100s of ceramics in parallel for big capacitance

Putting them in parallel won't get you a high voltage. Are you using high voltage ceramics?

--

Rick

If any one of them short you have a failure, so I would not get too close to the rated voltage if reliability is important. MLCCs do fail short by times.

--sp

On Tue, 16 Sep 2014 12:44:19 -0700 (PDT), Fibo Gave us:

Placing two capacitors in parallel does NOT increase its voltage handling capacity.

Be careful not to heat the junctions any more than the initial solder operation, or their performance will degrade immensely.

No, but isn't there a better way to do it?

Do you need low inductance/fast discharge?

How much voltage and capacitance?

--

John Larkin         Highland Technology, Inc 

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

One line questions usually deserve no more than a one line reply, but I'm bored waiting for updates, so you get some obvious questions and a little more than a one line reply.

What are you really trying to accomplish (besides burn money)? Is this a product, hobby project, or repair job? How much capacitance, voltage, and volume do you need?

As others have noted, you don't get higher voltage by paralleling capacitors. There's nothing wrong with paralleling or serialing(sic) hundreds of caps, presumably MLCC type, except that your yield will suck, it won't work for very long, and you will have some really entertaining problems to troubleshoot (depending on your circuit design). One at a time:

MLCC caps require careful handling. A machine can do it. I've tried replacing them by hand and had difficulties. If you overheat them, bend them, stress them in any direction, or look at them cross eyed, they will crack and short. With 100 caps all soldered together, identifying the shorted cap is difficult and replacing it impossible.

MLCC caps also don't like to be stressed. The dielectric is the same barium titanate used to make ultrasonic transducers. Give the MLCC cap some ultrasonic waveform, and you have a loudspeaker. The generated ultrasonics will mechanically couple through the solder connection, to the PCB, and into anything fragile. IC wire bonds, suspended component leads, other MLCC caps, mechanical components, etc can be vibrated to death. I've seen it happen on computah motherboard where the ripple current through the caps from the CPU is sufficient to tear apart the caps.

Mounting such caps is always a problem. See: Notice the parallel ceramic (porcelain) cap assemblies. On Pg 3 of: notice that various mounting schemes for such parallel cap assemblies. Notice that all the cap ends terminate in an "L" shaped assembly which can be allowed to flex slightly. That's the key to having them survive. If they don't flex, they'll break (something).

The troubleshooting aspect has helped pay for some of my decadent and lavish lifestyle. A client presented me with a design review problem for his project. The board was experiencing a high failure rate on a wide variety of components. When installed in the product, components on other boards soon rapidly began to fail. It took a while, but I eventually narrowed it down to his 75KHz(?) switching power supply, which was full of MLCC caps. They were mounted vertically from the PCB, and held in place with a stiff "C" bracket. When I put my coffee cup down near the board, and saw standing waves on the coffee surface, I guessed that ultrasonic mechanical vibrations from this assembly was causing the failures. Probing around with a "Bat Detector" confirmed my guess. The easiest way to prove it was to replace the MLCC caps with larger conventional electrolytics and retest. That worked and the board was soon fixed with a fast redesign. Now, do you want to deal with troubleshooting headaches like this?

Besides, why bother making a series capacitor assembly when you can buy 10KV MLCC caps?

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

all by themselves will cause you a problem during transients.

as a mental experiment, put two caps in series. Now ramp up the voltage across them to twice the voltage rating of one cap. Now you know those two caps are NOT identical in value, nor identical in exr, nor series nH, so during the time of the ramp up you can ALL the voltage across one cap, blow it short? then take out the other cap. Now envision the leakage of one cap is not the same as the leakage of the other cap, so once you're at high voltage, you don't have equal voltage on both caps, one will HOG the voltage, the low leakage one. Again, blow the cap

To avoid this I had to put small resistance in series destroying the effective esr, AND put in distributed parallel resistance to maintain the balanced voltage, destroying the high leakage value.

Well yes, you do need balancing resistors. See Pg 13.

I've never added series resistors in a series connected capacitor string. To equalize the voltage across a string of series connected capacitors (or diodes), the resistors go across the caps, not in series.

However, if I happen to be worried about equalizing the current through parallel connected capacitors, it might be expedient to add some series resistors. I've never seen that done, except maybe to reduce inrush current, where a PTC thermistor inrush current limiter is sometimes used. That also conveniently retains the low ESR once the PTC thermistor warms up.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

Something like that is generally done for banks of super-capacitors.

But as I understood the OP, he wanted to to get a high capacity, high voltage capacitor by using multiple low capacity, high voltage capacitors in parallel. This avoids any issues with over-voltage on the caps - they are already designed to withstand the voltages that will occur.

I can't see any serious problems with his idea. There are a few things to watch out for, however. The voltage rating of the caps should be significantly larger than the maximum voltage used in order to get the full capacity of the caps. You might find that your ESR/ESL is /too/ low - current rushes could become huge. And you may have to consider current flow patterns, oscillations, etc. But I believe the principle is right.

True, was not clear whether any caps would be placed in series.

But even with a myriad in parallel, watch out for the potential resonances that, if energized often, could shorten the life of these caps. [as you stated]

100's of little surface mount ceramic capacitors in parallel in order to ac hieve a high voltage high capacitance rating?

Thanks for all the responses, I'm basically making a something to replace a n electrolytic cap for an application that doesn't allow electrolytics.

I'm thinking about using a bunch of high voltage 2.2uF ceramic caps in para llel (no caps in series) but was wondering if I was missing something, and gotchas that I'm not thinking of. Yes the resonance and what I'm doing to m y ESR as far as high current surges go, I'm going to dig deeper into those issues.

Jeff,

I really appreciate the long response, and thanks for the links and pointin g out the potential mechanical challenges.

much thanks!

Could you use polymer aluminum caps? They work at low temps and don't dry out like wet caps. How about tantalums? Neither are high voltage, so a series stack would be needed.

There are high-value film caps around, too. And classic oil-filled paper.

Make sure the caps have enough capacitance when full voltage is applied. Some dielectrics, like Z5U, lose 80% of their C at rated voltage. All ceramics except C0G do this to some extent.

--

John Larkin         Highland Technology, Inc 

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

I'm curious. What devices do not allow electrolytics? Are you referring to tantalum electrolytics, have thermal runaway problems? Or aluminum electrolytics that have lifetime problems at high temperatures. There are plenty of other capacitor types that might be more appropriate.

Thank you for finally disclosing exactly one number. Of course, it's useless without knowing:

1. The total capacitance desired.
2. The total voltage desired.
3. The operating temperature.
4. Any ripple current? What frequency?
5. Expected lifetime?
6. Any compliance requirements?
7. Cost or mechanical (size) limitations? (optional)

It would also be nice to know:

1. Is this for a production product, repair job, or hobby project?
2. Any requirements to repair the assembly?

Y'er welcome. However, I would appreciate it if you would disclose more numbers and mot ignore my questions. You'll find that the quality of the answers and suggestions are directly related to how much information you supply.

Reducing MLCCs' piezoelectric effects and audible noise

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

Great question! You know this already, but easy to use LTspice [free circuit simulation tool] to analyze the configuration you actually physically must use. Remember the rule of thumb, inductance is about

10-20nH per inch, luckily parallel inductors divide like resistors. Also, *if* you can use 'planar' connections, where the connection is wider than long, you have a hope of being successful at sticking 100's in there. Just be sure to extend the planes about an extra inch around the caps, else the edge caps will have significantly different impedances than the internal caps. Big mounting pads and multiple vias, which DON'T punch one giant hole in the next plane, but have individual cutouts for individual vias. This is the 10MHz, 20MHz, 100MHz, and up to 1-2GHz effects. You can simulate the structure's effect, too.

If you build, come back let us know what happened!

Maybe someone could make a product by putting gobs of 0402 or whatever ceramic caps onto a flex circuit and rolling it up into an electrolytic capacitor can.

To make a 1000uF/25V equivalent would take 212 0603 X5R capacitors at

0.07 each = $15 in capacitors alone (to replace a few cent part). Is there someone who will pay $50-75 each for them?

--sp

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward" 
speff@interlog.com             Info for manufacturers: http://www.trexon.com 
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

With good connection-length matching, you can make a nifty slow propogation delay line that way. It could be tuned for a variety of frequency-selective uses, if that's what you're after. Do you ask because you FOUND a board with hundreds of capacitors?

Den onsdag den 17. september 2014 23.46.57 UTC+2 skrev Spehro Pefhany:

you could go for 47u in 1206, digikey have some for 0.35

-Lasse

On Wed, 17 Sep 2014 17:46:57 -0400 in sci.electronics.design, Spehro Pefhany wrote,

Audiophile market?

putting 100's of little surface mount ceramic capacitors in parallel in order to achieve a high voltage high capacitance rating?

replace an electrolytic cap for an application that doesn't allow electrolytics.

parallel (no caps in series) but was wondering if I was missing something, and gotchas that I'm not thinking of. Yes the resonance and what I'm doing to my ESR as far as high current surges go, I'm going to dig deeper into those issues.

Even C0G / NP0 ceramic caps lose C with voltage. C0G is just a low temperature coefficient. I used to test them for just that property combination.

?-)