Ceramic capacitors - measuring and acceptable tolerances

[jewelie]

Hiya

I'm going around in circles trying to work out if I've just got another load of duff ceramic capacitors or not.

I got an eBay mutlipack from China many moons ago but only sat down recently to try to measure them.  I don't have an LCR meter, just a Vici VC97, a Fluke 87 (original), an Arduino Nano clone and an oscilloscope. 

Using just the DMMs all the pF ones seemed to check out as pretty accurate; however, about half the nF ranges seemed way out to me (>25%.)  As an example, the 100nF ones all measured lower than all the 68nF ones, they appeared to be *that* much out. 

I took this as the cost of learning and ordered some replacements from the UK eBay auction: #121620084284 that listed 10% tolerance - obviously they were a bit more money but seemed like it would be worth it.

The replacements have arrived and I'm seeing much the same issues, especially with the 104 s / 100nF - they seem to be around 65nF to 85nF.

Doubting myself, I tried implementing some of the Arduino capacitance measuring circuits & programs and got pretty much the same result.

Wondering if this was just a very low frequency issue I tried the oscilloscope method (using 1KHz square wave at 5v peak-to-peak to charge the 100nF via a 1Kohm resistor.)  I've only an old CRO but with a bit of careful aligning, photographing, image cropping and pixel measurements, followed by some calculations...

...I still got about the same results!

I got my Vellemen Ceramic Capacitors kit out and measured some of those (a 47nF, 100nF) and got pretty much exactly the right readings on the meters.  Also apparently good accuracy with a MLCC 1uF.

I don't mind if I pay peanuts for cheap inaccurate junk from China without checking for any declaration of tolerance first and receive crap.... but it's a different matter if I pay over the odds from a UK seller stating 10% tolerance.

So, am I doing something really stupid, or are these capacitors ALSO not actually within 10% tolerance after all?   

Warm Regards
Julie
x

PS Can I get away with those leaded MLCC bead caps instead of disc ceramics for more hobby/experimental purposes?  Are they likely to be more accurate?  (The 105 / 1uF ones that I've got so far certainly appear to be.)

[T3sl4co1l]

Do a test with DC bias.  Do another one at a series of temperatures up to 100-150C or so.

You've probably got Z5U, so they age horribly even if they were made to nominal values first.

The smaller ones are probably Y5P, X7R, or C0G (white).

Tim

[ebclr]

Believe or not this work quit well

http://www.aliexpress.com/item/High-Quality-Best-Price-Newest-Diode-Triode-Capacitance-ESR-Meter-MOS-PNP-For-LCR-For-T3/32622886308.html?spm=2114.01010208.3.10.MQfMNe&ws_ab_test=searchweb201556_8,searchweb201602_4_10037_10017_507_10033_10032_401_10040,searchweb201603_3&btsid=34f5d12e-3df3-4624-856e-73bb13cca2c1

[wraper]

Heat the capacitor with a soldering iron, cool it down, then measure again. Class 2 and 3 capacitors do age by dropping capacitance, heating resets it. Also if those are Z5U / Y5V or something similar, if your meter gives high enough DC offset, it already can noticeably impact the capacitance. But if you hold them with your fingers, heat from them can also drop capacitance by say 10 nF, they are highly unstable.

Components of this classification are fixed, ceramic
dielectric capacitors suited for bypass and decoupling or other
applications in which dielectric losses, high insulation resistance
and capacitance stability are not of major importance. Z5U exhibits
a predictable change in capacitance with respect to time and
voltage and displays wide variations in capacitance with reference
to ambient temperature. Capacitance change is limited to +22%,
-56% from +10°C to +85°C.

http://www.kemet.com/Lists/ProductCatalog/Attachments/69/KEM_C1004_Z5U_SMD.pdf

[jitter]

Wondering if this was just a very low frequency issue I tried the oscilloscope method (using 1KHz square wave at 5v peak-to-peak to charge the 100nF via a 1Kohm resistor.)  I've only an old CRO but with a bit of careful aligning, photographing, image cropping and pixel measurements, followed by some calculations...

...I still got about the same results!

Well, you meaured in different ways and the results corroborate eachother.

I got my Vellemen Ceramic Capacitors kit out and measured some of those (a 47nF, 100nF) and got pretty much exactly the right readings on the meters.  Also apparently good accuracy with a MLCC 1uF.

I don't mind if I pay peanuts for cheap inaccurate junk from China without checking for any declaration of tolerance first and receive crap.... but it's a different matter if I pay over the odds from a UK seller stating 10% tolerance.

At least the UK seller can be contacted and confronted.
You might also want to try some from a well known and reputable company like Mouser.

So, am I doing something really stupid, or are these capacitors ALSO not actually within 10% tolerance after all?   

Unless you have types that alter capacity depending on the applied voltage, I'd say no.

If you haven't seen it already, check out this video:



Edit: do you know the type of your ceramic capacitors (e.g. NPO, X7R or Y5V)? In your UK eBay link, the type is not listed.

[jewelie]

Edit: do you know the type of your ceramic capacitors (e.g. NPO, X7R or Y5V)? In your UK eBay link, the type is not listed.

Thanks for that. 

I probably can't afford to kit myself out with stuff from mouser given the £12 P&P charge each time.  There's been a few things in the past that I could only get from mouser but the P&P charge made it uneconomic - there's an admittedly particularly rubbish TV waiting to go to landfill because of this (it's worth £2  to replace the broken connector, but not £14.)

eBay seller has just got back to me and told me that the manufacturer now says the tolerance is -20%/+80% and NOT +/-10% after all.  I'm not happy, if I just wanted what I had before then I would have stuck with what I had before which also cost a lot less too.  It wasn't as described so the lot (£12 worth) will be going back.

I'm having a real job stocking up on affordable adequately accurate nF range caps for the hobby box. 

[wraper]

I'm having a real job stocking up on affordable adequately accurate nF range caps for the hobby box.

There is no real +/-10% once you go above above few tens of nF maximum (those are extremely rare), usually few nF - class l ceramic (C0G/NP0). Capacitance will drift a lot with temperature, ageing and voltage applied. So if you care about do precise capacitance for particular usage, class ll and lll ceramic capacitors are not for you. So either you are stuck with a few nF max or use film capacitors instead.

[wraper]

or are these capacitors ALSO not actually within 10% tolerance after all?

Technically they may be well within 10%. If you open the datasheet for class ll or lll ceramic cap, you'll see how capacitance is measured and there will be something like cap needs to be heated up and cooled before measurement.
I checked that listing. I see that seller have edited tolerances (to be safe than sorry). However +80/-20% can be real only for higher capacitance and crappy ceramic (for disc ceramic capacitors you can be sure that will be crappiest ceramic possible once you are above few nF).

[jewelie]

or are these capacitors ALSO not actually within 10% tolerance after all?

Technically they may be well within 10%. If you open the datasheet for class ll or lll ceramic cap, you'll see how capacitance is measured and there will be something like cap needs to be heated up and cooled before measurement.
I checked that listing. I see that seller have edited tolerances (to be safe than sorry). However +80/-20% can be real only for higher capacitance and crappy ceramic (for disc ceramic capacitors you can be sure that will be crappiest ceramic possible once you are above few nF).

Heat the capacitor with a soldering iron, cool it down, then measure again. Class 2 and 3 capacitors do age by dropping capacitance, heating resets it. Also if those are Z5U / Y5V or something similar, if your meter gives high enough DC offset, it already can noticeably impact the capacitance. But if you hold them with your fingers, heat from them can also drop capacitance by say 10 nF, they are highly unstable.

Components of this classification are fixed, ceramic
dielectric capacitors suited for bypass and decoupling or other
applications in which dielectric losses, high insulation resistance
and capacitance stability are not of major importance. Z5U exhibits
a predictable change in capacitance with respect to time and
voltage and displays wide variations in capacitance with reference
to ambient temperature. Capacitance change is limited to +22%,
-56% from +10°C to +85°C.

http://www.kemet.com/Lists/ProductCatalog/Attachments/69/KEM_C1004_Z5U_SMD.pdf

Thanks guys.

Inspired by your collective input, I conducted some more, very rough, tests.

Some background info-

To give it a large heat-up-and-cool-down cycle, I used a soldering iron in direct contact with the capacitor for a few seconds, as if soldering it.
To do minor temperature sensitivity tests, i.e. ~10'C above room temperature, I found that holding tightly between my finger and thumb worked pretty well (tests suggested that went to about 33'C, room temp was about 21'C.)

The two meters I used appear to measure capacitance in slightly different ways-

The (original) Fluke 87 seemed to run at about a fixed 2Hz-ish cycle.  The curves are the standard looking charging curve followed by loooong flats, so I suspect is still constant voltage using 1.2V to charge the capacitor fully and then discharge fully, whatever the values (changing the capacitor didn't change the 2Hz cycle), so I guess it's just measuring how long it takes to get to 63.2%, twice a second. 

The Vichi VC97 uses a source voltage of about 0.7V open circuit, and appears to cycling the capacitor up-and-down to the infamous ~ 63.2% of this (I measured about 0.44V), so I'm guessing it's using a comparitor to measure this vs a 63.2% / 0.44V ref to swap between charge and discharge and then measuring frequency to determine the capacitance over time?

The measurements-

A tantalum 1uF - the closest I have to a reference. 

I didn't mess with this by giving it a heat cycle because I've only got one of these and I need it for a Vref circuit later.

@21'C
Fluke 87 = 0.97uF. 
VC97 = 0.96uF (adding 10'C made no difference.)

A 47nF from my Vellemanp K/CAP1 kit - standard unspecified tolerance/material ceramic disc capacitor.

Giving it a soldering iron heat cycle and cool down raised it about 5-10nF.

@21'C
Fluke 87 = 56nF
VC97 = 51nF (adding 10'C reduced it down to 36nF.)

A 47nF disc ceramic from the recently ordered packs discussed earlier in this thread.

Giving it a soldering iron heat cycle and cool down raised it about 5-10nF.

Fluke 87 = 49nF
VC97 = 43nF (adding 10'C reduced it down to 34nF.)

A random eBay leaded Monolithic Multilayer Ceramic Capacitor 1uF.

Giving it a soldering iron heat cycle and cool down made 0.5nF difference - barely any change.

Fluke 87 = 0.96uF
VC97 =0.96uF (adding 10'C reduced it down to 0.95uF - barely any change.)


To me it looks like the finger heat test, plus seeing different readings between the Fluke 87 and the VC97, seem quite a good indicator of the general stability of any random capacitors I have?

From that, even though the material/tolerances weren't stated, the eBay 1uF monolithic multilayer ceramic capacitor pack looks a fair compromise for random hobby experimenting purposes?  I've some more of different values coming from HK that I can compare with soon too (0.22uF and 0.47uF.)  Might these be the way to go for me, for relatively low voltages - ordering a pack at a time and testing them when they arrive?

Warm Regards from wet Derby, UK
Julie
xxx

PS OMG that £10 (~US$15) 1970s 10Mhz Oscilloscope I got (through the help and direction of folk in another thread) has come in so handy of late.  I seem to be turning to it almost every time I start playing with electronicy things now.  The pure analogue lines, looking so perfect and unaliased, are soooo preddy to look at too. 

[jewelie]

Believe or not this work quit well

http://www.aliexpress.com/item/High-Quality-Best-Price-Newest-Diode-Triode-Capacitance-ESR-Meter-MOS-PNP-For-LCR-For-T3/32622886308.html?spm=2114.01010208.3.10.MQfMNe&ws_ab_test=searchweb201556_8,searchweb201602_4_10037_10017_507_10033_10032_401_10040,searchweb201603_3&btsid=34f5d12e-3df3-4624-856e-73bb13cca2c1

From memory I believe that's a version of an open arduino-ish project that's quite a labour of love.  I've been thinking of getting one but I can't really justify it as I've been spending too much time of late, I've now got a few meters and a 'scope, and buying one pre-built seems to be a bit of a cheat.  Also, I gather there's an error on many/some of the boards where a 2V5 zener diode is being used where a 2V5 Vref should be being used (or omitted entirely)?

[ebclr]

Maybe we can use those capacitors as temperature sensor

[jewelie]

Maybe we can use those capacitors as temperature sensor

Hahaha.  That's just what I thought.  They could detect just my body heat from a distance of a few cms.   

[mikerj]

It's hard to believe just how poor the stability of a typical Class 2 MLCC cap is, in particular the voltage sensitivity is huge and gets worse with decreasing package size.

A circuit I have been involved in debugging recently had an LDO with a 1uF X5R 0201 (imperial) 6.3v cap on it's output.  Unusually this particular LDO requires a minimum low ESR capacitance on it's output for stability, and 0.4uF is recommended.  At the LDO's output voltage of 2.9v the 1uF cap can not meet this requirement, taking into account tolerance, life and temperature (~0.36uF).  A 2.2u 6.3v cap in the same package size just barely meets the requirements (~0.5uF).  The 2u2 cap doesn't give double the capacitance under the exact same conditions because it's voltage coefficient is even worse than the 1uF cap.

I strongly recommend anyone unfamiliar with the characteristics of these devices to study the datasheets, and never use a smaller package size than you actually need. Be aware the highest value parts in any particular size will likely have worse stability.  Note the disclaimer in the small print of some major manufacturers that suggest these parts only be used in devices with a design life of 5 years or less.

[wraper]

A random eBay leaded Monolithic Multilayer Ceramic Capacitor 1uF.

Giving it a soldering iron heat cycle and cool down made 0.5nF difference - barely any change.

Fluke 87 = 0.96uF
VC97 =0.96uF (adding 10'C reduced it down to 0.95uF - barely any change.)

Then it is class II capacitor (X7R or X5R). Although they are relatively stable with temperature, voltage bias significantly affects the capacitance, up to 80% of capacitance drop at rated voltage in worst cases, usually less though. The fact that heating didn't add up a lot of capacitance, means either you didn't heat it up enough (must exceed Curie point internally, around 125^o C) or capacitor was really new, just from the factory. Although they are much better than class 2 capacitors, they still must not be used for passing audio signal (if you care about audio quality) or timing, unless you can live with something like 50+ % tolerance.

[wraper]

Be aware the highest value parts in any particular size will likely have worse stability.  Note the disclaimer in the small print of some major manufacturers that suggest these parts only be used in devices with a design life of 5 years or less.

Leaded MLCC capacitors can have whatever inside. They are basically SMD capacitor with leads soldered and then potted. Often it can be 0402 size capacitor inside, which is very tiny compared to such capacitor's overall size. I find that leaded MLCC usually are worse than plain SMD of reasonable size like 0603 or 0805. For example 0.47 uF 25V MLCC: