Noise in tantalum SMT caps?

A very loud bang?

--
SCNR, Joerg

http://www.analogconsultants.com/

"Spehro Pefhany"

** This a trick question ??

Like how many ohms in a coulomb ?

....... Phil

Yes I am puzzled too, although he must have some reason for asking...

I did find that tantalums have very low microphony (I could not detect it) compared to high K chip ceramics which have lots. It makes them a good choice for filtering any kind of reference voltage.

--

John Devereux

Celine Dion?

Polarity marking is backwards so it is starting to break down?

--

    Boris Mohar

I'd guess just the thermal noise of the esr, which is only a couple tenths of an ohm maybe.

John

No sub-10Hz voltage noise as aluminum electrolytics under bias are supposed to have?

I'm looking at a design (not my own) where there's an RC on the output of a voltage reference feeding a Sigma-Delta 24-bit ADC. The R is causing a full scale gain error of, oh, 17bits. 8-( Over 1%.

Is the reference input current constant with input voltage? Temperature? I doubt it. 8-(

Look at the leakage current. The electrons rattle as they go through. The more leakage, the more noise.

Dunno. I recently tested some standard alums and some tants, posted elsewhere, and the alums had serious chemical things going on inside; the tants and polymer aluminums much, much less.

It certainly varies as a function of adc activity, which can vary in some systems, even for delta-sigmas.

Well, the cap-versus-temp bahavior of a tantalum isn't great. So if there were standing DC on it, as there is here, minute temperature fluctuations will modulate C and, since CV is conserved, V would wobble around. The same thing would happen on hi-K ceramics.

The R is maybe a bad idea, if unbuffered by an opamp.

John

There most certainly is and it has been characterized as 1/f^2 down to the millihertz region. From what I gather from the research papers available on the web, most of this noise is due to thermal instability around dielectric microleakage areas which self-heal via a chemical transformation of MnO2->Mn2O3. These are characterized by a gradual buildup of noise to a peak and then a substantial drop once the healing has occurred. Also the very construction of the things makes them a hotbed of tunneling and current trapping mechanisms as well as possessing a nearly shot noise effect due to a metal-insulator-semiconductor conduction mechanism from anode to cathode. And all of these things are not even close to stationary, varying all the time with a very slow trend into deterioration with aging. This is all transparent to the majority of applications requiring bulk capacitance at low ESR in small packages, but entirely visible as a reference for a 24-bit A/D conversion application.

Ouch! Did they split the grounds? That can cause part of it. I've never had a case where that brought the ENOB down this much, but it put a measurable crimp into it.

There's probably a lot more that's not constant.

--
Regards, Joerg

http://www.analogconsultants.com/

What's the magnitude of the currents? The source impedance of the reference (or of the series resistor) then determines the noise voltage. I've done analog systems that were good to a few ppm stability, with tantalums bypassing refs and excitation amps.

The resistor is a bad idea. What's its value, Spehro?

Numbers, gentlemen!

John

I've used tantalums at a point in a noise-filtering circuit where I'd for sure have noticed noise much above 10 nanovolts/rtHz, and not seen any. If you'd like, I could make a measurement with an AD797 amplifier connected to a couple biased-up caps, driving an HP89410 analyzer, and get a noise spectral density plot. I'm curious enough about it now I may just do that anyway...

Cheers, Tom

Same experience here. Tantals are either very quiet or very loud (as in

*BANG*). Quite nicely suited for filters if the tolerance were smaller.

--
Regards, Joerg

http://www.analogconsultants.com/

FWIW I would love to see this. I have been using e.g. 2k2 feeding a

22uF tant as a reference noise filter. (Feeding a high impedance opamp input). As I mentioned chip ceramics are a disaster here because of microphony and perhaps temperature fluctuations.

--

John Devereux

Can't say that. I use ceramics all the time, never had such issues in a Ref distribution and some of that is rather sensitive.

--
Regards, Joerg

http://www.analogconsultants.com/

[...]

IIRC I measured about a mV disturbance across the cap, when tapping the PCB. This was when using some 0805 1uF X7R. I had to go through the circuit and replace them with tantalums, in the critical areas, which fixed the problem.

ref-[R]-.----- opamp> --. [C] LED ----> Photodiode -> AC amplfier -> output Basically I was looking for a small fluctuation on a large background signal.

Hook up a 1M resistor and a 10u chip ceramic, with wires, to a voltage source. Put a scope on AC across the cap. If you move the wires you strain the body of the cap - this results in a visible output voltage change.

I assume the same thing is going on when the PCB is flexed.

--

John Devereux

I'll try that. Interesting, I always give my circuit boards a good banging and never saw issues like that. Of course we don't have to use RoHS solder out here :-)

--
Regards, Joerg

http://www.analogconsultants.com/

Have you tried C0G ceramics? You can get 0.1uF in 1206 size these days, though a bit pricey.

When I get a chance I'll set up the (rather simple) noise spectral density thing on the tants. Doubt I'll do any microphonics testing on them though; it'll just be to see if they are stable under bias. Per John Larkin's comments, I'll try it in "free air" and with some sort of thermal insulation to at least slow down the changes. That's one area where C0G should shine. Though they are rated +/-30ppm/C, you can get batches that are much lower. I accidentally mixed up some X5R and some C0G 1000pF caps the other day, and sorted them by putting a soldering iron on them while measuring (I know, I know, they weren't worth the effort, but it was an interesting exercise). Some of the C0Gs changed by about .01%, 100ppm, between room temp and way too hot to touch -- that's a lot less than 30ppm/C.

Cheers, Tom