Ultra high reliability capacitors

Hi all,

I'm working on a design at the moment requiring very high reliability at elevated ambient temps, up to ~70degC. I know electrolytic have certainly increased in reliability over recent years - but are still commonly the failure point in any design that uses them.

It seems solid tant capacitors are now becoming the cap of choice in high reliability applications. However, two questions.

a) I believe the common failure mode of a tant cap is short, as opposed to open with an electrolytic. So, as a second safety layer - I suppose I could use a Polyfuse or similar in series with the bulk tant caps to prevent shorts IF failure occurs b) Would potting a device (with a compound with high thermal conductivity, not cheap I know...) with electrolytic caps extend their life by preventing the caps from drying out?

Thanks!

EDIT - http://www.vishay.com/docs/40126/ds04053.pdf It would seem good quality solid tants have internal fuses in them already!

The common failure mode of a 'tant' is fire, followed by a short. The fuses they started putting inside them are to prevent the pyrotechnics.

If you put in a fused tant, does the failure mode change to 'open'? (I dunno).

Anyway, the trick for reliable tants is to really, really over-specify their voltage rating.

Hi Fungus,

Well yes - I would presume the failure mode would be referred to as open, technically - as a fuse blows - and surely, this fuse should blow before any fire!

However, some parts of my application require roughly 470uF, with a max voltage rating of ~40V - something that tants seemingly can’t provide. What are my options - considering I don’t want to use electrolytics?

Do you mean just reliability or lifetime? Aluminium electrolytics are very reliable when new, but suffer from age/heat related deterioration. Some people avoid tantalums altogether because of the failure mode (rather than the chance of failure). The failure mode of tantalums is a metal-oxidiser fire (like thermite).

Why do you need 470uF at 40V - for high reliability you might want to think outside that box? What are the ripple-current requirements - that affects self-heating which will be an issue when already running hot.

Ceramic capacitors can be employed if you can afford it (470uF is a lot, but 47 x 10uF MLCC's isn't beyond the realms of feasibility, but you get very poor tempco performance.

Do you mean just reliability or lifetime?

Well, I'd define reliability as intended operation over a long lifetime - what would you define it as?

Aluminium electrolytics are very reliable when new, but suffer from age/heat related deterioration. Why do you need 470uF at 40V - for high reliability you might want to think outside that box?

The caps are the main caps in an h-bridge driving currents up to ~10A, the chip in use is recommended as having 470uF per 10A load current, and my supply voltage is up to 30V. However, I wonder if for reliability I simply have a TVS diode to absorb any spikes that switching may cause from any parasitics, and omit the capacitance.

Sigh... aluminum electrolytics are not leaky! They will not dry out unless they are going bad. They are hermetically sealed.

Potting them is a -very- bad idea! Larger Al electros are prescored on top so they will split, rather than explode like firecrackers. Sealing them up tight will make that firecracker into something considerably more powerful. Having a high thermal conductivity for a potting compound won't help - that only helps when you are trying to remove heat, where the object in question is at a much higher temperature than ambient. But if your aluminum electrolytics are hotter than the ambient temperature, they are already failing.

Don't be temped by the new solid electrolyte aluminum capacitors. The lifetime goes down -much- faster with elevated temperature than it does for standard damp electrolytic aluminum capacitors.

FYI, 10uF ceramic capacitors are -very- voltage sensitive, the capacitance goes down very quickly with increased voltage. Keep that in mind when designing with them.

We probably also need to define "long lifetime" but I have a several 4000µF 50V electrolytics that have been operating (at ~40V) just fine for about 40 years now. They have an 85°C spec but are just used at room temperature, they're filters in the power supplies of audio amps. I did have a rectifier fail after a few years though. Just anecdotal evidence FWIW.

polymorph: Sigh... aluminum electrolytics are not leaky! They will not dry out unless they are going bad. They are hermetically sealed.

No, Aluminum Electrolytics are NOT hermetically sealed. Their seals MUST allow some venting to occur. As the electrolyte is consumed, hydrogen gas builds up. That gas must be allowed to diffuse through either the case or seal, otherwise it would explode. By definition, Aluminum Electrolytics are not hermetic.

polymorph: Don't be temped by the new solid electrolyte aluminum capacitors. The lifetime goes down -much- faster with elevated temperature than it does for standard damp electrolytic aluminum capacitors.

While it is true solid organic aluminum capacitors do have reduced lifetime at elevated temperatures (like all components), that loss of life does not occur until well beyond 85C.

Many people mistake the "qualification" of solid aluminums to be the same as "endurance" for wet electrolytics. This is absolutely not the case. Solid Organic Polymer capacitors do not have a wear out mechanism like the "wets". The Polymer is the actual cathode material, not an electrolyte. (This is not the case with OS-CONs which use a solid salt AS an electrolyte.)

Also, don't compare Solid Aluminums from a non-name Asian company. Tier 1 Solid Aluminums are excellent components.

jtw11: It seems solid tant capacitors are now becoming the cap of choice in high reliability applications.

Solid Tantalum-[u]Polymer[/u] capacitors are becoming the choice for high reliability applications.

Here is a paper from Raytheon about why they are switching to Polymer-Tantalum in their applications: http://ecadigitallibrary.com/detail.php?cid=27&pid=1965

Traditional Tantalum-MnO2 are the ones that ignite on failure. While it is possible to make a highly reliable MnO2, the cost goes from 10s of cents to 100s of dollars (seriously.)

jtw11: a) I believe the common failure mode of a tant cap is short, as opposed to open with an electrolytic

All caps can fail short. The most common for tantalum is short. Most common for electrolytic is also short. Ceramic can fail short. Film very rarely fail short, but can. The key difference with wet-aluminum electrolytic caps is that their wearout mechanism is to fail open.

Keep in mind that "reliability" and "lifetime" are very different things. Tantalums have lifetimes on the order of 1000s of years. However, their most likely failure is during power-on. Fortunately they self-heal which makes them more robust over time. I explain more in this posting: http://www.baldengineer.com/blog/2013/09/13/ouch-the-arduino-gsm-shield-has-a-pretty-serious-design-flaw-with-its-capacitors/

jtw11: I suppose I could use a Polyfuse or similar in series with the bulk tant caps to prevent shorts IF failure occurs

It depends on the application, but some times a fuse can be used. You have to watch picking a fuse value that would limit the cap's ability to charge or discharge itself.

It would be a very good idea to consider a Polymer-Tantalum and de-rate the voltage by 10-20%. If you de-rate 50% its failure rate drops significantly. It is very important you properly de-rate the applied voltage of a Tantalum (MnO2 or Polymer). De-rating does not improve lifetime (well it does, but not to an order anyone cares). Instead de-rating improves the reliability by reducing the chance for a dielectric breakdown to occur during power-on.

jtw11: with electrolytic caps extend their life by preventing the caps from drying out?

You really don't want to put an aluminum electrolytic. As it ages, it releases hydrogen gas.

jtw11: EDIT - http://www.vishay.com/docs/40126/ds04053.pdf It would seem good quality solid tants have internal fuses in them already!

Tantalum with a built-in fuse is a "specialty" product. If the data sheet doesn't say it has a fuse, it doesn't. Again, I would recommend using Polymer-Tantalum over any variation of MnO2. I don't know of anyone putting fuses in Polymer-Tantalums because of their improved reliability.

Polymers-Tantalums have a lower failure rate, lower ESR, and in the event they do fail, are unable to ignite.

The reliability as many here have said depends on a lot of things including who manufactures them.

I find that I am replacing electrolytic caps in all sorts of equipment, expensive well known branded gear to el-cheapo not mainstream manufactured equipment and they all suffer the same degree of cap failure.

Most swell and open their tops or leak out the bottom, or end up with a very large ESR. Most electrolytics I come across, brand new stock or failed caps have been made by many different manufacturers.

Since the popularity of switched mode power supplies, my work load has increased, repairs mainly being caps, in particular values below 100uF and operating voltages above 50V. Sometimes you can almost guarantee the littlest cap will be the culprit. I have had caps that fail short circuit, high ESR, low capacity, test as a diode, or think its Christmas and blow tinsel ALL over the PCB.

Buying from the main component suppliers that can provide a spec sheet is probably the only way to make sure you are going to get what you pay for and working well within the specs.

Tom.. :) my ten cents worth....

Aha, James - you must be the notorious capacitor expert on these forums whom I've heard so much about.

Thank you so much for your comprehensive and detailed reply, I shall be re-reading this a few times today.

I do intend to pot my design, and knowing now that aluminium electrolytics release hydrogen during use - will certainly not be using them, and will be switching to tant polymer caps.

I have also just been reading your blog post re the GSM shield failures - very interesting to see the effects of derating version polymer and non-polymer tants!

Thanks again.

Once again James has taken me to school. Which means I know more things now than I did before.

polymorph:
Once again James has taken me to school. Which means I know more things now than I did before.

Pales in comparison to how many times you’ve schooled me.

Blush.

So, as I said I would - I've just been rereading this thread...

Whilst I see from many a source now that polymer tantalum are highly reliable caps - the main bypass caps used for power applications tend to be of the order of 1000s of uF. My input voltage can be expected to be as high as 30V, but no more.

From the searching I've done, it seems no polymer tantalum exists this large to suit such a requirement.

So, do I simply omit my bypass caps - and ensure that PCB trace parasitic properties etc are minimized as much as possible?

Also, I've got some high side P FETs driving loads with PWM, I had also planned on having each of these FETs with their own ~470uF caps (again up to 30V) to significantly reduce any noise that the switching would cause, clearly - one cannot do this with solid tants.

So, what is one expected to do?

jtw11: Also, I've got some high side P FETs driving loads with PWM, I had also planned on having each of these FETs with their own ~470uF caps (again up to 30V) to significantly reduce any noise that the switching would cause, clearly - one cannot do this with solid tants.

So, what is one expected to do?

Make the PCB/box bigger and use polyester+ceramics....?

Enclosure and PCB wise, I have a size to which I simply cannot exceed - for a number of reasons. I will be looking at polyester film caps however, but PCB space is getting very tight indeed.

What do you mean when you say use ceramics though? Ceramics are used extensively throughout my design for decoupling, bootstraps etc. But I thought using large value ceramics was a BIG nono due to the high tendency for cracks to form?

the main bypass caps used for power applications tend to be of the order of 1000s of uF

Sorry, bypass caps bypass rf or noise gliches in supply rails caused by digital switching or external emf, they are strategically placed physically around a pcb. They are usually of the order of 10uF down to 0.01uF depending on what you want to bypass.
Caps in the 1000s of uF are termed filter or smoothing caps to smooth out ripple from AC to DC conversion.
What is it that you want to do, bypass or smooth?
Each type of cap has its own unique characteristic and hence application.

Thanks James, Solid Tantalum-Polymer capacitors may help with a common failure of a piece of equipment I work on.

Tom… :slight_smile:

Sorry, bypass caps bypass rf or noise gliches in supply rails caused by digital switching or external emf, they are strategically placed physically around a pcb. They are usually of the order of 10uF down to 0.01uF depending on what you want to bypass.

I would refer to those as decoupling caps, as does most of the documentation I've seen.

I know bypass caps perhaps isn't the ideal description, but a number of sources still refer to high value capacitance in power applications as bypass, do they not?

But yes, you're right - the requirement here is filtering out noise created by PWM drive of various loads, in addition to bulk input capacitance to DC-DC converters etc.

I know bypass caps perhaps isn’t the ideal description, but a number of sources still refer to high value capacitance in power applications as bypass, do they not?

What are “power applications”? That is a very generic term. In this case, it does appear that you are talking about capacitors placed around the circuit in such a way as to smooth out current spikes that would otherwise propagate around the circuit via the VCC lines. If you are finding it necessary to place many large capacitors around your circuit, perhaps a design review is in order? Proper layout of power and ground lines helps, too.

But yes, you’re right - the requirement here is filtering out noise created by PWM drive of various loads, in addition to bulk input capacitance to DC-DC converters etc.

Switch mode power supplies and load controls usually have fairly high frequencies and therefore don’t require really high capacitance.

May we see the schematic?

May we see the schematic?

Right now, no - but not being i'm being a pain :), but because this is just a general 'what-if' discussion at the moment, and in fact you've just taken the words out of my mouth.

If you are finding it necessary to place many large capacitors around your circuit, perhaps a design review is in order? Proper layout of power and ground lines helps, too.

I haven't yet found it necessary at all, as I'm yet to design and have the first PCB fabricated for this project so I have no idea if any issues are currently present. I'm rather thinking in advance, that if some filtering was in order - what are the more reliable options compared to the usual aluminium electrolytics.