quite expensive ).

[wraper]

Tantalums are completely ok but you must derate voltage used by at least 50%, better 70% than specified. Some manufacturers even write this in their datasheets. http://dkc1.digikey.com/us/en/tod/Kemet/tantalumcapacitors/tantalumcapacitor.html

[poorchava]

Conventional tantalum capacitors are currently one of the weakest links in high reliability electronics.

They are very sensitive to soldering conditions, because too fast heating or cooling will develop random microcracks which can result in failure (instant or in the field). Microcracks which are born during production are in most cases not detectable because oxidized tantalum electrode is porous and you therefore you cannot see the microcracks inside the electrode. There are processes which use special voltage/current ramps to "heal" the cracks, but believe me, it doesn't always work .

Other thing is, that they usually fail in rather spectacular fashion (a direct quote from one manufacturer's presentation "it will burn under water, at temperatures up to 700*C").

On the other hand ceramic capacitors have extremly low esr which causes stability issues in converters and they are also prone to cracking (only in a bit different way).

[wraper]

I'm servicing some lab devices made in 93-98 and are running 24/7 non stop. There are hundreds of tantalum caps in each, yet didn't saw any faulty. But I needed to change all electrolytic caps because almost half of them were dead. You never should use tantalum caps of dubious origin because like poorchava said, if they fail there can be small firework.

[asgard]

I think that, for reasons of board real-estate, design life, and fire hazards (good grief it's only sinking 5V) I will just have to accept that I have to use the Al electrolytics for now.  In addition, since the converter is the last stage before the output I cannot rely on connection of purely resistive loads, which puts a bit of kibosh on using tantalums, blast it.  Dave is always on about inevitable failure of wet electrolytics crapping out when the electrolyte has dried up, and I was hoping to avoid that.

[SArepairman]

I like tantalum capacitors electrolytics look ugly, smell bad and they are big. Cheapskate capacitors.

[JoeyP]

Conventional tantalum capacitors are currently one of the weakest links in high reliability electronics.

Other thing is, that they usually fail in rather spectacular fashion (a direct quote from one manufacturer's presentation "it will burn under water, at temperatures up to 700*C").

Damn, that's like a Dreamliner battery. I knew they failed spectacularly, but didn't realize it was that extreme. Personally, I refuse to use them in any design. I spent much of the 80's and 90's repairing test equipment, and was always amazed at just how often tantalums failed. I'm sure I replaced more than 1000 of them during my career. Modern versions may be more reliable, but I've never been convinced of it. And on the subject of reliability, the only thing that comes close to the failure rate of tantalums is reed relays. Won't use them in a design either, unless there is simply nothing else that will do the job.

[wraper]

They are much more reliable than electrolytic caps but the bad thing about them is how they fail. From my experience they fail if they are crappy old ones from 80's, were not derated in voltage or just china crap. Actually I've seen almost only china crap failed on modern equipment. Usually they didn't fail in spectacular way but it is more than possible.

[JoeyP]

They are much more reliable than electrolytic caps but the bad thing about them is how they fail. From my experience they fail if they are crappy old ones from 80's, were not derated in voltage or just china crap. Actually I've seen almost only china crap failed on modern equipment. Usually they didn't fail in spectacular way but it is more than possible.

These days I tend to avoid electrolytics also, in favor of ceramic caps. However, it's hard to beat the cost vs capacitance ratio of electrolytics when you need a large value.Though electrolytics are essentially guaranteed to eventually fail, they do so in a less destructive way and are easy to troubleshoot. In my entire career, I've seen exactly *1* aluminum electrolytic cap fail shorted under normal operating conditions.

[SeanB]

Wet tantalum caps are a different unit totally however, they last almost forever in a glass frit seal case and only blow if you subject them to reverse voltage ( they go off like bullets in a fire) but the wet ones in a regular rubber sealed case ( slightly cheaper) only last about the same time as electrolytics, and fail for the same reason, drying out of the electrolyte from evaporation through the seal.

[Conrad Hoffman]

Everybody forgets the millions of these things that have been used successfully and last near to forever without failure, unlike aluminum electrolytics. The only downside is the failure mode (shorted) and you must never use tantalums where the current isn't limited, and always derate the voltage. They also don't tolerate reverse voltage of any sort. Used properly they will outlive all of us.

I just rebuilt an old RF generator and replaced a pile of small aluminum electrolytics with dipped tantalums, knowing I'll never have to service those again. The power supply got surplus mil wet tantalums so the only two electrolytics remaining were the main filter caps, which were also replaced with new.

Tantalums do have some slightly weird properties that (IMO) make them unsuitable for almost all audio applications.

[poorchava]

Tantalums DO tolerate reverse bias.  The case is that tantalum-tantalum oxide-manganium/magnesium construction behaves a bit like a rectifier.  This means that the capacitor will work ok under reverse bias as long as voltage is very low. I've seen some papers on that,  and the common conclusion was,  that if you run a 35 or 50 volt cap under reverse bias of 3.3 or 5 volts,  it will most likely work on (that is: experiments did not reveal reduced lifetimd).  Obviously applying rated voltage in reverse direction = boom.

Electrolytics also suck,  because they are temperature sensitive,  and lead free soldering process has to be tightly controlled or they get damaged.  They are also bulky and tend to fall off the pcb when subject to vibration or mechanical shock.

[lgbeno]

IMHO, I try to use ceramics wherever and whenever possible and only deviate what I'm outside of the bounds if the technology.  The only time that I chose to use tantalum was when I needed a lot of bulk capacitance and wanted a little bit of esr to cut down on the inrush for the upstream power supply.  (Space constrained application) There were 220uf tantalums available rated at 16V.  I did blow some IP too when testing, to me they are very temperamental, you don't have to go very high above their rated voltage for them to turn into vapor.

One note on ceramics, because of the almost non-existent ESR, a filter made from ceramics might not need as much total capacitance to get the same amount of ripple attenuation as other cap technologies.

Like in my application above though, Low ESR can be a double edged sword because on inrush.  At the end of a long cable, the high inrush dI/dt, could get you some high voltage spikes without much cable inductance.


Sent from my iPhone using Tapatalk 2

[poorchava]

Well, ceramics suck too in their own way.

Their main failure mode is cracking as a result of thermal or mechanical stress. For a robust pcb design you need to lay out ceramics in such a way, that longer side of the capacitor is perpendicular to bending vector (i don't know if that is the right term) to minimize the displacement of capacitor leads.

Other thing is that then they are soldered in reflow oven and the solder cools down, it shrinks and tends to tear the capacitor sides apart, which results in "flaking" cracks (i mean layers seem to come off from the capacitor). This may or may not lead to electrical failure, but you never know.

Generally for a hi-rel products ANY capacitors are bad, but ceramics are probably the least fragile type and they are preferred. There are also "open mode" capacitors, which are designed to fail as an open circuit (eg. will not blow up, because they shorted the main logic supply rails). The downside is that they are a bit more expensive and are available with lower capacitances for any given package and voltage rating.

[GK]

I can't remember how many tantalum capacitors I've had to replace over the last 10 years of maintaining a suite of bore hole logging tools full of them. After ~20 years of existence they started going short at an exponential rate.

[Conrad Hoffman]

Hmm... what's the temperature down that bore hole? On the reverse voltage thing, I'm way more conservative now. After testing a bunch of tantalum caps removed from some audio equipment, the few that ran under even a couple tenths of a volt reverse voltage were the only ones that had higher leakage current after 30+ years, and they didn't come back to normal under forward bias. Thus, even though slow, I have to assume damage was being done and even slight reverse voltage is to be avoided.

[woodchips]

Just don't forget that the wet tantalums used sulphuric acid as the electrolyte, also that the cathode was a piece of solid silver making them attractive things to be recycled.

[nctnico]

I avoid using tantalum caps. They are basically bombs (oxidizer and fuel). If you put too much current into them, they ignite. If you reverse the voltage they ignite. Assembly can be prone to errors as well. I have seen people put them in circuits in reverse on several occasions. On an electrolytic the - is marked, on a tantalum the +. Last time one exploded right into my face. Fortunately I have glasses but I needed a new pair after that incident. In an earlier mishap they burned right through the PCB so people had to go out to check about systems on 40 sites all over the country. Another problem is that to make tantalum caps they need to cut down rain forrest to mine tantalum.

Electrolytics don't have to be a problem when choosen properly. Their lifespan is easy to calculate.

[lgbeno]

Tantalum supply dries up from time to time as well which affects lead times.  Not fun of you do just in time manufacturing...


Sent from my iPhone using Tapatalk 2

[GK]

Hmm... what's the temperature down that bore hole?

These were low temp tools that would very seldom see 60 deg C. Maybe being vibrated over hundreds of thousands of kilometers of pot-holed and corrugated dirt road and track played a part.