Tantalum substitute

[ezalys]

Hey all, I'm just looking for a substitute for the TAJB106K016R 10 uF caps used for supply decoupling in some high end analog devices dev boards. I've been suggested to replace them with high voltage spec'd MLCCs, but if these are a serviceable substitute, why would analog have spec'd tantalums to begin with? Have MLCCs improved substantially in the last 5-10 years or something?

[mikeselectricstuff]

Some appnotes are very old, or derived from older ones, so quite possible that 10uf ceramics were expensive or not available when the appnote was written.

[KT88]

Tantals have more ESR than MLCCs. This could provide dampening to circuits that tend to ring.
At the output of some regulators the ESR helps to stabilize the loop.
An MLCC in series with a resistor could provide some artificial ESR. It could be bridged if it turns out not to be nescesary...
Additionally I agree with Mike's statement.

Cheers

Andreas

[Jay_Diddy_B]

Hi,

KT88 +1.

You need to be a little careful. The original parts have quite a high ESR. The designer may have used the ESR to provide damping:




You might be better off with electrolytics.

Can you share ADI Dev board No.?

If you can use MLCCs use them at 1/2 the rated voltage or less. They loose capacitance with bias voltage.

Jay_Diddy_B

[KT88]

You need to be a little careful. The original parts have quite a high ESR. The designer may have used the ESR to provide damping

...that's what I stated.
At the voltage range of tantals I would prefer to use MLCCs + additional resistance for several reasons:
- The resistance (ESR) can be dialed in to the optimum value.
- It won't degarde over time like an electrolytic.
- More power (ripple current) can be handled without degradation.
However if cost ist the biggest concern, an electrolytic might be a better choice...

[Jay_Diddy_B]

@KT88

We are in agreement.

Jay_Diddy_B

[maginnovision]

Is there an issue with using properly (de)rated tantalums?

[ezalys]

EVAL-AD5780

https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/eval-ad5780.html

is the board I'm thinking of.

[Jay_Diddy_B]

Is there an issue with using properly (de)rated tantalums?

There is not really a big technical issue. Engineers have used tantalum for a long time.

One issue is tantalum is a 'conflict mineral'. This means you may have to certify that you source of tantalum is 'clean'.

Google 'conflict mineral' if you want to learn more.

In many cases there is no need to use tantalum and other solutions are better. You just have to watch out for power supplies that rely on ESR for loop stability and designs where the ESR is used for damping resonances.


Jay_Diddy_B

[Jay_Diddy_B]

EVAL-AD5780

https://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/eval-ad5780.html

is the board I'm thinking of.

I looked at the user guide, UG-256, for that board I didn't see any places where I would be concerned about using MLCC caps in place of the tantalum caps.

Make sure that you use 25V MLCC caps.

Check that the voltage regulators that you are using for VDD and VSS are okay with ceramic output capacitors.

Jay_Diddy_B

[KT88]

I fully agree with Jay_Diddy_B on the conflict minerals...

The EVAL-AD5780 is an almost obvious example for the dampening purpose of the tantalum caps. Ther are ferrite beads in use which already provide some dampening - but not enough. The second part of the puzzle is the high-res SAR-ADC. These parts create very nasty current spikes that excite every resonant circuit on the board - and there are a lot...
The ESR of the caps can absorb a lot of this energy pretty quickly.
There is one possble trap though: multilayer caps have a significantly lower capacitance when biased - in some cases down to around 20% of the nominal capacitance...

[Vovk_Z]

If I'm not wrong TAJ series is of a general use type and easy to get. I don't see a reason to change it to anything else (non-tantalum).
There must be a strong reason to change it to non-tantalum type. And of cause you need to check if that other type can be used there taking into consideration all the parameters (ESR and it's stability, if it can be lower, resonating frequency and signal frequency range, temperature range, humidity variation, vibration, ... etc).

[Jay_Diddy_B]

Hi,

Here is a quick small signal LTspice model with the Wurth 74279204 Ferrite Bead:




This show the response to a 100mA current step with and without ESR damping. The circuit is better without the damping.

Jay_Diddy_B

[Howardlong]

If I'm not wrong TAJ series is of a general use type and easy to get. I don't see a reason to change it to anything else (non-tantalum).
There must be a strong reason to change it to non-tantalum type. And of cause you need to check if that other type can be used there taking into consideration all the parameters (ESR and it's stability, if it can be lower, resonating frequency and signal frequency range, temperature range, humidity variation, vibration, ... etc).

Reliability and failure mode? Tants typically die short, not great in a decoupling scenario, where there could well be secondary failures too as a result.

[KT88]

For a more realistic result it is better to use more complex models of your capacitors. The source for the lumped element models I used, can be found on the AVX website: http://www.avx.com/design-tools/
A look into the frequency domain tells a lot more than a pulse response plot...
For an even more realistic simulation it would require to also include trace impedances which I intentionally left out.
The result shows that there is no one size fits all.
In the example of the AD5780 the resonant peak around 100MHz may hamper the performance of the ADC significantly.

[Jay_Diddy_B]

Hi,

Can you check the AVX models?

These impedances that I have highlighted seem very high.




They are so high that if I remove those four branches that they are in, it has very little effect on the overall result:




These two models give essentially the same results. Here they are overlaid:





I have attached the model

Regards,
Jay_Diddy_B

 eevblog_ESR_JDB.asc (5.14 kB - downloaded 32 times.)

[KT88]

Jay_Diddy_B, you are right, I messed up the 10uF circuit - they are all nHs... The ohms in the 100pF circuit are correct. Thanks for spotting it!
https://spicat.avx.com/product/mlcc/chartview/04025C104KAT2A
https://spicat.avx.com/product/mlcc/chartview/12105C106KAT2A
The peak at 10MHz (not 100MHz  ) isn't as bad as before anymore - but still present.
And you are also right that the simplified model doesn't make a big difference in this case - in other cases it matters.

Cheers

Andreas

[Jay_Diddy_B]

Hi,

Still not a big deal. Those two parallel branches are for making the 'wiggle' in the impedance curve between 3 and 4 MHz.





I suspect that AVX obtained the values by curve fitting to their measured results.



Regards,
Jay_Diddy_B

[Vovk_Z]

Here is a ....
....
This show the response to a 100mA current step with and without ESR damping. The circuit is better without the damping.

But what if you'll put L2 (some inductance, like L1) in series with current source I1?

[SilverSolder]

[...]
I suspect that AVX obtained the values by curve fitting to their measured results.
[...]

Maybe we are watching the response of their probe, LOL!

[Siwastaja]

Look at the circuit:

1) If tantalum is used "for no reason", i.e., the regulator feeding it is stable with near-zero ESR, just replace with MLCC,

2) If the regulator feeding the tantalum requires ESR, use MLCC + series resistor. Sometimes a simple aluminium electrolytic is also suitable, if you need some ESR but don't need it to be as carefully controlled as in the original tantalum circuit.

3) Look if the actual capacitance value is important, or not. If important, use a large, expensive, C0G ceramic. If unimportant, an X7R or similar will do.

[Hydron]

Tantalums on the power input might also be about not blowing the thing up when you hot-plug the power. Ultra low ESR MLCCs plus some lead inductance from the PSU can spell instant overvoltage death if you plug in an already-on PSU (with output caps charged etc).

Have seen this kill stuff regularly - I've designed stuff that is meant to be hard-wired into a supply (and therefore shouldn't be at risk) yet found half a production run blown up by the guys on the factory floor not reading the test procedure and hot-plugging the power during QC testing (they'd made a cord with a switch on the DC output side - the switch dying when trying to repeatedly interrupt 48V DC prevented more carnage!).

Linear has a good app-note on it: https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf
Throwing in a few moderate-ESR electrolytics (or I assume tantalums) in parallel fixes the issue normally.

[JPortici]

If I'm not wrong TAJ series is of a general use type and easy to get. I don't see a reason to change it to anything else (non-tantalum).
There must be a strong reason to change it to non-tantalum type. And of cause you need to check if that other type can be used there taking into consideration all the parameters (ESR and it's stability, if it can be lower, resonating frequency and signal frequency range, temperature range, humidity variation, vibration, ... etc).

Reliability and failure mode? Tants typically die short, not great in a decoupling scenario, where there could well be secondary failures too as a result.

MLCC also fail short (unless specified otherwise) which is why you can often see two of them in series in automotive boards

[splin]

Boy, o boy - AD screwed up right royally there! Presumably they put a new graduate on the job?

All the capacitors, ceramic and tantalum are rated at 16V. But:     

The following external supplies must be provided:

• 5 V between the VCC and DGND inputs for the digital supply of the AD5780. Alternatively, place Link 1 in Position A to power the digital circuitry from the USB port via the SDP board (default).
• 7.5 V to 16.5 V between the VDD and AGND inputs for the positive analog supply of the AD5780.
• −2.5 V to −16.5 V between the VSS and AGND inputs for the negative analog supply of the AD5780.

So every tantalum apart from C22 and C43, and many of the ceramics can be operated above their rated voltage! And bear in mind that the minimum recommended voltage derating for tantalums is 50% so those shouldn't be used above 8V! Is this why you are looking for replacements?   

Niobium oxide are often a good replacement for tantalums having similar characteristics but are safer having much higher ignition energy. And, as far as I know, Niobium isn't a conflict mineral - but I haven't checked. They are cheap and only require 20% voltage derating. Unfortunately I don't know of any rated above 10V so are not relevant here.

On this development board 25V MLCCs should be fine for decoupling with a couple of caveats as others have pointed out:

1) Ceramics lose capacity with DC bias so don't use Y5U or Z5U - stick to X5R/X7R or COG. The tantalums are size 1210 so X7R should be OK - smaller sizes lose more capacity. A 22uF,35V X5R 1210 may well be less than 10uV at 16V. The actual value for decoupling is hardly critical in this case though.

2) Beware low ESR ceramics at power supply inputs as pointed out by Hydron. C3, C4 and C43 are the ones to watch here. It would be a shame to blow up such an expensive board over such a seemingly innocuous parts. Aluminium or organic polymer electrolytics (eg. OS-CON) caps should be a good alternative (but *not* the ultra-low esr versions for the same reason).

SDG Electronics did a video about ceramic caps at power supply inputs destroying boards:

[David Hess]

The general alternative to solid tantalum capacitors are high frequency or low ESR aluminum electrolytic capacitors with about 4 times the capacitance to achieve the same impedance.

And bear in mind that the minimum recommended voltage derating for tantalums is 50% so those shouldn't be used above 8V!

33% or even 25% voltage derating for solid tantalum capacitors is sufficient in current limited applications.