Tantalum caps

[Marco Trapanese]

A friend of mine, a good old analog engineer always talks about using
tantalum caps as bypass in power supply. I wonder if that's still a good
practice and if they needs a 100n in parallel to improve the
high-frequency behavior.

I try to avoid electrolytic caps (Al) if I don't need a large capacity.
Instead I prefer to use ceramic (up to several uF) because they *should*
have a better frequency response.

Where is the truth? I'm talking about standard power supply filter, near
an LDO or even a small DC/DC.

Thanks
Marco

 

[Phil Allison]

"Marco Trapanese"

A friend of mine, a good old analog engineer always talks about using
tantalum caps as bypass in power supply. I wonder if that's still a good
practice and if they needs a 100n in parallel to improve the
high-frequency behavior.

** It never was good advice and still is not.

Fitting a 100nF cap in parallel is totally superfluous in most cases.

I try to avoid electrolytic caps (Al) if I don't need a large capacity.
Instead I prefer to use ceramic (up to several uF) because they *should*
have a better frequency response.

** A ceramic or film cap of a few uF will have a lower ESR than an electro
( tant or Al ) of the same value. But a 100uF ( or higher ) low ESR electro
will cost way less and do the job of bypassing way better.

Where is the truth? I'm talking about standard power supply filter, near
an LDO or even a small DC/DC.

** Tants are so fragile that if there is ANY chance of a large current
spike, a short or the application of reverse voltage of even 1 volt - then
use an Al electro. Also it is wise to derate the voltage by a factor of 2
or 3 times.

Fact is, most tantalums will not stand repeated shorts when charged to their
rated voltage while Al electros laugh at such treatment.

A jaded engineer once said: " Tantalum caps are perfectly reliable, long as
you put a fuse in series and a reverse diode in parallel".


.... Phil

 

[Marco Trapanese]

Il 21/09/2011 14:24, Phil Allison ha scritto:

** It never was good advice and still is not.

Fitting a 100nF cap in parallel is totally superfluous in most cases.

The advice you're talking about is the use of tantalum cap or the 100n
in parallel?

** A ceramic or film cap of a few uF will have a lower ESR than an electro
( tant or Al ) of the same value. But a 100uF ( or higher ) low ESR electro
will cost way less and do the job of bypassing way better.

Ok, I got it.

** Tants are so fragile that if there is ANY chance of a large current
spike, a short or the application of reverse voltage of even 1 volt - then
use an Al electro. Also it is wise to derate the voltage by a factor of 2
or 3 times.

Fact is, most tantalums will not stand repeated shorts when charged to their
rated voltage while Al electros laugh at such treatment.

Good point.
So, changing the question: when would you use a tantalum cap?

A jaded engineer once said: " Tantalum caps are perfectly reliable, long as
you put a fuse in series and a reverse diode in parallel".

LOL

Marco

 

[Phil Allison]

"Marco Trapanese"

Phil Allison

The advice you're talking about is the use of tantalum cap or the 100n in
parallel?

** Both.

All a parallel 100nF cap will do is create a narrow impedance dip at about
5 MHz.

Ok, I got it.



Good point.

So, changing the question: when would you use a tantalum cap?

** They are sensibly used where the DC supply voltage is low, there is no
chance of a charge or discharge current spike nor a polarity reversal AND
the small size is crucial. Eg a mobile phone.

Low leakage apps are often touted for tants - however leakage is not
constant with temp and suffers big sample to sample variations, so best use
a film cap for that too.

A jaded engineer once said:

" Tantalum caps are perfectly reliable, long as you put a fuse in series and
a reverse diode in parallel ".

He avoided them entirely.


.... Phil

 

[Nico Coesel]

A friend of mine, a good old analog engineer always talks about using
tantalum caps as bypass in power supply. I wonder if that's still a good
practice and if they needs a 100n in parallel to improve the
high-frequency behavior.

I try to avoid electrolytic caps (Al) if I don't need a large capacity.
Instead I prefer to use ceramic (up to several uF) because they *should*
have a better frequency response.

I avoid tantalum because they are very prone to shorts. They also may
burn a hole in a PCB if you are unlucky.

Where is the truth? I'm talking about standard power supply filter, near
an LDO or even a small DC/DC.

Ceramic is a good choice but sometimes the ESR of a ceramic is too
low. For example: I like to use LM1117 LDO's but I need to put a 0.5
Ohm resistor in series with the 10uf ceramic at the output.

 

[Marco Trapanese]

Il 21/09/2011 15:44, Phil Allison ha scritto:

** They are sensibly used where the DC supply voltage is low, there is no
chance of a charge or discharge current spike nor a polarity reversal AND
the small size is crucial. Eg a mobile phone.

Low leakage apps are often touted for tants - however leakage is not
constant with temp and suffers big sample to sample variations, so best use
a film cap for that too.

Ok, thank you!
Marco

 

[Spehro Pefhany]

They detonate from high dV/dT. The MNO2 is an oxidizer and tantalum
powder is fuel. Too much pulse current ignites it. But it's erratic,
and people who have never experienced it tend to discount the
possibility.

Board layout has nothing to do with it. Design with MNO2 tantalums
should avoid high dV/dT or derate voltage by 2:1 or better 3:1.

ftp://jjlarkin.lmi.net/Fried_Tant_1.JPG

ftp://jjlarkin.lmi.net/Fried_Tant_2.JPG

IME, it's less of a problem where you have internal relatively wimpy
power supplies.

When you give the customer power supply connections or allow hot
plugging of PCBs it's more likely to show up.

 

[Spehro Pefhany]

Why would board layout affect whether a tantalum explodes?

John

Failure to put 15" of 10-mil wide 1/2 oz copper in series with the
cap.

 

[Joerg]

Why would board layout affect whether a tantalum explodes?

It usually doesn't. Tants can be fine for a few months. Then one fine
morning the birds start to chirp, fresh coffee smell wafts through the
lab ... tsst ... *POW* ... a smoke alarm starts blaring, a stench wafts
through the rooms and *POP* .. *BAM* ...

A VP of production once said "And then it was like popcorn".

 

[Joerg]

Tantalums are wonderful, reliable parts if you respect their dV/dT
failure mechanism. They are IEDs if you don't.

We had them derated. Yet still ...

Bottomline is that I don't trust those things. For timer purposes, ok,
but even there I prefer ye olde CD4060 or 74HC4060 over tantalums.

 

[Fred Bartoli]

Joerg a écrit :

We had them derated. Yet still ...

Bottomline is that I don't trust those things. For timer purposes, ok,
but even there I prefer ye olde CD4060 or 74HC4060 over tantalums.

I've designed some unusual power supplies boards with about 100 (yep,
one hundred) 33uF/25V tantalums, for bypassing +/-15V supplies. 100A RMS
into the caps.
The customer produced more than 2K boards to my knowledge which are
running almost 24/24 since 10 years and I've yet to hear of a
catastrophe scenario.
Of course it was carefully designed

 

[Joerg]

Joerg a écrit :

I've designed some unusual power supplies boards with about 100 (yep,
one hundred) 33uF/25V tantalums, for bypassing +/-15V supplies. 100A RMS
into the caps.

1A ripple each and only 40% derated? With such small capacitors that is
like Russian Roulette

The customer produced more than 2K boards to my knowledge which are
running almost 24/24 since 10 years and I've yet to hear of a
catastrophe scenario.

You were very lucky. I have seen designs where they weren't. And the
shirt of a guy with dozens of tantalum holes in it. Luckily he wore a
T-shirt underneath and it didn't go through.

Of course it was carefully designed

Could you explain that some more? What's the secret trick?

 

[Rich Grise]

A friend of mine, a good old analog engineer always talks about using
tantalum caps as bypass in power supply. I wonder if that's still a good
practice and if they needs a 100n in parallel to improve the
high-frequency behavior.

I try to avoid electrolytic caps (Al) if I don't need a large capacity.
Instead I prefer to use ceramic (up to several uF) because they *should*
have a better frequency response.

Where is the truth? I'm talking about standard power supply filter, near
an LDO or even a small DC/DC.

I haven't done a whole lot of analog, but a "rule of thumb" I've used on
most of my digital connect-the-dots boards has been about a 10 uF tantalum
at the point where the power enters the board, another at the opposite
corner, a 1uF tant. for each row of chips, and a preferably .1 uF
monolythic at each chip if I have any on hand or can grab a handful from
the stockroom when nobody's looking. ;-) Otherwise, .01 discs.

Hope This Helps!
Rich

 

[Rich Grise]

They detonate from high dV/dT.

John, why do your power supplies have such high dV/dT?

Thanks,
Rich

 

[Nico Coesel]

I like tantalum capacitors for good performance in the temperature range
and the convenient SMT packaging. Despite of horror stories told by some
people, I never had any issues with capacitors blown or an LDO
oscillation. I attribute those stories to poorly made design and/or
board layout.

I guess you never worked in a repair department Besides that
tantalums may explode when mounted wrong. Last year one blew up right
in my face when testing a board which just came in from the assembly
company. Fortunately I wear glasses otherwise I would have suffered
serious eye injuries.

 

[Joerg]

I haven't done a whole lot of analog, but a "rule of thumb" I've used on
most of my digital connect-the-dots boards has been about a 10 uF tantalum
at the point where the power enters the board, another at the opposite
corner, a 1uF tant. for each row of chips, and a preferably .1 uF
monolythic at each chip if I have any on hand or can grab a handful from
the stockroom when nobody's looking. ;-) Otherwise, .01 discs.

Yo, Rich, in case you do another one of those: 1uF and 10uF ceramic have
been invented quite a while ago. Que estan mucho mejor. Oh, and they are
cheap.

 

[John S]

A friend of mine, a good old analog engineer always talks about using
tantalum caps as bypass in power supply. I wonder if that's still a good
practice and if they needs a 100n in parallel to improve the
high-frequency behavior.

I try to avoid electrolytic caps (Al) if I don't need a large capacity.
Instead I prefer to use ceramic (up to several uF) because they *should*
have a better frequency response.

Where is the truth? I'm talking about standard power supply filter, near
an LDO or even a small DC/DC.

Thanks
Marco

Never use a tantalum. They suck. Or, actually, they blow!

 

[Rich Grise]

Never use a tantalum. They suck. Or, actually, they blow!

I've never seen a blown-up tantalum cap, and until I came hear I'd never
even heard of one.

Thanks,
Rich

 

[Phil Allison]

"Rich Grise"

I've never seen a blown-up tantalum cap, and until I came hear I'd never
even heard of one.

** Most designers shied away from them by the late 1970s - because of bad
experiences.

The horrible things are like red flags to repairers, soon as you see even
ONE on a PCB you say to yourself " well, there's the problem ".



..... Phil

 

[Guest]

They detonate from high dV/dT. The MNO2 is an oxidizer and tantalum
powder is fuel. Too much pulse current ignites it. But it's erratic,
and people who have never experienced it tend to discount the
possibility.

Board layout has nothing to do with it. Design with MNO2 tantalums
should avoid high dV/dT or derate voltage by 2:1 or better 3:1.

ftp://jjlarkin.lmi.net/Fried_Tant_1.JPG

ftp://jjlarkin.lmi.net/Fried_Tant_2.JPG

John

Yep. We used to use tantalums for everything, until we had a board with
a 470uF 16V tant across the power-in jack, for bulk decoupling. The PSU
was 9V at a couple of amps but the short-circuit transient curent was
much higher, about 18 amps for a few tens of ms IIRC. Fortunately all
the "*BANG* + smoke + flames on PSU connection" incidents happened
during initial testing rather than in front of the customer. We
replaced it with an aluminium electro - no more problems.

Even when run under very benign conditions they can go leaky - we had
another board which used the same tant to hold up the supply for some
CMOS RAM during battery replacement. Current was limited to a milliamp
or so. A few boards came back with no power to the RAM, the cap
measured a few hundred ohms. Remove the cap from the board and apply
rated voltage (ramped up slowly) from a bench PSU -> nice fireworks
display.

These were all AVX TPS series. Anyone have any better experience with
other brands?

R.