Tantalum Capacitors

That is somewhat of a crazy schematic. I bet some of the NU resistors could be some 0 ohm resistors. They are mainly jumper wires made inside a blob of material that resembles a resistor in size. They have been used to be in inserted by machine to jump over places where a circuit trace can not be made.

No, we did it locally, every IC that was uncoupled with a tantalum, we put a small resistor from the power rail to the tantalum and the problem was solved. I clearly remember reading this advice in some tantalum's datasheet or application note. As those IC's consume very little, the DC voltage drop over the resistance was negligible, but the reduction in current spikes through the tantalums was considerable. They didn't blow up anymore and neither did the IGBTs.

joe

Can somebody decode "postprocessing fixture" back into English? I have idea what the author is talking about or what the "fix" really was.

He ramped up the supply slowly with a current limit, to give the tants a chance to clear the damage.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Is it about clearing damage or just to show the added resistance prevented the failure? I found the story a bit hard to follow.

--

Rick

The point being that by adding the resistor, you've increased the "ESR" of the cap(-resistor). You might just as well use an aluminum cap in its place if ESR doesn't matter.

No there is still a low-ESR decoupling for the IC. I do that a lot since it isolates the IC rail from spikes on the main power rail. (Not with tants though).

--

John Devereux

He is talking about this...

Vcc ---/\/\/\---+-------+ | | IC = | | --- --- - - Not this...

Vcc ----+----/\/\/\-----+ | | IC = | | --- --- - -

--

Rick

Really? Click on the graph near the bottom of the page. Note that the low-ESR aluminum electrolytic is only slightly worse than the equivalent tantalum. If polymer caps were added to the graph, it would be about the same as tantalum.

The reason we used tantalum (in marine radios) over electrolytics was that they were smaller, lasted longer, were sealed, more stable capacitance over temperature, more stable ESR over temperature, and were easier to handle in a wave solder and washing machine environment. While each benefit for tantalum is admittedly minor, the combination of all the aforementioned benefits made them quite a superior device. The only downside was the cost, which limited their use to areas where a very low ESR was needed.

Your "current limiting resistor" sounds like something that would raise the ESR of the device by the resistor value. Much depends on the ripple current, which presumably in a switching power supply filter cap, is quite high.

When playing with ESR's of less than 1 ohm, minor variables such as PCB plating thickness and trace width/length become significant. When the lowest ESR is at the series self resonant frequency of the capacitor, the selection of type, value, voltage, package, etc also become important. Much of the RF circuitry involved in a radio requires broadband bypassing. That rapidly becomes an exercise in capacitor selection based on series resonant frequencies and lowest overall ESR. It was not unusual to have 3 different bypass caps in parallel at key locations, such as the corners of PCB's to chassis ground points. Adding a series resistor to the tantalum cap would not have worked for obtaining the lowest possible ESR.

Did the ripple voltage on the power supply line increase with the added series resistance?

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I guess I should mention that Intech had two divisions. I worked for the marine radio division. There was also a "modular products" division that made military grade modules (A/D, D/A, amps, etc). Both divisions shared many of the same components including a wide selection of tantalum caps. Most of the modules ran on +15/-15 VDC and used 35 VDC rated tantalums. As I vaguely recall, there were no aluminum caps used in anything that had to work from -40C to +105C. If tantalums were that failure prone, they would never have survived in a mil spec environment.

Some typical radio boards. This is Intech M3600 2-30 MHz 150 watt PEP synthesized SSB radio circa 1977(?): The boards are a mix of purple potted electrolytics and blue or orange colored tantalums. No failures in 10 years of similar radios.

There also seems to be an aging mechanism involved. I'm the not so proud owner of several Wavetek 3000B service monitors: Some of the tantalums have shorted over the years. I'm replacing them as I blunder forward. No fires, smoke, or discoloration in about 10 years of fixing these. I also have some other equipment with similar tantalum problems. I recently repaired an M3600 radio which showed no evidence of deteriorating or failed tantalums. Has something changed in the last 40 years in how tantalums are made?

I think you mean dI/dT which provides the heating necessary to ignite the tantalum. That all sounds logical, but doesn't explain why a similar amount of heating caused by normal ripple current doesn't set fire to the capacitor. I've seen some heat darkened tantalums operating normally without ignition. Like the bulging electrolytics and burning LiIon batteries, I suspect there's been some changes in production methods (like skipping important steps to save pennies).

Nice photos. Having done post mortem failure analysis a few times, I like to look at the damage and try to guess what was the cause. It's fairly easy to inspect the remains and estimate the violence of the failure. For tantalum, there's usually something left of the wire leads or carbonized slug. It gets hot, belches flames, carbonizes, falls apart, which finally breaks the connection.

However, the OP mentioned that: "The first thing I notice when looking inside is that the small SMT 100uF 10V tantalum capacitor C109 has completely vacated - it appears to be gone, blown right off the board. There are some little fragments rattling around in the case." That's not what I consider to be a conventional tantalum burn failure. The cap should have looked like the one in the above photos. Something caused this one to explode rather than burn, which is why I suggested that a much higher voltage wall wart was involved. I don't think the tantalum was at fault simply because it was the first thing to blow and was the most obvious physical failure.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

An aluminum electro in parallel with a smaller ceramic makes a nice lead-lag network for the voltage regulator, if you do it right.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

One thing that doesn't show up on the power supply schematic sheet is the zillions of chip bypass caps on other sheets. One might have tens of uF of paralleled super-low ESR ceramic caps on a big power pour.

A tantalum seems to have the right ESR to damp the whole mess, even when a regulator should be unstable with just the ceramics.

One test for stability is to apply a pulse load and see how the regulator reacts.

I saw one big board (part of an Anritsu DRAM tester system) that had

3000 bypass caps. It's not unusual to see an FPGA appnote that recommends a hundred caps or so per chip.

Look at it like this (you might switch to fixed width font). ____

• ---|____|----------------- small | R |tant --- --- --> to IC Vcc/GND | |

- --------------------------

It's a filter now. Still the same (or even better) ripple suppression. (Almost) no lowering in supply voltage for the IC. Tantalum as near as possible to the IC in order to take account of the EMC induced voltages also, which are caused by the HF high current switching.

joe

Right. It was (at that time) even recommended.

joe

Correct.

joe

}snip{

Are you sure? I understood that the I=C.dV/dt was responsible for the damage. The higher dV/dt, the higher I. And the longer the dV/dt continues, the hotter the C gets.

}snip{

joe

Yes

}sorry for snipping, but I hate scrolling a lot{

If you read the post carefully, you'd see that between power supply and C there is a small R, which gives a tiny reduction of voltage to the IC, and then there is the C // Vcc-GND. So the IC sees the same (or even smaller) ripple from the power supply, EMC suppression is still the same (or even better due to more damping of the ringing) and everything went fine from then on.

Yes. And all those factors are uncertain to a high degree. Therefore it could be advisable to just add a resistor of a known value, if your design permits it.

Indeed, not in that case. But where we had the problem, it worked well.

Indeed.

After finding out the problem, we (finally) read up to the specs and recommendations, found the suggestion to add small R's, did it, and the problem was gone. We did not measure ripple voltage. I guess that would have been useless because the problem was caused by EMC, and the measuring probe would probably have suffered the same problem and not given the right picture.

joe

No, the point is that there is no reason to use a tantalum cap if you're going to blow its ESR with a resistor. Just use an aluminum.

Nevermind!

I sit corrected. ;-)