Hello,
I've recently buy some 5% MKP capacitors (very expensive : 0.46$)
The serie is pretty bad and some of them are out of tolerance.
Is it a common problem ?
I've recently buy some 5% MKP capacitors (very expensive : 0.46$)
The serie is pretty bad and some of them are out of tolerance.
Is it a common problem ?
What do you mean by 'pretty bad', what is the tolerance then?
How did you measure the quality?
Jan
How did you measure the quality?
Jan
Last edited:
What would I do... I would first try and use a totally different measurement technique to verify that a problem exists. It may well be a problem but you need to be sure. Could the caps be fakes ? (perhaps from one of the popular auction sites), or are they from a recognised supplier ?
I'm guessing you have used a meter of some sort, maybe a cap range on a DVM etc. Try something different, try another meter if possible, or try and rig up a circuit that will highlight the tolerance error.
If these are small value caps then a small precision oscillator is easy to knock up, one that uses the cap as a timing element. Use a frequency meter and see if the oscillator varies by as much as the initial tolerance error would suggest.
I'm guessing you have used a meter of some sort, maybe a cap range on a DVM etc. Try something different, try another meter if possible, or try and rig up a circuit that will highlight the tolerance error.
If these are small value caps then a small precision oscillator is easy to knock up, one that uses the cap as a timing element. Use a frequency meter and see if the oscillator varies by as much as the initial tolerance error would suggest.
I don't have a well calibrated professionnal high precision capacimeter, i've tested many 1% to 10% MKT and MKP series and this is the worst ever.
The first serie seems not fantastic, i was expecting that expensive 5% film capacitors should meet the manufactuers specifications... even cheap 10% are better.
Should i contact the manufacturer ?
The first serie seems not fantastic, i was expecting that expensive 5% film capacitors should meet the manufactuers specifications... even cheap 10% are better.
Should i contact the manufacturer ?
According to your readings, the MKP 5% are all within 5% with the exception of the last two highest and I would suspect the meter. The R82s are all OK and the 10% also.
What is the actual problem?
What is the actual problem?
Statistics, the median value is pretty out of target.
Synonym of a bad managed quality production
The first set of Epcos cap measurements, spec'ed at +-5%, range from 15.13µF to 15.8µF. Not sure what the value is supposed to be, but just looking at the values that you provided, I calculate the RMS value (the geometric mean) to be 15.39µF.
The cap farthest from that mean is the 15.8µF cap, which is 2.65% away from the RMS value of your measurements. So, the cluster of cap values about their geometric mean ranges around half of the faceplate tolerance, which should be acceptable.
If we assume that your measurements are accurate to 0.01µF, and that the faceplate value is supposed to be 15.00µF, the 15.8µF cap is the outlier, and its error is 5.33%, ignoring any of the statistics of your sample.
Is this really something awful that the manufacturer has to apologize for?
I didn't do statistics on the rest of the caps, but at a casual glance, your cap measurements, which have not been substantiated or accredited in any way at all, seem to vindicate the manufacturer, especially if we assume a measurement error on your part and the incorrect temperature during your measurements.
Basically, the spread of these caps is about half of what the faceplate tolerance suggests, so regardless of your measurements, the deviation of this lot of caps from its mean is far better than +-5%.
Are they defective? You have the right to hate whatever you buy and complain to the manufacturer as much as you'd like - not sure what that would get you, since your measurements are sloppy, but heck - have fun!
For my money, these caps seem to be produced pretty repeatably. If accuracy is so important and you're buying a +-5% part, but you want a tighter tolerance, you need to measure and match the parts you want to use, and you have to get used to rejecting parts that you can't figure out how to use.
It's either that or learn to roll your own caps... but honestly, I would not call the caps you measured "fraudulent" at all. Most likely, your measurements are fraudulent: prove this to us otherwise.
The cap farthest from that mean is the 15.8µF cap, which is 2.65% away from the RMS value of your measurements. So, the cluster of cap values about their geometric mean ranges around half of the faceplate tolerance, which should be acceptable.
If we assume that your measurements are accurate to 0.01µF, and that the faceplate value is supposed to be 15.00µF, the 15.8µF cap is the outlier, and its error is 5.33%, ignoring any of the statistics of your sample.
Is this really something awful that the manufacturer has to apologize for?
I didn't do statistics on the rest of the caps, but at a casual glance, your cap measurements, which have not been substantiated or accredited in any way at all, seem to vindicate the manufacturer, especially if we assume a measurement error on your part and the incorrect temperature during your measurements.
Basically, the spread of these caps is about half of what the faceplate tolerance suggests, so regardless of your measurements, the deviation of this lot of caps from its mean is far better than +-5%.
Are they defective? You have the right to hate whatever you buy and complain to the manufacturer as much as you'd like - not sure what that would get you, since your measurements are sloppy, but heck - have fun!
For my money, these caps seem to be produced pretty repeatably. If accuracy is so important and you're buying a +-5% part, but you want a tighter tolerance, you need to measure and match the parts you want to use, and you have to get used to rejecting parts that you can't figure out how to use.
It's either that or learn to roll your own caps... but honestly, I would not call the caps you measured "fraudulent" at all. Most likely, your measurements are fraudulent: prove this to us otherwise.
Statistics, the median value is pretty out of target.
Synonym of a bad managed quality production
Your median calculation is wrong: for the first series of EPCOS caps, only 3 caps have a larger value than your median, and 17 have a value lower than your median. By definition, your median value is wrong, assuming that your measurements are right in the first place.
I was thinking about inform the manufacturer about a defective batch.
There is a quality issue when the median value of the batch is not centered on the target value... perhaps my batch is too small 😀
Done, i'm waiting for their answer.
I haven't presiced "statitics of what 😀"
(15,13+15,13+15,13+15,16+15,27+15,28+15,29+15,29+15,35+15,37+15,41+15,42+15,44+15,45+15,47+15,47+15,47+15,73+15,77+15,8)/20
= 15.59
Taget value = 15
We are at +4% of the target... critical for a median value.
Kemet +1.6 ok 🙂
Wima -0.2 ok
Apart from that one or two possible outliers (And your measurement accuracy is in question), those caps look fine to me.
There is absolutely no requirement that the distribution be symmetric about the target value, you see all sorts of distributions. I have ordered (back in the day) a batch of 5% wire wound resistors that all turned out to be within 1% of a value that was 3% low... Perfectly valid, and it saved the manufacturer 3% on their expensive resistance wire.
Back then you also used to see two peaks in the distribution when buying 5% parts, because they had selected out all the 1% parts from the batch to sell at a premium, apart from some specialist things you don't see that so much these days.
If you design for a 5% tolerance then anything within 5% should work, so why worry about it?
Regards, Dan.
There is absolutely no requirement that the distribution be symmetric about the target value, you see all sorts of distributions. I have ordered (back in the day) a batch of 5% wire wound resistors that all turned out to be within 1% of a value that was 3% low... Perfectly valid, and it saved the manufacturer 3% on their expensive resistance wire.
Back then you also used to see two peaks in the distribution when buying 5% parts, because they had selected out all the 1% parts from the batch to sell at a premium, apart from some specialist things you don't see that so much these days.
If you design for a 5% tolerance then anything within 5% should work, so why worry about it?
Regards, Dan.
Ok, there is no problem, thanks for your answers.
I've buyed luxury 😀 MKP instead of standard MKT for active filters which apparently needs at least 2% tolerance capacitors...
perhaps i'm becoming a capacitor aristocrat.
I've buyed luxury 😀 MKP instead of standard MKT for active filters which apparently needs at least 2% tolerance capacitors...
perhaps i'm becoming a capacitor aristocrat.No,
The median value is not part of the specification.
If the specification states tolerance ±5% then that is what all the product must meet.
If you measure a component with uncalibrated equipment, then you have no argument and no valid data. You need to measure sufficiently accurately to have data to present to the manufacturer.
Is that so? 5% tolerance means they should be within 5% of the stated value, not that they should be evenly spread (or normally spread) about the stated value. A group where all are 4-5% low (or high) would meet the spec.silverprout said:
Yes, i've said it before, i'm looking for an advice.
I'm never satifyed of the worst case scenario.
multimeter.
I've proposed to the manufacturer to send him for free the supposed out of tolerance capacitors...
Whether you are satisfied is irrelevant to a manufacturer who has clearly stated a spec and met it. If you want a normal distribution around the nominal value then you may have to pay extra for that or do your own selection.silverprout said:
- Status
- Not open for further replies.