It takes a very sensitive test system to measure any film capacitor distortion, especially at what is far below the voltage rating
Spoke too soon. One probe clip was at the wrong position. 
Here is a proper measurement. Still, difference to OP measurements is large.
Here is a proper measurement. Still, difference to OP measurements is large.
No, the 50Hz peak is well below the cap distortion - this is clearly some sort of MLCC inside the fake film cap body, and the residual shows little evidence of 50Hz or 150Hz. 100Hz peak is the tricky one as its the 5th harmonic of the test signal and 2nd harmonic of the mains...
What was the voltage rating of the WIMA? Can you repeat the same test with a resistive divider? I am surprised the MKP should generate -94 dB 3rd harmonic. If this were true, I wouldn't know how we can have audio equipment with -120 db THD using MKP caps.Spoke too soon. One probe clip was at the wrong position.
Here is a proper measurement. Still, difference to OP measurements is large.
View attachment 1251290
Edit: what did you use for DAC, amp, ADC?
You wee right to be suspicious. So was I. 🙂
At first, result didn’t rise alarm as OP had similar distortion. But, measurement setup with resistor is, in principle, the same as was used by C. Bateman for his famous article about capacitor distortion.
I found that used signal generator was at the too low temperature and went to high distortion. It is EOSC10KV3 signal generator which has about -150 dBc at < 3 V output. Used ADC is Cosmos + active notch filter.
Repeated test has shown distortion at -130 to -150 dBc. I will make an update after lunch.
Improvise measurement rigs and hobby equipment are full of potential pitfalls.
At first, result didn’t rise alarm as OP had similar distortion. But, measurement setup with resistor is, in principle, the same as was used by C. Bateman for his famous article about capacitor distortion.
I found that used signal generator was at the too low temperature and went to high distortion. It is EOSC10KV3 signal generator which has about -150 dBc at < 3 V output. Used ADC is Cosmos + active notch filter.
Repeated test has shown distortion at -130 to -150 dBc. I will make an update after lunch.
Improvise measurement rigs and hobby equipment are full of potential pitfalls.
As soon as a varying voltage is across the cap, X5R X7R dielectrics have a piezo effect, changing size, even up to vibrating themselves off the pcb in extreme cases. using them as ac coupling caps, the voltage on them is constant, nothing to worry about.
the other distortion mechanism is dielectric absorption, the charge cannot be provided immediately by the cap.or when you discharge a cap, and open up the short, suddently there is some charge on it again. integrators do not like that. so only propylene or mica caps are used there.
film caps that have self healing properies should be named self-destructing caps. I have seen examples of german brand (made in spain) film caps where hardly any metal coating was left over after being used in an EV fast charger dcdc converter. The spanish factory has been closed now.
the other distortion mechanism is dielectric absorption, the charge cannot be provided immediately by the cap.or when you discharge a cap, and open up the short, suddently there is some charge on it again. integrators do not like that. so only propylene or mica caps are used there.
film caps that have self healing properies should be named self-destructing caps. I have seen examples of german brand (made in spain) film caps where hardly any metal coating was left over after being used in an EV fast charger dcdc converter. The spanish factory has been closed now.
Then I don't understand why you swept the frequency from 20 Hz. At that point the fundamental would be attenuated by 20+ dB so the THD+N will looks artificially worse. An FFT at 2.5 kHz would be more meaningful.
Have you validated your setup with a known good cap?
If you stick with 100 nF you should have no trouble getting various different film capacitors and MLCCs. The distortion of C0G/NP0 ceramic should be lower than your test setup. Polypropylene won't be far behind.
Tom
Such high-K dielectrics have pretty high voltage coefficient too. If you have signal voltage across the capacitor and the capacitor has high voltage coefficient you'll introduce a signal-dependent error which shows up as harmonic distortion.
Tom
Yes, it is almost impossible to measure distortion of a PP capacitor. OTOH, X7R looks like what OP measured. This would be another record in inventive Chinese parts faking, X7R inside film capacitor case. Every cent counts. 😉
When measuring the distortion of a capacitor, I think it is best to measure the voltage across it when it is driven by a sine wave current.
I recommend an inverting integrating circuit as shown below. It can be output to distortion meter at low impedance. Electrolytic capacitors also distort, but if the voltage is 100 times that of the DUT, the voltage across them is only a small amount and can be ignored.
Using the 5532, you can obtain an amplitude of 10Vrms with a supply voltage of ±18V, and if R1 ≥ 3.3kΩ, you can measure up to -120dBc at 10kHz or less.
I recommend an inverting integrating circuit as shown below. It can be output to distortion meter at low impedance. Electrolytic capacitors also distort, but if the voltage is 100 times that of the DUT, the voltage across them is only a small amount and can be ignored.
Using the 5532, you can obtain an amplitude of 10Vrms with a supply voltage of ±18V, and if R1 ≥ 3.3kΩ, you can measure up to -120dBc at 10kHz or less.
A risk is the opamp distortion is often higher than the capacitor distortion, for PP and C0G for instance.
I've just triaging all my ostensibly PP film caps to remove the non-PP types. The QA403 has internal OPA1612 and OPA1632 buffers and you can just use a simple RC or CR divider direct and get that -120dB measurement - anything much higher and its a fake or polyester or a manufacturing flaw (there seems to be some fluctuation amongst the better caps). I'm not so interested in precision measurement here, just is it a good PP cap...
I also managed to label some of my ceramics as C0G (they had minimal markings) in the process.
Next I need some sort of jig for testing SMT caps - it would be good to know how well PPS ones perform (PP and PS are not available surface mount as they melt in the reflow oven!)
I've just triaging all my ostensibly PP film caps to remove the non-PP types. The QA403 has internal OPA1612 and OPA1632 buffers and you can just use a simple RC or CR divider direct and get that -120dB measurement - anything much higher and its a fake or polyester or a manufacturing flaw (there seems to be some fluctuation amongst the better caps). I'm not so interested in precision measurement here, just is it a good PP cap...
I also managed to label some of my ceramics as C0G (they had minimal markings) in the process.
Next I need some sort of jig for testing SMT caps - it would be good to know how well PPS ones perform (PP and PS are not available surface mount as they melt in the reflow oven!)