An alternative to solid tantalum capacitors | Hacker News Reader

An alternative to solid tantalum capacitors

electronicsweekly.com

86 pointsrbanffy44 comments

Dad used to be head of QC at Kemet (he retired from there in the early 90s). They were known for high-reliability Tantalums (used in various ICBMs, undersea cable repeaters, and the System 360). These days they make many different styles of capacitors and other components.

They did 100% inspection - twice, to ensure they shipped a good product. Like many electronics production lines, the yield of working product was important to track. And they didn't have the ability to ship a capacitor with part of it disabled, like Intel used to do with the 80486DX and 80486SX (the SX was a DX with a non-functioning floating-point math processor).

Story time: They received Tantalum via truck delivery. And one day the truck driver failed to clear a rail crossing in time and was hit by a train. He was fine, but the expensive metal was scattered along the rail bed. Obviously they couldn't use it after the accident, as it was contaminated. So they paid employees to go out there and pick up the pieces so they could be returned to the smelter and get credit for it.

Tantalum died years ago. It was a suicide.

Multiple scarcity events, skyrocketing prices, and then in 2014 you couldn't buy them even if you wanted to.

Purchasing agents were seen chasing design engineers with pitchforks by torchlight.

Ceramic MLCC's have filled their spot. Better, much easier to use and safer.

And the modern poly hybrid electrolytics are amazing, as well.

Tantalum poly caps are probably great, but the fear is still there and too recent - so I doubt they will get very popular except in small specific instances where they are absolutely better than anything else.

Ceramic MLCC's aren't without their problems, especially in the smaller packages. Just take a look at a voltage vs. capacitance curve in an 0402 package ceramic cap. When used at the rated voltage, they don't have anywhere near the capacitance they state. Also the temperature vs. capacitance curves are less than appealing.

That said, I think a lot of engineers out there have been burned (sometimes literally) by using tantalum caps in the past and are likely not to design them in in the first place these days. I know we moved away from them completely in my last job because enough exploded to scare people off.

EDIT: See dragontamer's comment below for more. Explained better than I did.

Do you have any suggestions for what to use for a regulator eg. lm2940 - that needs something like 22uF, low ESR on the output pin for stability? MLCCs are expensive and not very available in this capacitance range.

Sorry, I can't really recommend anything without more information. There's just too many variables.

I briefly looked at the datasheet for the regulator and it has some constraints for the output cap that should help guide your choices. For example, you need to look at what temperature range you expect the regulator/cap combo to be working at, as certain dielectrics are sensitive to temperature changes [1]. If you need stability over a wide range, you might go with an X7R or better, otherwise X5R might do. You will also have to look at derating the capacitance for the operating temperature - you might find that in a hot environment you actually need a 47uF cap or greater to meet the minimum 22uF output capacitance requirement of the regulator.

The other main thing to consider is operating voltage. The regulator you list is adjustable, so you will need to pick an output cap that has an appropriate capacitance at the voltage you plan to run the regulator at. The curves drop pretty steeply, so it's really important to analyze this.

For ESR you should look at manufacturer websites. They often have a "picker" tool that shows graphs of that (it's dependent on frequency!). You probably won't have trouble finding caps with a low enough ESR value for this application as long as you stay away from electrolytics.

Also, you can consider paralleling some lower value capacitors to achieve the capacitance you need, and reduce the ESR a little. I wouldn't use it as an ESR lowering strategy however [2]

[1]: https://www.maximintegrated.com/en/app-notes/index.mvp/id/55...

[2]: http://electronics.stackexchange.com/questions/152549/esr-an...

For those wanting to learn facts (as opposed to lore) the video at Hackaday gave me more insight on capacitors then any other single resource I've come across.

http://hackaday.com/2016/07/09/capacitors-are-simple-right/

https://youtu.be/ZAbOHFYRFGg

How ultracapacitors work (and why they fall short)

https://gigaom.com/2011/07/12/how-ultracapacitors-work-and-w...

The article focuses heavily on the angle of tantalum ore supply -- but it's worth noting that Niobium capacitors [1][2] have been gaining prominence since the 2002 for the same reason. Niobium-based capacitors have some desirable properties [2] but are roughly comparable to tantalum-based ones; and Niobium ore is much more abundant.

[1] http://www.avx.com/products/niobium/ [2] http://www.imaps.org/adv_micro/2004may_jun/pages/article_nio...

Tantalising news. The alternative is still tantalum-based if I understood correctly, but uses less tantalum. The article doesn't mention how much less.

Well, "solid" doesn't have to mean "not a liquid/gas/plasma", it can also mean "all the way through". For instance a table made of "solid oak" is not trying to say it's not a liquid, it's trying to say it's not veneered particle board.

That's how I interpreted the use of "solid" here, but I'm no electronics engineer of course.

"Solid" in electronics means the former (state), not the latter.

It means solid as in the phase of matter in this case. The electrolyte in a tantalum capacitor can either be a solid or liquid.

I don't think that's universally the case, for example solid-state storage refers to the lack of moving parts.

https://en.m.wikipedia.org/wiki/Solid-state_storage

Edit: though it does appear to be the case in this context, just to be clear.

Solid capacitors are ones that do not contain liquids. Most electrolytic capacitors use the junction of a metal and a liquid as their active element. Solid ones use the junction of two solids.

The past three companies I've worked for that were doing circuit board design expressly forbid the use of tantalum capacitors because they are surface mounted (SMT) two-terminal parts that have polarity. In other words, tantalum caps have a + and a - terminal and are damaged when reverse voltage is applied. Blanket ban on tantalum use was to avoid mistakes during PCB assembly in manufacturing where caps are oriented incorrectly.

Edit: Maybe I misunderstood the reason for a blanket ban of tantalums. The fact remains, the ban was real.

What are some of the places where tantalum capacitors are required and ceramic capacitors just can't get the job done?

As a hobbiest, I usually see them in high stability power supplies.

Probably because when designed right (they are far more finicky then ceramic) they typically offer higher energy density.

Higher capacity you mean (for a given volume), that and also very low leakage current so if you want to make a very efficient device then tantalum might pay off. Another reason is that electrolytic caps are usually wound aluminum sheets which act as a coil. If you don't want any parasitic inductance tantalum is better in that application (better frequency response).

Tantalum caps have much worse leakage current than ceramics.

The debate for most designs is not tantalum vs. electrolytic, it's electrolytic vs. solid vs. ceramic. And ceramic usually wins.

They have lower internal inductance and resistance than other electrolytics. That is very goo for making the low-pass filters one uses on power supplies.

They're usually used as bulk capacitance, particularly at the input side of switching regulators. Thing is, their ESR is too high to be used alone so they're usually used in parallel with an MLCC ceramic. I've had success replacing both on the input stage with an aluminum polymer electrolytic (not standard aluminum) due to their low ESR and high capacitance (and better inrush rating). I usually use ceramics on the output.

[1] http://www.ti.com/lit/an/slta055/slta055.pdf

This is a great writeup on the various types from Kemet:

[2] http://www.kemet.com/Lists/TechnicalArticles/Attachments/5/A...

There are many reasons to not use tantalum caps (inability to handle voltage transients requiring derating being the primary that comes to mind), but I've never heard of a company instituting such a policy because their assembly process was so poor.

Did they also blanket ban diodes?

Might be a retcon. There are sometimes good reasons not to use tantalum aside from the explosion and fire inevitability.

At really high temperatures (engine computers etc) tantalum loses voltage rating down to less than 50% in some cases at the limit whereas ceramics lose capacitance with high temp (weird but true for X7R dielectric). This is a general rule of thumb and I'm sure you can find mfgrs whitepapers explaining how their expensive series is immune to this general class effect.

In the bad old days tantalum volume was like 100 times smaller for a given capacitance than a ceramic but its not the bad old days anymore and its like 3:1 ratio for some MLCC and because of voltage derating a tantalum derated to 25V (aka a 50V+ cap) might not be all that much smaller than a 25V rated MLCC. Size of cap is no longer a relevant criteria, but customers are more excited than ever about burning exploding personal electronics, so ...

Another fun one is leakage current for tantalum is about three orders of magnitude (whoa!) higher than MLCC. Its possible to build little ultra low power IoT things that sleep most of the time and a tantalum leaks more mAH than the processor but a MLCC would waste less.

High frequency response is fun because neither is inherently better but in some applications one or the other is better. Generally the MLCC is going to have a higher Q resonance at a higher frequency. If that freq is irrelevantly high to the app and EMI issues, thats a great cap, whereas maybe for some apps sometimes tantalums could be resonant, admittedly at a lower Q, in the region of freqs important to the app or EMI/EMC issues.

Not that ceramics are problem free. Piezoelectric effects are hilarious. Aging after heating drops ceramic capacitance, its possible to solder a SMD cap and watch its value drop a couple percent over the next day or so.

As with all effects you can acquire manufacturers whitepapers explaining how their extremely expensive exotic series cap can be soldered in place of another of the same family if there's a design mistake and you need to correct it in production, so sure you can pay a lot of money for tantalums that work at 230C or temperature stable ceramics.

Something I don't understand is ceramics being ceramics mean they should be good up to lava like temps and alloys being alloys mean you should be able to find something with a matching coeff of expansion thats good up to liquid metal temps, so why can't you buy ceramic caps rated to orange-hot temps? I haven't looked into it deeply enough to find out. There are practical reasons wet tantalums can't run at cutting torch temperatures but ceramics superficially don't seem to have those limits.

Low ESR applications.

Ceramic Caps have extremely low ESR as well however.

Well, show me a 10uF, 10V below 1 Ohm.

Ummm.....

http://psearch.en.murata.com/capacitor/product/GRM219R61A106...

Try below 0.01 Ohms, between the frequencies of 100Hz to 1MHz, in the 0805 form factor. 10uF and 10V as requested.

Ceramics are extremely good at low ESR / low ESL / decoupling applications. I mean, so are Tantalum caps, so they're both good for this purpose. If anything, Ceramics might have better ESR.

He, thank you very much, I will look into this. As a self-taught electronics designer ( https://mecoffee.nl ) I find the amount of different components ( at Farnell, Mouser et al ) absolutely intimidating.

Could you provide some insight into why for tantalums the ESR is specified, while for these ceramics they are not? How should I go about filtering them at Farnell/Mouser ?

Thanks again.

> Could you provide some insight into why for tantalums the ESR is specified, while for these ceramics they are not?

Ceramics are highly frequency dependent. Almost every parameter changes with frequency: capacitance, ESR, ESL... you gotta look at charts.

For the most part, when picking a Ceramic Cap... you pick a dielectric (NP0 for voltage and temperature stability and accuracy... X5R for "high" values like 10uF).

> How should I go about filtering them at Farnell/Mouser

The only methodology that seems to work for me is to learn the various types of dielectrics. For the most part, the properties of NP0 ceramic capacitors are the same across all the companies.

Class 1 (typically NP0) and Class 2 (typically X7R... but X7S, X5R or whatever is fine) are your main types. Class 1 for temperature and voltage stability and accuracy at the nanoFarad range, Class 2 for "high" values like 10uF.

the ESR on MLCCs are negligible

Should check out this from an expert who knows! I learned more in this video then almost anything else.

http://hackaday.com/2016/07/09/capacitors-are-simple-right/

https://youtu.be/ZAbOHFYRFGg

This seems to be a reasonable article about low ESR selection : https://abacus.avnet.com/wps/portal/abacus/article/understan...

MLCC are used in broader-frequency applications. The loss angle tangent varies much less than ESR over frequency, and so MLCC caps (and film caps at higher voltages) are more likely to specify the loss angle tangent instead.

> What are some of the places where tantalum capacitors are required and ceramic capacitors just can't get the job done?

Ceramic Capacitors are a class of class of capacitors. We need to break things down further... Your "NP0" ceramics have great overall characteristics, but are very small... 0805 NP0 ceramics top out at 0.01uF (and are quite expensive at such "high" values).

To get higher capacitance in a ceramic capacitor, you switch over to lower-quality ceramics, like X5R or X7S. X7S gets to 100uF or higher in the 0805 form factor, but are both temperature dependent and voltage dependent.

Yes, Voltage dependent capacitance.

http://i.imgur.com/mnqbVJC.png

Here is Murata's "Simsurfing" page, which describes the characteristic on a 100uF 0805 X5R Capacitor.

A 100uF X5R ceramic capacitor by only be 15uF by 5V. Its capacitance changes with voltage. Strangely enough, for most purposes (ie: decoupling), that's not an issue. But any precision circuit (ie: OpAmps or timers) will prefer to use a more voltage-stable capacitor.

Someone else mentioned power-supplies and energy storage. Yeah, a X5R ceramic capacitor has much less energy storage because it loses capacitance as it stores energy (and "high quality" NP0 ceramics just don't have much capacitance to begin with... so no point using those for energy storage).

More issues, see: http://www.kemet.com/Lists/TechnicalArticles/Attachments/93/...

Don't electrolytic caps have the same issue?

Almost every part has the same issue except for the most basic, so that's not the reason.

Past 3 companies?!? What?

How do you handle diodes, LEDs, batteries, and electrolytic caps? Those are also two terminal devices with polarity. Heck how did you even build cables correctly?

I've worked with voltage regulators from Linear that explicitly call out using tantalum caps in the recommended design.

A good reason not to use tantalum caps is that 10 years down the line or so they seem to transform into automatic firestarters.

But if you aren't able to correctly install parts with any polarity then the only parts that you might be able to use are resistors, coils and capacitors without polarity (which tend to be low capacity), which would make for rather boring designs.

Ceramic caps are usually very low capacity, tantalum and electrolytic caps (much) higher.

>Ceramic caps are usually very low capacity, tantalum and electrolytic caps (much) higher.

30+ years ago, this was true. Not anymore.

Show me where to buy a 1000 uF 50V ceramic cap for a reasonable price.

That's a stupid demand. You said "very low capacity". You can get MLCC caps in the ones and tens of microfarads easily and cheaply now, which isn't "very low". "Very low" is picofarads. Ceramics advanced from that level ages ago.

Interesting article, it reads like they are trying to rehabilitate the tantalum capacitor market.

Back when NetApp was got a deal to provide filers for IBM to re-label the word came from IBM that all Tantalum capacitors had to be designed out of the system. There was an interesting project in engineering to purge the designs of all tantalum capacitors. IBM's reasoning was that they had been the source of fires as I recall.

As a result I find myself always hesitating to use one (even in silly one-offs) and other engineers I've met have similar habits.

Great article. Was not aware there is this much stigma around tantalum caps. There are times when I am forced to avoid ceramic caps for better high frequency handling. It looks like polymer is the way to go - better longevity and no ethical issues.

Funny that the alternative also uses tantalum and in its solid state... I know, I know, "solid tantalum" refers to a technology and the alternative is a new technology, or at least an improvement over the older one.

Reply on news.ycombinator.com