Holy c(r)apacitor dielectric, Batman!

It was funny, all those comic-book detectives zooming up and counting pixels on an image file whose name is Charging_47u.JPG

He is astonishingly literal, to the point of disability. I'm a tad schizophrenic, which is both more fun and more productive.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

4) TDK (URL is pre-loaded with a part I looked up)

These web-generated spec-sheets are great (if scary). (below)

Two 2.2uF X7R 0805 parts, one 16V, the other 25V-rated.

The 25V part is 1.60mm thick, the 16V part is 1.25mm thick.

The 16V-rated part is down to 1.1uF @ 12VDC; the 25V-rated unit retains 1.8uF at the same d.c. bias.

Cheers, James Arthur

Ooops, sorry--that 25V-part is 1206, not 0805. I missed it in the data swarm.

The better-part's d.f. is 3% versus 7.5% for the inferior part. That suggests those two 'X7R' ceramics are *not* the same.

Cheers, James Arthur

Thanks for the links James. I use a through hole TDK I went looking for it's part number on DK and DK had the spec sheet for me! (They call it a characterization sheet.) here,

gives this,

George H.

That's a helpful tip, thanks George.

This whole exercise has been a rude shock (harrr!).

I need to make an accurate, flat-topped rectangular pulse slew 1.2kV in 10nS into a small capacitive load; the 2.2uF ceramic caps I inherited for the bulk storage are down 75% under bias (series-connected for higher voltage).

At that performance, getting adequate bulk storage takes two 4x2 arrays of $5 caps., hogging real estate I'm already hard-pressed on.

I'm beginning to wonder if ceramic is up to the task, and whether film capacitors might need pressing into service here.

Given the dC/dV debacle, my faith in all of these suppliers is sorely tested at this point.

Cheers, James Arthur

Huh.. I know of kV but I don't want to go there. Is there room for a through hole film cap? (That seems to be the only piece of C-V space left for film caps.)

George H.

I did a high-voltage (1200 volt max) PLZT pulser and used film caps, the Kemet AC line "XY" types. Some of them were still alive at 8KV.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

[...]

Some of the online data is pretty good. E.g. the samsung one seems excellent (apart from being flash-based, but it is worth enabling assuming you can do it on a temporary basis)

Graphs with cursor readout of dc bias, ripple current, impedance, everything, for each individual part. And you can select multiple parts and plot their curves on the same graph for comparison.

They can be very bad, down to ~10% of their rated capacitance at rated voltage. And the curves are sort of inverse exponential so most of the capacitance vanishes after the first few volts..

--

John Devereux

I had not previously known about such huge variations in capacitance due to voltage. Thanks for the heads up.

Here is an article from around 2012 that highlights this unfortunate, deceptive, and perhaps dishonest phenomenon:

The good news is that this is most pronounced in higher values and smaller physical sizes. I use a minimum size of 0805 and most of the capacitors I have used are 1 uF and below. The article also shows that higher temperature actually counteracts the negative voltage coefficient, somewhat. Recently I bought some Murata high value 0805 capacitors (22 uF, 25 V), along with other values for general purpose use as filters and energy storage for high frequency switching supplies.

The part number is GMR21BR61E226ME4L.

I used the Murata web tool

and found:

5 V 9 uF 12 V 4 uF 15 V 3 uF

A 1 uF 25 V capacitor in 0805 (part number GRM21BR71E105KA99) is as follows:

5 V 0.8 uF 12 V 0.7 uF 15 V 0.7 uF 25 V 0.5 uF

That's much more reasonable.

Paul

Fair point. We owe you an immense debt of gratitude even if you didn't use an analogue storage scope for your measurements.

That's a neat method of smoothly interpolating in SPICE.

In my case the bulk storage's absolute value isn't that important, provided it ensures that C under bias exceeds a minimum requirement needed to keep output error within tolerance.

Be warned, too, that Z5U and X7R are *classes* of dielectrics, comprising a range of materials with widely varying capacitances under bias.

Cheers, James Arthur

I'm mulling options, but I'm leaning toward the idea that I can scrimp on capacitance IFF I refill it before the next pulse arrives. I think I can ensure that. But it's not ideal.

Other options include - stacking 2220 ceramic caps (ick), or - film.

The smallest film caps (TDK B32674-series) are 12(w)x32(l)x19mm(h) for

1.5uF @ 630V.

Lowest profile so far is Vishay, MKP1848S53070JK2A (3uF/700V, 24x32x12mm).

Cheers, James Arthur

Encouraging, thanks.

One of my limitations is that I'm not generally allowed to try things first, yet they have to work the first time. So, it's good to hear that film caps have been successfully used.

Thanks.

James

I've known of the problem forever--even posted measurements here a ways back, e.g.

But what has surprised me here is that X5R and X7R have become so awful-- from extensive measurements years ago I'd thought this was a Y5V/Z5U thing, and that the X's were well-behaved.

E.g., from that old post,

1) Taiyo-Yuden JMK316BJ106KL-T, 10uF/6.3v, X5R, 1206 10.5uF/0v, 10.0uF/3.6v, 9.6uF/4.5v

Thanks for the Samsung link--I'd missed it. I've added it to our list of web-based resources for checking ceramic capacitors' capacitance loss under d.c. bias.

1) Yageo has a search-by-part-number page. Enter your search, then look at the right-most results column for a link to "DC bias, AC voltage, TCC, ESR & |Z|" for your part #. That's decent.

2) Kemet has K-Sim, a web-based simulator that generates plots for certain of their part numbers:

3) Murata

4) TDK (URL is pre-loaded with a part I looked up)

5) Samsung

Cheers, James Arthur

There needs to be another digit to specify the degree of crappiness in this respect.

--

John Devereux

The scary thing is that my list's entries above were chosen from candidates whose dielectric is expected to be *well*-behaved! The more notorious dielectrics are simply awful.

My immediate problem has been this '2.2uF'/450V 2220 X6S capacitor measuring 400nF under bias!

I could not find that part number in their tool. The closest p/n I found was GRM21BR61E226ME44 (your part number with a 'ME44' suffix instead of 'ME4L'), a '22uF'/25v 0805-sized X5R cap that's 2uF @ 20VDC. Sweeeeet!

The 1210-sized GRM32ER71E226ME15 (X7R), '22uF'/25V is still only 5uF @ 20V. Pretty awful.

Yep.

Cheers, James Arthur

I suggest "crapacity," a unitless measure calculated as

actual capacitance at rated voltage crapacity = -----------------------------------, expressed as a percentage. advertised capacitance

Cheers, James Arthur

Some of the small AC line parts are rated for a few hundred volts RMS, tested at 1500 DC, and stand 3KV DC. I've tested the bigger ones to

8KV, where I ran out of power supply.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Gotta be a typo. It's probably "ME44L". The 'L' suffix is the reel/tape type code.