X5R vs X7R MLCC

Derate 3:1 on voltage, or they may detonate. With mandatory 3:1 derating, they are worse than a ceramic!

I posted a trick for using ceramics on the output of LM317s.

They are erratic. Some types work fine, and another batch of the same parts explode.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

John Larkin wrote

I bought an assortment of values of caps like in the bottom link some years ago on ebay, only a few left now, not one problem, much of that stuff is on 24/7.

Sure 35 V with 12V input or 5 V on it, always have some margin.

But also in industry, all I remember is 1 case where a tantalum exploded, took some cards in 19 inch rack with it. Was nothing much left of it, so we do not know if it was in the wrong way.

Much better than those electrolytics, and much better ESR in the long run I think. When using electrolytics I often put 100 nF or something in parallel, in case of tantalums not.

MnO2 tantalums are ignited by high current, namely dV/dT, so are poor choices to put on supply rails, unless severely derated on voltage.

Polymer aluminums and polymer tantalums are OK. Polymer aluminums are great, but ESR is too low for some regulators.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

C0Gs have low voltage coefficients but don't come in big values. Just lately people are making big ones, like 0.5 uF, but I don't know about those. Kemet claims "No capacitance change with respect to applied rated DC voltage."

Just hope your C-meter has no voltage coefficient! Maybe add Phil's second resistor.

Why wait til Friday?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

They're sensitive to thermal history, for one thing. I've had good luck with them on the output side of 78xx-type regulators, but I sure don't use them on the input side.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

A polymer aluminum and a pulse-rated sub-ohm resistor make a good combination for that. I usually start with a zero-ohm jumper.

Putting a lead capacitor on the feedback network of a 317-style regulator helps with that too, at the price of somewhat degraded noise.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

While waiting for the new ones I decided to use this method to test some of my current stock of Samsung parts. These are ceramic 470nF 50V X7R in an 0805 package.

Samsung P/N: CL21B474KBFNFNE Source: Digi-Key

I tested these from 0-30V (I normally expose these to 15-24V in use). Here are my results:

Impressed they start a bit higher than rated, but the capacitance drops off faster than expected.

Here is what their datasheet shows:

Hmm, something isn't right. I should be on that upper red curve (X7R 50V). They show a 10% decline in capacitance by 30V, while I am measuring over 50%!

Very good chance I'm doing something stupid. This is my test setup:

I suppose it is possible the resistor on the right is affecting the readings from my LCR. Maybe I'm missing something else silly...

ts value

..

Interesting article. I don't have good statistics because I've never gotte n manufacturing to focus on documenting things rather than just getting uni ts fixed and shipped. What I've seen is that we have about two failures in 100 units on a 150 uF tantalum cap. The failure is not hard to find once you are aware to look for it.

I'm wondering how slow "slow" is for the voltage ramp. "I had the idea to build a capacitor-postprocessing fixture. Its function was to slowly ramp u p the voltage applied to the PCB with enough current capacity to power ever ything on the PCB but with sufficient internal resistance to limit transi ent capacitor-clearing fault current."

I have jumpers around sub ohm current measurement resistors for the UUT. I f The current sense resistor could be replaced with a current limiting resi stor. We already have a step in the test procedure to turn on power while watching LEDs that indicate the power supplies are up. We could add a step of removing the zero ohm jumper, turn on power while watching the LEDs, th en install the jumper and continue testing.

With a current limit resistor would a normal power supply ramp up be ok to clear the short without damage? I believe the PSU devices can provide 1 Am p. I guess it doesn't really matter so much. If a large enough resistor t o prevent damage is used and the short doesn't clear the voltage would be d ragged down enough to flag the problem still. If it does clear we are ahea d of the game.

It doesn't take long to replace the failed caps. So the question is whethe r this would use more time than just finding and fixing the shorts when the y blow.

Rick C.

Tesla referral code +--+

Tants are fine steady-state. It the turn-on transition that ignites them.

Usually my boards are well seeded with 1 uF bypass caps on power planes.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

You got snookered by the datasheet--those curves are almost always for

100 nF. You need the _characteristics_ link from the Digikey listing (if it's back up--they were all 404 when I looked a couple of days ago).

That's right. With time constants that large you'll need to wait several seconds between measurements, of course.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

That link still 404s for me. I would be very interested to get my hands on the 'characteristics' sheet.

I indeed had to wait about 30 seconds between readings for things to settle.

It will be interesting to compare with the new ones.

Tantalums are erratic. Some later batch may behave/fail differently.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I reported it to Digikey. If enough people do that, they may take notice.

Yup. To make it more interesting, makers use different numbers to refer to the same part, at least judging by the C(V) curves, which are exactly the same shape but cut off at different voltages, so that a 16V part loses 30% of its capacitance at 16V where the curve ends, and the 35V part has the same curve but continues down to -70% at 35V.

Clear as mud, probably intentionally.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

A cap from ADJ to ground seems to make an LM317 regulator ceramic cap tolerant. 10 nF is usually about right. I'd expect that to reduce HF noise a little.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Phil Hobbs wrote

Nice story, just learned something.

Interesting. Lead caps help buck switchers deal with large output caps, but work the other way on 317s?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Yeah, from back when you could learn something from EDN. :(

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

20 or 30 nF and no R6 works about as well.

Lead caps work with switchers that have a fb pin that's ground referenced. I tried a cap across the upper divider resistor of a 317, and that makes ringing worse. The cap to ground was an empirical (fiddled) discovery. It actually works.

I need to verify this for LM1117; we use a lot of those.

I sometimes use a 317 with FB grounded, to make 1.25V for an ARM core. That doesn't work with my trick.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

OK I first tried JL's method. And then Phil's. Phils's sorta worked but the cheap meter would sometimes lose it's mind. JL's worked fine... but I made a factor of two mistake in the math.. (I forgot the phase effects.)

I saw no change (V = 0-60V) in the 0.01uF cog, nor in a 0.1 uF cog. A 0.01 uF X7R only changed by ~10%. 0.1uF X7R decreased to about 35% of it's 0 V value at 60V. A 1 uF x7r was down to 20% at 60V. These were all 50V caps, though different manufacturers.

Thanks all for the simple test setups.

George H.