Fake polypropylene capacitors from China?

[Krampmeier]

I ordered some polypropylene foil caps from china, which I wanted to use in a ripple filter for a PWM DAC. I chose this type of caps because the dielectric absorption of other capacitor types can increase the settling time of the DAC significantly. I have already seen that effect with the tantal caps I have in my circuit today.

I compared the chinese caps to some others I found in the lab, and got some surprising results.

Method:
I set a Keithley 2461 sourcemeter to 1 µA constant current and a voltage limit of 2 V. Then I discharged the capacitors by shorting them and then charged them with the SMU. The pictures show the current when the capacitors reach the voltage limit. I think that with an ideal capacitor, the current would drop to zero very fast. Any residual, decreasing current would indicate dielectric absorption.

I used 3 different 1 µF capacitors, and 2 470 nF parts of the chinese caps in parallel, which also resulted in about 1 µF (the capacitance was above the nominnal value).
As you can see from the attachments, the chinese capacitors seem to have less DA than the X7R and the tantal capacitors, but very much more than the Wima MKS4 (polyester) ones. I used to think that PP caps should be even better in this respect than polyester.

Am i just wrong in my assumptions or conclusions, is my method wrong, or have I been betrayed?

Here's a link to the "PP" caps I bought: https://www.amazon.de/gp/product/B07J2ZTSGH/ref=ppx_yo_dt_b_asin_title_o09__o00_s00?ie=UTF8&psc=1

[edpalmer42]

I'm not an expert on this.

Wikipedia shows a different method of measuring dielectric absorption.  https://en.wikipedia.org/wiki/Dielectric_absorption . 

This method doesn't really disagree with yours, it's just looking at discharge behaviour rather than charge behaviour.  It's also  easier to derive a number from the measurement just by measuring the 'phantom' voltage.

Ed

[TimFox]

When I check capacitors to see if they are polypropylene, I use a good LCR meter (DE5000) and measure the Q factor.  Genuine polypropylene (and polystyrene) should have Q around 1000 (a practical limit for this measurement), but polyester (Mylar) is roughly 100.

[Sylvi]

Hi
Polypropylene and polystyrene are pretty much ideal dielectrics. In audio they exhibit zero distortion where polyester adds third-order harmonics.

Polyprop also has that nice self-healing effect making it ideal for use in power supplies.

Polystyrene is next to impossible to get in anything but tiny values.

Buying from China takes some care. There are good and bad vendors, so don't paint them all with the same brush. For a critical performance component it is safest to use reliable broadline distributors like Digikey and Mouser as they work hard to assure that their sources are reliable.

[Gyro]

If this is for a one-off project then decently large value Polystyrene capacitors are available NOS from ebay sellers in the former Soviet Union. These tend to be close tolerance and good quality, for little money.

[spec]

Hi
Polypropylene and polystyrene are pretty much ideal dielectrics. In audio they exhibit zero distortion where polyester adds third-order harmonics.

Polyprop also has that nice self-healing effect making it ideal for use in power supplies.

Polystyrene is next to impossible to get in anything but tiny values.

Buying from China takes some care. There are good and bad vendors, so don't paint them all with the same brush. For a critical performance component it is safest to use reliable broadline distributors like Digikey and Mouser as they work hard to assure that their sources are reliable.

Just in case: 

https://www.digikey.co.uk/products/en/capacitors/film-capacitors/62?k=&pkeyword=&sv=0&pv1293=19&pv909=4&sf=0&FV=ffe0003e%2Cfffc000a&quantity=1&ColumnSort=0&page=1&nstock=1&pageSize=25

https://industrial.panasonic.com/content/data/CP/PDF/Film/film_ecwfa_e.pdf

http://www.kemet.com/Lists/ProductCatalog/Attachments/43/KEM_F3106_R75.pdf

https://www.wima.de/wp-content/uploads/media/e_WIMA_MKP_4.pdf

https://www.mouser.co.uk/pdfdocs/CDE715pFilmCaps.pdf

[tszaboo]

Take a scalpel, and take it apart.

[Krampmeier]

Thank you yery much for all the thoughtful answers!
I was very aware of the fact that these cheap caps are probably garbage, that's why I tested them first. I thought it would be worth trying, and it would certainly be a good opportunity to learn something...

So, let's go...

Q Factor measurements
I don't have a high quality LCR meter at home, but my 4192A impedance analyzer should give a rough idea of the Q factor, even though the analyzer it is a bit out of spec.

Measurements at 1 kHz, 1V amplitude, 1V bias:
1 µ X7R: C = 0.992 µF, Q = 73
1 µ Tantalum: C = 0.994, Q = 41
1 µ WIMA Polyester: C = 0.924 µF, Q = 242
470 n Chinese PP: C = 0.509 µF, Q = 198
2*470 n Chinese PP in parallel: C = 1.007 µF, Q=208

Measurements at 10 kHz, 1V amplitude, 1V bias:
1 µ X7R: C = 0.956 µF, Q = 77
1 µ Tantalum: C = 0.958 µF, Q = 6.4
1 µ WIMA Polyester: C = 0.915 µF, Q = 100
470 n Chinese PP: C = 0.503 µF, Q = 85
2*470 n Chinese PP in parallel: C = 0.996 µF, Q = 85

Measurements at 100 kHz, 1V amplitude, 1V bias:
1 µ X7R: C = 0.949 µF, Q = 36
1 µ Tantalum: C = 0.917, Q = 1
1 µ WIMA Polyester: C = 0.897 µF, Q = 64
470 n Chinese PP: C = 0.492 µF, Q = 46
2*470 n Chinese PP in parallel: C = 0.973 µF, Q = 43

EC/EN 60384-1 method
I charged each capacitor to 10.0 V for about 10 minutes (I am way too impatient to wait a full hour), and shorted the whole pack of them for 10 seconds directly after that. Then I disconnected the caps and let them sit without any connections. After 15 minutes, I measured the residual voltage using a Keithley 2010 DMM in 10 V range.

1 µ X7R: 0.138 V -> 1.38% (low, but still plausible)
1 µ Tantalum: 0.233 V -> 2.33% (plausible)
1 µ WIMA Polyester: 0.026 V -> 0.26% (plausible)
470 n Chinese PP: 0.076 V -> 0.76% (not plausible, should be 0.05 to 0.1% )

Conclusion

So it is very clear now, these capacitors are not what the seller advertises.

I pried one of the caps open, and in fact there is some metalized foil inside. No idea how to determine which kind of material this is though. At least there is not ceramic capacitor buried inside 

I checked the polystyrene caps on ebay, but those are in the single digit nF range maximum. I'll rather need something like 330 nF, so I'll stick to PP. The Wima MKP4 series is easily available here, so I'll get some of those and have a look.

Thank you to everybody who sent a reply again!


--
Edit: Clarification of method, correction of value at 1 kHz

[spec]

+ Krampmeier

Fascinating results- thanks for taking the time to do all those tests and write them up.

When you get some genuine PP caps it would be great to see the test results for those.

By the way, for low-voltage (normal) electronics, I find the Panasonic ECWFA range to be the best.

Just in case: there is a gotcha with film capacitors. To get the higher voltage ratings, the manufacturers will put N capacitors in series. So it is best to check this and for this reason it is dangerous to go for too higher working voltage, even if that may seem an advantage at a cursory evaluation.

From the results that you got, it would seem that the Sino-caps do not have a PP film dialectic.

[Gyro]

I checked the polystyrene caps on ebay, but those are in the single digit nF range maximum. I'll rather need something like 330 nF, so I'll stick to PP. The Wima MKP4 series is easily available here, so I'll get some of those and have a look.

Excellent analysis.

Sorry, I should have been more explicit on the search terms, they take some finding. Try a search for 'K71-7 capacitor'

e.g.

https://www.ebay.co.uk/itm/10-x-0-5uF-250V-1-Polystyrene-Audio-Capacitors-K71-7-USSR-NOS/202534528702?hash=item2f27ffa6be:g:hkEAAOSwkDtcD9BO:rk:6:pf:0

https://www.ebay.co.uk/itm/K71-7-POLYSTYRENE-USSR-Military-capacitor-0-316uF-2-250V-NEW-Lot-of-10/263444758577?hash=item3d56880031:g:7O8AAOSwGzlTtVJ-:rk:7:pf:0

Physically large but pretty good value for 10pcs, especially if you combine postage. Very good quality and stability.

[Wolfgang]

I may be mistaken, but "real" polystyrene caps are all NOS, because nobody makes them anymore. The ones I have are all "Styroflex" or the same from the former east block countries. Only moderate capacitance values, but very precise and no aging, leakage or dielectric soakage. Who makes new ones in the same quality and is trustworthy ?

[spec]

I may be mistaken, but "real" polystyrene caps are all NOS, because nobody makes them anymore. The ones I have are all "Styroflex" or the same from the former east block countries. Only moderate capacitance values, but very precise and no aging, leakage or dielectric soakage. Who makes new ones in the same quality and is trustworthy ?

Polystyrene caps are rare nowadays because, in spite of their excellent electrical properties, they are relatively large, not compatible with SM, and are physically weak. They are also expensive. Another thing about polystyrene caps is that they always have very thin self-leads, which is not good for conductivity.

We had a laugh at work when someone put a PCB with polystyrene caps through a wave-soldering machine

But they sound fantastic in audio equipment. They are also great for filters and RF work.

Polystyrene caps are still made and are available new, as well as NOS:

http://www.farnell.com/datasheets/1869697.pdf

https://docs-emea.rs-online.com/webdocs/157e/0900766b8157e132.pdf

https://www.hificollective.co.uk/components/polystyrene.html

https://uk.rs-online.com/web/c/passive-components/capacitors/polystyrene-film-capacitors/

https://cpc.farnell.com/lcr-components/fsc-160v-6800pf/capacitor-160v-6800pf/dp/CA05867?mckv=sx5FCDr5d_dc|pcrid|224694277432|kword||match||plid||slid||product|CA05867|pgrid|48077370878|ptaid|pla-369572310414|&CMP=KNC-GUK-CPC-SHOPPING&gclid=Cj0KCQiAm5viBRD4ARIsADGUT262AYdYVQJKA_5bMf6IqW64RrYPHYJZFx2dTk0foLfqWodhA8AGKWgaAi58EALw_wcB

https://www.cricklewoodelectronics.com/CPS1N0.html?gclid=Cj0KCQiAm5viBRD4ARIsADGUT26bDv85Gpj8QSzG5pFuL2AUnsafthr_8pY0w3q4UG248nHLpTZcEpsaAtWNEALw_wcB

https://www.rapidonline.com/LCR-FSC160V220PF5-220pf-160V-5-Polystyrene-Capacitor-12-0522?IncVat=1&pdg=pla-339135155671:kwd-339135155671:cmp-757438067:adg-44804851896:crv-207912323492:pid-12-0522:dev-c&gclid=Cj0KCQiAm5viBRD4ARIsADGUT25yxfIuPlO_qi0wzVYSa61sawVb0rM1D-6whf7WzM-vPNlCzzKYvS8aAuGyEALw_wcB

[Krampmeier]

Thank you for the links!

I don't really need the stability of the polystyrene caps, but as the russion NOS seem to come at about the same price as new PP caps, I'll consider them. Will check the physical size difference first though.

Does polystyrene really give an advantage over PP in the frequency range below 1 MHz?

It will possibly still take some time until I actually buy parts because I am still working on the schematic and fighting with KiCad.
I might order some different types for curiosity and will post results here once I get the parts.

[spec]

Does polystyrene really give an advantage over PP in the frequency range below 1 MHz?

In general polystyrene is best if you need the last gram of performance at low currents, say for a filter, or audio pre amplifier, but PP tends to be best if any appreciable current in involved. But there is not that much in it. They are both excellent caps.

At one time, polycarbonate were the choice, but DuPont, or  Bayer, etc stopped making the  polycarbonate film. Wima made some really nice compact polycarbonate capacitors which were a good price too. At that time, I didn't bother with any other capacitor type, unless there was another overriding factor. They were little green beauties. 

[bson]

How many do you need???  Just buy real MKP parts from a trusted distributor.  If they're for AC coupling then a 10% part is fine; the linearity of MKP will be far more important than the exact value.

E.g., https://www.mouser.com/ProductDetail/Panasonic/ECW-FD2W105KB for a 10% part.  $4.05 for 10.

[MasterT]

///
EC/EN 60384-1 method
I charged each capacitor to 10.0 V for about 10 minutes (I am way too impatient to wait a full hour), and shorted the whole pack of them for 10 seconds directly after that. Then I disconnected the caps and let them sit without any connections. After 15 minutes, I measured the residual voltage using a Keithley 2010 DMM in 10 V range.

1 µ X7R: 0.138 V -> 1.38% (low, but still plausible)
1 µ Tantalum: 0.233 V -> 2.33% (plausible)
1 µ WIMA Polyester: 0.026 V -> 0.26% (plausible)
470 n Chinese PP: 0.076 V -> 0.76% (not plausible, should be 0.05 to 0.1% )

Conclusion

So it is very clear now, these capacitors are not what the seller advertises.
///

Some measurements I've done with my new project:
https://www.eevblog.com/forum/projects/absorption-insulation-capacitor-meter-arduino-based/msg2168257/#msg2168257

Capacitors tested at rated 100V, except HV ceramic (1kV) that I put in just for verification of the test gear, and CBB (630V).
My data for WIMA is different, (bought at Mouser) - well within spec. for polypropylene  0.05%.

Yellow-china is most likely polyester (Mylar) fake.


CBB also bought from China-ali, and do surprisingly well - same as WIMA MKP4. (Rated 630V - )


Ceramic - out of scale in 5 seconds.

[Kleinstein]

One could check the dielectric from the torn open capacitor. PP film has a relatively low melting point. The likely alternative materials like polyester and polycarbonate are higher melting point. Usually PP caps are also physically larger than mylar ones. So the size will often tell.

There is a small chance that the caps could still be PP film, but just poor quality with way to much humidity inside or similar. If the production process is poor DA can be higher than the ideal film.

PS caps usually have better insulation than PP caps - one reason is that they are usually PS all the way to the leads, while PP caps are often encapsulated in some kind of epoxy and there may be some surface leakage there. Sill PP caps are usually good enough unless dealing with fAs.

[Krampmeier]

I got a load of Wima MKP2 polypropylene last week, and also added a Panasonic ECWFA and some silver mica caps for comparison.

With the sourcemeter, I did not see any DA or leakage with neither of them. The 2461 is not as sensitive as MasterT's project, which I find realls awesome. A very simple design and great results!

The Wima MKP2-100 1,0µ polypropylene cap read 0.966 µF and Q=100 at 100 kHz, clearly better then the genuine MKP4 polyester or the fake PP from Amazon. The fake ones are much smaller than the genuine ones, so everything matches up.

I compared some 330 nF caps 10 100 kHz as well:
Wima MKP2-100 330N: C=326 nF, Q=265
Wima SMDID03330TB00KS00: C=342 nF, Q=420
Panasonic ECWFA 330N 250: C=333 nF, Q=340

The largest silver mica cap I got was 10 nF (CLEC brand), and it read Q=Overflow at 100 kHz and Q=1300 at 1 MHz. Even at 5 MHz, the old 4192A measured 11.5 nF and Q=235 for that one.

All the measurements were done with open and short compensation, but without a proper fixture and with an uncalibrated instrument. So the absolute values should be taken with a grain of salt.

[floobydust]

Polystyrene are still in production, Xicon available from Mouser. Up to around 10nF and excellent for audio gear. I'd have to dig for the THD tests of various dielectrics, it was an Audio Precision rep that did the tests.

[Doctorandus_P]

With what kind of a stupid meter are you measuring those things?
Is it able to put the axis on zero and have single digit numbers on the tick marks? 

I'm jealous.

[Krampmeier]

Doctorandus_P, this is a Keithley 2461 high current SMU, and it's not mine. I am just lucky enough to have one at work which I can use for such experiments during my lunchbreak 
The off numbers also annoyed me a bit, but the smooth interaction with that screen kinda makes up for it. You should have seen the histogram functions, quite mind-boggling and very usable...

You can set the scale manually, in a 1-10-100-... fashion, but as soon as you use the touch screen to move or zoom the graph, the numbers get off just as you saw.

The Keithley 6500 is a more affordable multimeter, which shares the same UI. It should have a similar resolution, but of cause it is not a SMU.

[Doctorandus_P]

You can set the scale manually, in a 1-10-100-...

No built-in 1 - 2 - 5 sequence? Redicilous.

as soon as you use the touch screen to move or zoom the graph, the numbers get off just as
you saw.

I really hate those off numbers.
It makes me stare at the index for 20 odd seconds in an atempt to read them. For me this is so distracting that I completely forget what I was measuring.

I'm still jealous though.
I read some spec's of the keithley's and saw some demo's from youtube/the signal path, Sharriarriaaahhrg has a tower of them standing on his desk.
Fortunately this stuff is so expensive that there is no use of even dreaming about it.

I do not know how good Keithley is in details for the GUI.
If I were to design such graph functions I would shift the axis ticks together with the data to keep round numbers on the axis. This may even be available in some menu 3 levels deep, but in this crasy world such simple Ideas are probably patented and impossible to sell / use in a normal way.