"N_Cook" wrote in message
...

On Mar 18, 10:24 pm, "Arfa Daily" wrote:
Anyone read anywhere what the failure mechanism is for solid dielectric
caps
of the tantalum variety ? In my experience, no matter what the value,
working voltage, or format (bead, bullet or box), they always seem to
fail
leaky. Not open or short (well, very occasionally short). Just leaky.

Example. Today, I had a Mesa Boogie combo cross my bench. Very odd
problem
in that when the 80Hz slider in the graphic was advanced in the 'boost'
direction, the audio suddenly went very distorted, and then disappeared.
There was also a slight 'scratchiness' to this pot, which did not feel
like
a bad or dirty track.

When I got the graphic pots board out, it was actually quite a simple
affair, with each of the 6 bands having just a pot, one resistor, one
choke,
and one cap. On the 80Hz channel, this cap was a 3u3 tantalum bullet, and
it
was 2k leaky. Why ? The device is under no voltage stress at all in this
position, being subject to low signal levels only. I wonder if it's some
kind of internal 'growth' like the dreaded tin whiskers, which causes it
?

A new cap (used a 3u3 tant bead that I had in stock) restored normal
operation of the equaliser, and all scratchiness in that band
disappeared.

Arfa


wrote in message
...
I recall hearing a long time ago that tantalum caps do in fact tend to
grow metallic whiskers inside, between the plates, which ultimately
bridge the cap. Whether it's tantalum or some other metal used in
construction, it is a poor conductor, and in a power supply, will
immediately burn out and the cap will return to normal. If the cap is
left for a long long time without power applied, the bridge can become
quite substantial, and the heat generated next time power is applied
can blow the cap apart. I remember the moral of the story is that
tantalum caps are good only for decoupling and power filtering, where
there is guaranteed to be sufficient current to deal with the shorts.
They are attractive for low signal applications because of their size,
but I've never used them for the above reasons.

Now, is any of this true? I have no idea. But it would account for Ta
caps in signal applications failing in a leaky manner.


Indeed it would ...


some useful info here
http://my.execpc.com/~endlr/reliability.html



Very interesting

Arfa