A few weeks ago, I found myself in need of a single 22µF Tantalum capacitor for use with a 3.3V voltage regulator. Since I didn’t require any other components, I felt a bit silly placing an order with Digi-Key, so I emailed a couple of friends to ask them if they had an unused device skulking around at the bottom of their spare parts drawers.

In response, my friend Rick Curl emailed me to say that he’s largely stopped using Tantalum capacitors in his designs. Intrigued, I asked him why this was, and he responded as follows:

For a long time, I’ve been in the habit of generously sprinkling Tantalum bypass capacitors between the supply and ground buses of my analog and digital PCB designs. It’s good engineering practice to prevent noise and spikes on these buses. However, back in 2011, a couple of things happened that caused me to rethink my choice of Tantalum capacitors.

I had received a hand-assembled prototype PCB containing a bunch of surface mount parts. As I powered it up, I was thinking that I would probably need to make a few minor tweaks before releasing the design to production. About that time, there was a loud BANG… BANG… BANG, and I noticed the bench supply that was initially supplying 5V at about 0.25A was now reading ~0.5V, while the current was ~2A. I quickly shut the supply down and took the PCB outside, leaving a trail of smoke and a nasty odor behind.

Max here; I’m assuming that the nasty odor was from the charred circuit board and not Rick himself, but we digress…

I stood beside the road out front and examined the board. There were little charred holes in the PCB where some of the bypass capacitors had been. Looking at the remaining capacitors that had not yet failed, I realized that they’d all been installed backwards. It turned out we had somehow acquired a different brand of capacitor than the ones with which the assembler was accustomed, and she had misinterpreted the polarity markings. My first thought was to get more of the original brand of capacitor, but I soon discovered that the price of Tantalum had more than tripled almost overnight, thereby driving prices through the roof. Combined with the fact that Tantalum capacitors almost always fail to a short if you reverse their polarity or exceed their working voltage, I decided that it was time to find some other alternative. I also decided that it would be really nice if I could drop the new components on the same surface mount footprint.

Aluminum electrolytics were not an option. Their ESR (equivalent series resistance) is too high for them to be effective as bypass capacitors, they’re too big, and — like Tantalum’s — they’re also polarity sensitive. Film capacitors might work, but it’s hard to find physically small ones with values above a microfarad or so. Eventually I started looking at Multi-Layer Ceramic Capacitors (MLCCs). It used to be that these were only available in lower capacitance values, but a quick search of Digi-Key’s website turned up values over 100 microfarads, even in a 1206 footprint! These are not polarized, so installing them backwards is not a problem. Could the solution really be as simple as dropping a ceramic capacitor in the footprint where I used to have a Tantalum? I started doing some research to see if there were any hidden “Gotcha’s,” and — as you might expect — there were. I found this technical article from Maxim Integrated to be particularly helpful.

What I discovered is that ceramic capacitors are generally split into two types, or classes. Class 1 is the stable, low-capacitance type that is often used in tuned circuits. Type 2 is available in much higher capacitance values, but with a couple of major shortcomings that we must keep in mind. The first of these shortcomings is the temperature coefficient. The chart below shows the temperature ranges and the capacitance variation over temperature. I usually use X7R, which is specified from -55 to +125° C, where the capacitance is allowed to vary 15% over this range.

It’s important to make sure that the capacitance remains within an acceptable range over the expected temperature range of the product.

Now that’s out of the way, we need to look at voltage coefficients. This is actually the bigger, but more obscure, issue. Take a look at a typical Y5V capacitor at half of its rated voltage as illustrated below.

As we see, the capacitance has dropped from 1.0µF to 0.3µF. That’s a HUGE loss! There are a few things you can do about this. One is to go to a larger case size. Notice that the above chart is for an 0805 part. Changing to a 1206 causes the capacitance to drop much less as voltage is applied. You can also select a higher voltage part or find a part that uses a different dielectric.

It is hard to measure the actual capacitance value while voltage is applied to a capacitor. Empirical testing of the actual in-circuit parts over a range of temperatures and voltages can help to understand if the capacitors are performing as expected. While the capacitor’s performance over temperature is not that difficult to predict, information about capacitance change as voltage is applied can be a bit more elusive (Murata has this very nice online tool that will allow you to see the effects of temperature, voltage and other characteristics for their capacitors).

By the way, the part about the voltage coefficient is apparently not well known at all. When I first started reading about it, I called the Murata FAE in Huntsville and asked him if I was reading the graph correctly where it indicated that I would lose 70% of the capacitance value at 50% of the rated voltage. He initially replied “Surely not — nobody would want capacitors that had that problem.” But he called me back the next day to say he had checked with the factory and that the chart was indeed correct. (Just to clarify, this voltage coefficient issue mainly occurs in with Class 2 / Bypass ceramic capacitors; it is much less pronounced in tantalum, aluminum electrolytic, and film capacitors.)

In closing, the discussions above are certainly not meant to be a thorough comparison of capacitor types. They’re also not intended to imply that Tantalum capacitors are bad. They are not. There is much more in-depth information available about the comparison between capacitor type, such as this technical article.

It’s my goal to share my experiences with you to keep you from learning some lessons the hard way (like I did). I was able to put ceramic capacitors on the same footprints as Tantalum ones on those boards, but only after careful research. I continue to use ceramic MLCCs on almost all my board designs today.

Wow. As you may recall, all I wanted was a single 22µF Tantalum capacitor for use with a 3.3V voltage regulator. Happily, while I was wading through Rick’s email, my chum Duane Benson rooted one out and dropped it in the post. Having said this, whenever a circuit calls for a Tantalum capacitor in the future, I shall certainly consider using a Multi-Layer Ceramic equivalent.

What say you? Do you have any capacitor-related experiences you’d care to share?

— Max Maxfield, Editor of All Things Fun & Interesting

Max covers programmable logic, microcontroller units, and prototyping for EE Times' Designlines. Over the years, he has designed everything from silicon chips to circuit boards, and brainwave amplifiers to steampunk "Display-O-Meters." He has a BSc in Control Engineering in from Sheffield Hallam University in Sheffield, UK.

Max covers programmable logic, microcontroller units, and prototyping for EE Times' Designlines. Over the years, he has designed everything from silicon chips to circuit boards, and brainwave amplifiers to steampunk "Display-O-Meters." He has a BSc in Control Engineering in from Sheffield Hallam University in Sheffield, UK.

A reported problem with MLCCs and one that we have experienced here at work is that when a MLCC is flexed, the capacitor cracks and fails. And this was with 0805 devices. The propsed solution is to place the caacitors away from the board edge and orient them in the direction of least flex especially when the board is borken out of the panel. See Kemet's "Flex Crack Mitigation"

@antedeluvian: ...when a MLCC is flexed, the capacitor cracks and fails ...

I'm much the same way myself these days LOL

But good info -- thanks for sharing

When I was about 19 I made myself a breadboard in a box with a power supply built in.  Pretty standard power supply, multi-voltage transformer, a couple of bridge rectifiers, some hefty Al electrolytics, and 7805 / 7812 / 7905 / 7912 regs.  I put a 22 uf 16V tantalum cap on the output of each reg.  it worked fine for yonks.  fast forward 30 years or so and I arrived in Australia with said breadboard, but I didn't need to use it for about a year.  When I did, I got the sound of a straining transformer (a muted buzz) and a bad smell, and two of the outputs were zero.  On investigation two of the tantalum caps on the outputs had gone short.  I replaced them with higher rated ones (I think 22 uf 35V) and it's been fine since.  Some time ago I read an article - I think in these very pages - about tantalum caps doing exactly this - failing after 20+ years.  I was conforted by the fact that I was not alone :-) but I've regarded tantys with some suspicion ever since.

Tantalum is mostly mined in the Congo. Extraction of the mineral is subject to murder, coercion, child soldiers and child labour. Use of it is usbject to morals and legislative requirements. See this article (which is one of many).

There are some alternatives like niobium.

Quite a lot to consider when all you want is a 22uF cap to quickly resolve a problem.

@antedeluvian: Tantalum is mostly mined in the Congo. Extraction of the mineral is subject to murder, coercion, child soldiers and child labour ...

I had no idea! I don't know what to say.

@antedeluvian:  Extraction of the mineral is subject to murder, coercion, child soldiers and child labour ...

You're absolutely correct.  When this came to light there were measures put in place to obtain tantalite (tantalum ore) only from "conflict free" sources.  Unfortunately, there weren't many.   Look at the effect on the price in 2011 when this came to be public knowledge: http://www.infomine.com/investment/metal-prices/tantalite-ore/all/

@Antedeluvian: "A reported problem with MLCCs and one that we have experienced here at work is that when a MLCC is flexed, the capacitor cracks and fails."

I was initially concerned about that too, but after using tens of thousands of MLCC's, the only time this was an issue was when I absent-mindedly placed a cap too close to a hole where a broaching spacer was to be pressed into the PCB. Those caps were perfectly justified in cracking.   Moving them away from the hole by a quarter inch resolved the cracking problem.

We've also encountered this problem where I work but only for certain capacitors placed in high stress locations. the solution is usually to modify the board layout to move the capacitor away from the stress zone. In a few cases, we've resorted to buying the "flex cap" parts when it was not practcal or cost effective to re-layout the board.

The "flex caps" are more expensive than normal caps but fix the problem and are  less expensive than board rework or a field failure.

I've witnessed the effect of voltage factor on MLCC decoupling a motor controller/driver running at 12V; 1V pk-pk input ripple at 100's of KHz on 22uF.  Try to pick the largest package the local PCB space can bear and lay the cap perpendicular to the long axis of a rectangular PCB, in addition to locating away from edges, mounting bosses, connectors, etc. Also beware that MLCCs can be microphonic; Tantalums, alas, are the easiest practical solution for decoupling sensitive analog front-ends and PLL's that may be subject to vibration, especially in the audio and ultrasonic spectrums.

Had a similar experience with some really old but NOS tantalum teardrop caps I used as decoupling in a hastily constructed sub-woofer low-pass preamp filter circuit.  Prior to releasing its magic smoke, accompanied by leg separation from the package, first power yielded a brief, dazzling incandescent lightshow from the little critter.  I guess I had the current limit too high.  Failing short gets ya every time!

The Conflict-Free Sourcing Initative (CFSCI) is an organization that supports efforts to remove illegal and illegitimate sourcing of tantalum, tin, tungsten and gold from from the Congo and other conflict regions.  Here is their website: 

http://www.conflictfreesourcing.org/about/members-and-collaborations/

You can check to see if your capacitors are manufactured by one of these companies and if one of these companies are your customers.  Note that there is a reporting requirement on whether your products contain components that contain minerals from these regions per Section 1502 of the Dodd-Frank Act.  Even if the Act is repealed (Trump), there is still a complinace standard that most companies have signed up for from the Electronics Industry Citizenship Coalition (EICC).  You may want to check out that website too.

When replacing tantalum caps with MLCCs, one should consider more than just matching the capacitor value. Matching the ESRs can also be very important, depending on the application circuit. For example, many voltage regulators require output caps with a finite ESR to ensure stability; some even forbit the use of MLCCs. MLCCs have much lower ESRs than tantalums, so when using MLCCs, additional series resistors may be required for stability.

Also, power supplies that use MLCCs for decoupling are much more prone to resonances and anti-resonances (for multiple, different values caps in parallel) that can cause detrimental power supply oscillations. Again, careful consideration of the power supply distribution network (PDN) impedance is required for proper decoupling capacitor selection.

On a side note, if a MLCC ever drops on the floor or table, it's best to simply throw it out. The mechancial stress of impact can cause MLCC microfractures and internal stresses that will only get worse during high-temp reflow soldering and also make the MLCCs more sensitive to future stresses, decreasing their reliability.

Hi Ed- Thanks for the Conflict-free sourcing link.  Since we no longer use tantalum capacitors that's one less thing to worry with.  I'll take a closer look at that website to see if any of our other components might be affected.

@2Torr: "Matching the ESRs can also be very important, depending on the application circuit."

That's very interesting. I definitely need to dig deeper into this. I just went back and checked the several types of linear regulators we use and they generally say something like "Make sure to use a low ESR capacitor placed close to the input and to the output using the shortest trace possible".  None of them say anything about the ESR being too low- but since you've got me thinking about it I think I do remember an article that cautioned against using ceramic caps for a specific device because it could cause overshoot and ringing.

At the time I took that with a grain of salt, but now I need to go back and try to find that information. Cany you point me to any information about this? 

Low ESR is generally not a problem for linear regulators but some of the older switching regulators rely on the capacitor ESR to damp out the ringing that might occur between the inductor and capacitor. Also some early LDOs might have a problem as well. Most newer parts are fine with low ESR but some of the older parts are still in common use so it pays to check your datahseets.

I agree with elizabethsimon. It's always good to follow the datasheet recommendations. Most of the stability concerns are with using ultra-low ESR MLCCs, but even regular MLCCs with low ESR can cause issues.

Here's a link to a TI App Note that details stability concerns with LDO regulators using output MLCCs:

AN-1482 LDO Regulator Stability Using Ceramic Output Capacitors

http://www.ti.com/lit/an/snva167a/snva167a.pdf

And for switching regulators, an App Note from Rohm may be helpful:

The Important Points of Multi-layer Ceramic Capacitor Used in Buck Converter circuit

http://rohmfs.rohm.com/en/products/databook/applinote/ic/power/switching_regulator/cera_cap_appli-e.pdf

and for MLCCs used on the input of DC-DC converters:

Ceramic Input Capacitors Can Cause Overvoltage Transients

http://www.linear.com/docs/24956

Here's another good overview App Note:

Engineers note: Capacitors are key to voltage regulator design

http://www.ti.com/general/docs/lit/getliterature.tsp?baseLiteratureNumber=snoa842

Another issue with large value MLCC's is leakage. They can tend to leak internally due to porosity through their dielectric ( pinholes formed during the screenprinting process ) and internal cracks and defects. This may not be noticeable with line-powered circuits but battery-operated low power circuits will die prematurely and finding the root cause can be a real bear. For battery-powered circuits I still prefer tantalum for larger value caps.

Thanks, 2Torr for those links.  There's some really valuable information there. I downloaded all of those app notes (the one from your other post too) and will keep local copies in my "Interesting documents" file.  AN-1482 was particularly informative.  

I learned some important stuff here.

@elizabethsimon: "Most newer parts are fine with low ESR but some of the older parts are still in common use so it pays to check your datahseets."

About a year ago I was having lunch with a semiconductor sales rep and the topic of bypass capacitors came up. I asked why many of the spec sheets for regulators specifically called for tantalums when it looked like ceramics would be more cost effective.  He told me that in most cases it was because the spec sheets were written when tantalums were cheap and plentiful and that ceramics would always be a good alternative.  Based on the information that has come to light in this thread it's obvious his advice was wrong!   

I find I remember design tips like these better when there's a bit of humor associated with them.  Although there's nothing funny about an LDO going unstable due to ESR,  I recall reading an app note or datasheet long ago (can't remember where) where the LDO's ESR stability plot with the widening at one end was described as "the Horn of Doom"  Any readers remember that one, and where?

From a non-EE: how do you test capacitors for functionality?

I know more or less which bits of a multitester measure voltage and resistance, but apart from looking for a short by checking for a low resistance between the ends, what tests could one apply? Is there a special kind of meter? (Especially to check for particular values, rather than go/no go.)

what tests could one apply? Is there a special kind of meter?

There are all kinds of possibilities, and to some degree it matters whether you have leaded or SMD parts and if the capacitors are in circuit.

You can start with the note "How to measure capacitance " from Fluke.

You can get DMM with the ability to measure capacitance- there are many, just search on Google. The Extech EX363 is one example.

In labs it is quite common to see a LCR (inductance, capacitance, resistance) meter. Again there are quite a few manufacturers.

Finally there are tweezer devices that will measure SMD devices in circuit.

Keep in mind that the value measure in circuit is affected by all the devices and parasitic effects connected to the capacitor 

Added after 10 minutes: Also there should be no power applied when you attempt to measure. On larger capacitors  it is also a good idea to connect a resistor (say 10K) across the terminals to allow them to discharge before you try to measure.

Many hand held and benchtop dmms have a capacitance test function, but testing the caps in circuit can lead to false readings.  Using a Fluke 177 meter on a board I have handy, two 106 tantalums read 17uF & 11uF, and 85uF & 9590uF for the two caps and both lead polarities.

When doing lab work and suspecting bad caps, I would always look at the voltage waveforms on a scope first, if they did not look right I would unsolder the caps and check them out of circuit with a meter (or just try replacing it).

Ceramic Input Capacitors Can Cause Overvoltage Transients http://www.linear.com/docs/24956

I have seen that first hand, on SSD drives which were supposed to be hot-pluggable.  We had some field returns where the regulator(s) were damaged and I was trying to understand and reproduce the failure.  The regulators had recently been changed from medium voltage (5-24V Vin) to low voltage (5-7V Vin) for cost and efficiency reasons.  I was able to reproduce a voltage surge (2x Vin) during the hot-plug event, so did our pilot manufacturing/test facility once instrumented to look for it.  The 2xVin (=10V) took the low voltage regulators out occasionaly whereas the older medium voltage could sustain the transient event fine.  The input and output caps were MLCC.

If you're Max you get an affordable component tester. See his review here:

http://www.eetimes.com/author.asp?section_id=30&doc_id=1329331

This reminds me that I want to add one to my shopping list...

Or if you have lots of money you might get an impedance analyzer which will tell you more than you want to know unless you're an EE :)

As has been mentioned in the other responses, none of these work with capacitors in circuit.

Thanks, Elizabeth and everyone else, for the answers.

It's a shame that capacitors can't be tested in-circuit; that makes it tedious to track down a problem suspected to be a bad one.  (Barring visual checks for horrible chemical effusions.)

@perl_geek "makes it tedious to track down a problem suspected to be a bad one. " It's actually a little worse than that. There are no identifying marks on ceramic capacitors, so it is impossible to know if there has been an assembly snafu.

The series R of a tantalum is beneficial with LRC filter circuits to control the "Q" and prevent ringing with transients.  This is much more space efficient than adding an extra resistor.  I agree that most "bypass" applications, right next to an IC should be ceramic.  For higher current devices, you may also need a resevoir to "refill" the smaller bypass caps at a larger time constant.  That is where the tantalums come in handy.

Max, if you have been around board design for much time you surely have a tantalum cap explosion story to relate. Back in the early 90's we used them with abondon. All over the cards of the time.

Once in the lab we had a large epoxy coated tantalum let loose and flaming balls of molten material erupted fom the board. The burning material proceeded to wedge between the wall and a mains electrical conduit. The tantalum flamed away and burnt the wall and left a terrible smell with smoke. The burnt wall was not repaired for years and the site was a constant reminder of the dangers of tantalums installed incorrectly.

On the boards - We fixed it with three terminal caps at the time. The two outside terminals were ground and the center the positive leg. No matter how you installed the caps they were correct by assembly.

Although it is about bypass capacitors it does have a bit more general information as well

Choosing and Using Bypass Capacitors

You must Register or Login to post a comment.

This site uses Akismet to reduce spam. Learn how your comment data is processed.