Tantalum Relpacement - decoupling

[LabSpokane]

I know that this is a very tired subject for some of you, but what is the best bet for replacing a tantalum decoupling cap?  Let's say 10uF on a 5V or 3.3V power pin.  Is a MLCC plus a series resistor for potential damping needs a interchangeable solution?  If I don't need damping I can just stuff a zero ohm in and not worry about it.

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

Thanks in advance!

[tautech]

I know that this is a very tired subject for some of you, but what is the best bet for replacing a tantalum decoupling cap?  Let's say 10uF on a 5V or 3.3V power pin.  Is a MLCC plus a series resistor for potential damping needs a interchangeable solution?  If I don't need damping I can just stuff a zero ohm in and not worry about it.

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

Thanks in advance!

As much as many of us hate them, when used carefully and of recent manufacture they are fine.
Care must be taken to select a much higher voltage headroom that you might with any other cap.
For this reason I'll use only 50V Tants and never subject them to more than ~ 1/2 that.

Because they are fast they can serve as both decoupling and bulk capacitance, that's useful when you're pushed for space. 

[T3sl4co1l]

I know that this is a very tired subject for some of you, but what is the best bet for replacing a tantalum decoupling cap?  Let's say 10uF on a 5V or 3.3V power pin.  Is a MLCC plus a series resistor for potential damping needs a interchangeable solution?  If I don't need damping I can just stuff a zero ohm in and not worry about it.

Yeah, that's fine.

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

Thanks in advance!

On supply bypass, I wouldn't think so.  Do mind that ceramics vary quite strongly, which might upset knife-edge designs.

For other applications, tantalum isn't a hazard: timing and coupling capacitors benefit from the reduced leakage* and excellent stability.  They can't run away and ignite without a high current supply to kick them off (which is why supply bypasses are so notorious).  For both these reasons, and others (microphony might be a problem!), (type 2) ceramic isn't so good in such applications.

*Except for intermittent "self healing" spikes, apparently.

Tim

[Zero999]

There's nothing wrong with tantalum capacitors. They're certainly more reliable than aluminium capacitors.

[nctnico]

There's nothing wrong with tantalum capacitors.

Everything is wrong with tantalum caps. They consist of fuel and oxidizer. See what just happened in China if you store those 2 close together. A tantalum can easely burn a hole through a PCB rendering the entire PCB into scrap.

[Zero999]

There's nothing wrong with tantalum capacitors.

Everything is wrong with tantalum caps. They consist of fuel and oxidizer. See what just happened in China if you store those 2 close together. A tantalum can easely burn a hole through a PCB rendering the entire PCB into scrap.

That's only when tantalum capacitors are not used properly: don't subject them to surges and they're fine and include proper over-current protection and it won't catch fire.

On the other hand aluminium capacitors the electrolyte can evaporate over time, the ESR will increase and capacitance decrease, leading to eventual failure, irrespective of how well the circuit is designed. On the other hand tantalum capacitors are solid and very reliable with no electrolyte to evaporate.

[vk6zgo]

I know that this is a very tired subject for some of you, but what is the best bet for replacing a tantalum decoupling cap?  Let's say 10uF on a 5V or 3.3V power pin.  Is a MLCC plus a series resistor for potential damping needs a interchangeable solution?  If I don't need damping I can just stuff a zero ohm in and not worry about it.

Yeah, that's fine.

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

Thanks in advance!

On supply bypass, I wouldn't think so.  Do mind that ceramics vary quite strongly, which might upset knife-edge designs.

For other applications, tantalum isn't a hazard: timing and coupling capacitors benefit from the reduced leakage* and excellent stability.  They can't run away and ignite without a high current supply to kick them off (which is why supply bypasses are so notorious).  For both these reasons, and others (microphony might be a problem!), (type 2) ceramic isn't so good in such applications.

*Except for intermittent "self healing" spikes, apparently.

Tim

"Excellent stability"?

Relative to an Electrolytic,maybe!
Back in the day,we used a LM567 tone decoder to turn TV transmitters "off" & "on" by removing & restoring the field syncs on a PAL TV signal.

We experienced a number of phantom turn offs in the middle of the day.
It turned out to be the tantalum used in the 50Hz CR network drifting in capacitance with temperature,causing the decoder to unlock.
"Shoe-horning" a massive Polycarbonate into the available space on the board fixed the problem permanrntly.

[Bud]

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

It is when you need to know that the capacitance in the operating mode will not change from the capacitor value. Ceramics experience horrible capacitance drop under DC. Tantalums do not do that. If a circuit was designed for minimum required capacitance (i.e. input/output LDO capacitors), using ceramics may result in instability.

[T3sl4co1l]

"Excellent stability"?

Relative to an Electrolytic,maybe!

Yes.  +80/-20% doesn't do a circuit much good, nor do Z5U ceramics, or even X7R when voltage and temperature changes are included.

Back in the day,we used a LM567 tone decoder to turn TV transmitters "off" & "on" by removing & restoring the field syncs on a PAL TV signal.

We experienced a number of phantom turn offs in the middle of the day.
It turned out to be the tantalum used in the 50Hz CR network drifting in capacitance with temperature,causing the decoder to unlock.
"Shoe-horning" a massive Polycarbonate into the available space on the board fixed the problem permanrntly.

A precision TV timing circuit definitely needs to be film.  PP, PC, PS or PPS would all be fine.

Tantalum has a pretty good positive tempco, see Fig.2:
http://www.kemet.com/Lists/TechnicalArticles/Attachments/93/2008-11%20Update%20-%20Ceramic%20versus%20Tantalum.pdf
Arguably, that might help in a power supply situation.  It's certainly better than the freezing-out-of-ESR that electrolytics do.

Tim

[technix]

For a few reasons I generally avoid using tantalum caps:

1) As mentioned they are NOT chemically stable. They can easily burst into flames in the fashion of thermite reaction if some surge got leaked in - for example when inductively or capacitively coupled into the circuit.
2) Tantalum is conflict mineral. I live in China so maybe I can source tantalum caps produced using local material (China does have tantalum mines, operated by state-owned companies) but I do have the concern about ending up funding armed conflict, inhumane mining or destruction of local environment if I used them.
3) MLCC and polymer capacitors actually have even lower ESR and for polymer caps better stability than tantalums. I have once designed a SMPS with MLCC output filter cap and in breadboarding MLCC performed better than both tantalum caps and Nippon Chemi-Con aluminum caps. I have secured a source of 50V 10µF MLCC in 0805 size and those will become my daily driver SMPS output cap.

[nctnico]

There's nothing wrong with tantalum capacitors.

Everything is wrong with tantalum caps. They consist of fuel and oxidizer. See what just happened in China if you store those 2 close together. A tantalum can easely burn a hole through a PCB rendering the entire PCB into scrap.

That's only when tantalum capacitors are not used properly: don't subject them to surges and they're fine and include proper over-current protection and it won't catch fire.

On the other hand aluminium capacitors the electrolyte can evaporate over time, the ESR will increase and capacitance decrease, leading to eventual failure, irrespective of how well the circuit is designed. On the other hand tantalum capacitors are solid and very reliable with no electrolyte to evaporate.

Electrolytics can be choosen/calculate to give a piece of electronics decades of service life.

[technix]

Are there any situations where a tantalum is still the only game in town?  Aside from needing the most capacitance in the smallest device possible?

It is when you need to know that the capacitance in the operating mode will not change from the capacitor value. Ceramics experience horrible capacitance drop under DC. Tantalums do not do that. If a circuit was designed for minimum required capacitance (i.e. input/output LDO capacitors), using ceramics may result in instability.

You can horribly derate MLCC voltage to avoid this culprit - my 5.3V SMPS used several 50V 10uF MLCC in parallel as the output filter cap.

[nctnico]

If a circuit was designed for minimum required capacitance (i.e. input/output LDO capacitors), using ceramics may result in instability.

Ofcourse not. Read the datasheet of the MLCC capacitor. There is a huge variation and voltage rating doesn't mean a thing but the datasheet has all the answers to select the right MLCC capacitor.

[vk6zgo]

"Excellent stability"?

Relative to an Electrolytic,maybe!

Yes.  +80/-20% doesn't do a circuit much good, nor do Z5U ceramics, or even X7R when voltage and temperature changes are included.

Back in the day,we used a LM567 tone decoder to turn TV transmitters "off" & "on" by removing & restoring the field syncs on a PAL TV signal.

We experienced a number of phantom turn offs in the middle of the day.
It turned out to be the tantalum used in the 50Hz CR network drifting in capacitance with temperature,causing the decoder to unlock.
"Shoe-horning" a massive Polycarbonate into the available space on the board fixed the problem permanrntly.

A precision TV timing circuit definitely needs to be film.  PP, PC, PS or PPS would all be fine.

Tantalum has a pretty good positive tempco, see Fig.2:
http://www.kemet.com/Lists/TechnicalArticles/Attachments/93/2008-11%20Update%20-%20Ceramic%20versus%20Tantalum.pdf
Arguably, that might help in a power supply situation.  It's certainly better than the freezing-out-of-ESR that electrolytics do.

Tim

At that time,Tantalums were the "latest & greatest"---they were being designed into all sorts of things,due to their small size,low leakage,etc.
They were thus,an obvious candidate for a  compact 50Hz  tone decoder,as the capacitor would otherwise be quite large.

This device only determined the presence or otherwise of vertical syncs,so was not what you would call a"precision TV timing circuit"
The LM567 at such low frequencies has a fairly broad capture range,but not quite wide enough to handle the "Tant's" drift.

I have also encountered many "cooked" Tantalums in power circuits,to the extent that I replace them with new capacitors of the same type if :-
(a) they look as if they have got warm sometime,& (b) if they are of an older packaging.
Re: this latter,I have noticed that some of the early Tantalum packages are more prone to failure than others.

Where time is money,it is easier & cheaper to replace suspect Electrolytics & Tantalums  "on spec".

A lot of equipment has very low value Electrolytics (around 0.47----2.0uF).
Replacement Electrolytics of such values are unobtainable anywhere I have looked.

Even Tants around those values are not always stock items,& Film caps are too large.
I have often replaced such small value Electrolytics with MLCCs with no problems.

[Psi]

A lot of modern Tantalums from reputable manufactures actually state in the datasheet that its ok to run them continuously at X volts. Where X is a typical derating you would find on many other components like 10-20%.
I guess they are trying to reassure engineers that its ok to use their tantalums without massive 50% derating.

eg, 6.3V tantalum are designed for 20% derating on 5V rails.

I think a lot of the panic over derating a tantalum comes from dodgy manufactures, fakes and/or old tantalum technology (new ones are miles better)

[Howardlong]

3) MLCC and polymer capacitors actually have even lower ESR and for polymer caps better stability than tantalums. I have once designed a SMPS with MLCC output filter cap and in breadboarding MLCC performed better than both tantalum caps and Nippon Chemi-Con aluminum caps. I have secured a source of 50V 10µF MLCC in 0805 size and those will become my daily driver SMPS output cap.

Aluminium Polymer caps are definitely worth a look, particularly for SMPS, in the 47uF to 1000uF or so range. Exceptionally low ESR, and solid dielectric so no drying out. No need to derate voltage either, at least not to the extent of some other cap technologies. Pricing is now reasonable too, so rather than multiple parallel caps so often seen on SMPS to reduce ESR, often a single Al Poly will suffice.

http://www.digikey.co.uk/Web%20Export/Supplier%20Content/Panasonic_10/PDF/panasonic-understanding-polymer-hybrid-caps.pdf

[vk6zgo]

I was a bit puzzled by the apparently well-known capacitance variation of ceramic capacitors with voltage,until I read the information linked by AcHmed99.

After discussing this effect,it goes on to say that such an effect was "rarely noticed with the thicker dielectrics of five and ten years ago".
 
This,of course,explains why it wasn't common knowledge back when I was replacing "Unobtainium" Electros with so-called "monolithic" ceramics.
These "through the hole" components,which I have always assumed were MLCCs with leads attached.would probably have used that earlier construction.

Logically,currently available"through the hole" components must use the same technology as SMD's.so they probably exhibit the same behaviour,& are subject to the same limitations.
Does anybody know if this is so?

[T3sl4co1l]

These "through the hole" components,which I have always assumed were MLCCs with leads attached.would probably have used that earlier construction.

Logically,currently available"through the hole" components must use the same technology as SMD's.so they probably exhibit the same behaviour,& are subject to the same limitations.
Does anybody know if this is so?

That is precisely the case.

By the way, for anyone shopping for their best choice of capacitor:

For a given nominal capacitance rating, get the highest voltage available (usually topping out at 50V), in the desired package size.  Examples:
All these are X7R.  Don't get worse than this if you have the option (X5R has a smaller temp range but same spec otherwise; anything Z5U or related is just crap).
50V 0.1uF 0805 -- 0.1 might also be available in 50V for smaller sizes (0603, 0402?), but whatever the highest voltage is, get that.
10uF 0805 -- 16V is usually tops, I think?
10uF 1206 -- 35 or 50V
47uF 16V 1210, maybe 1812

Don't get bigger than 1812, they're much more liable to crack.  IPC doesn't even really recommend going over 1210, but that might be coming from a maximum reliability (military?) angle.

C0G:
Small values (up to 1nF is economical): doesn't matter
2.2nF 630V 1206
10nF 630V 1812
Most economical for higher voltages and high dV/dt.  High voltage kind of precludes small package sizes (1206 at 630V or more is already pushing it).

FYI: type 1 dielectrics exhibit very little to no voltage effect.  I've blasted "100pF 50V" C0G disks at 500VAC (and as many kHz) with no effect (they got slightly warm from the current, that's all; a 2kV Y5P in the same place barely lasted seconds before effectively shorting out the oscillator).  Not that I would recommend or condone such abuse of ratings, but they are very robust, and well worth the cost when needed.

As for type 2 dielectric variability, I don't know that they've advanced all that much in a decade, but certainly since the early days.  I have some truly monstrous 120pF NP0 (now called C0G) disk caps.  Probably rated 500V, but probably don't break down until a spark physically jumps the pins (5-10kV).  Even the Z5Us were terrible in size.  But aside from tolerance (Z5U has always been the same +80/-20% definition), no one much cared about voltage coefficient or aging.  And they might not've noticed, because the voltage coefficient on such a huge capacitor is pretty small.

Also back in the day, voltage ratings might've been for safe operation AND meeting tolerance.  I made this plot from an unknown ceramic cap,



only labeled "223Z".  If it's rated less than 25V, it still meets spec (well, not currently.. but would've, and probably would again after additional annealing?), and there's no problem.

So where they got off removing voltage from the tolerance, if it was ever used in the first place, I don't know.  It's absolutely true that there's no relation today, and one must be constantly mindful of this.

So, if you're still wondering, what's the difference between voltage ratings, anyway?  Number of layers, and thickness of layers.  The chip cap is solid ceramic, through and through, so you're paying for that regardless -- what you're paying for is the number of layers (a production time cost, and a small cost in palladium or whatever they use), which is why there's a small difference in price across voltage rating (for a given capacitance; or vice versa).

The reason I would suggest getting the biggest (highest C, V for a given size; but also the most expensive) is because, this combination pushes the material limitations, so that you are most likely to get capacitors with a similar or equal curve of C(V in %), so you can safely use the same adjustment factors (rated voltage = 2-3x nominal) and get the same expected performance.

You can always cost-reduce from there, if you need to.

Tim