This is a rather tricky problem so I turn to people with deep theoretical and practical knowledge.
Any theories about this problem:
246 caps 1500 uF 35 V, Acrotronics type ESW http://www.arcotronics.com/download/aluminium.php?id=3
One of 500 caps "pops" at 28-32 volts only after they have been in soldering machine but even when we have taken caps directly from the package. One board hand soldered had a failure at 35 V after a couple of hours. All caps are tested with 300 mA/246 as charge current, very soft charging in other words.
We think that there is a production problem of the caps because the connections inside have sharp edges and the "booms" are always located at the sharp edges of the connector. The failure takes place after 1 min to several hours.
I wonder: Do we get a side effect of connecting so many caps in parallel? LC, oscillations, high voltages etc?
The short term solution is to changing to Panasonic. I'll guess the audiophiles will get happy about this 😀
The cap board (plus 4 more boards with less caps) will take 1800 A in charging current. This has been tested for a month now with excellent results.
Any theories about this problem:
246 caps 1500 uF 35 V, Acrotronics type ESW http://www.arcotronics.com/download/aluminium.php?id=3
One of 500 caps "pops" at 28-32 volts only after they have been in soldering machine but even when we have taken caps directly from the package. One board hand soldered had a failure at 35 V after a couple of hours. All caps are tested with 300 mA/246 as charge current, very soft charging in other words.
We think that there is a production problem of the caps because the connections inside have sharp edges and the "booms" are always located at the sharp edges of the connector. The failure takes place after 1 min to several hours.
I wonder: Do we get a side effect of connecting so many caps in parallel? LC, oscillations, high voltages etc?
The short term solution is to changing to Panasonic. I'll guess the audiophiles will get happy about this 😀
The cap board (plus 4 more boards with less caps) will take 1800 A in charging current. This has been tested for a month now with excellent results.
Attachments
Hi Peranders,
Have you used that company/brand for a while OK? If so, then I would say a manufacturing defect. Contact them. They'll appreciate the feedback to fix the issue and will likely credit your account.
If it's a new supplier, let me share with you my similar experience - don't cheap out on electrolytics run near capacity (excuse the pun).... the result is invariably *KABOOM!*
Good luck!
Have you used that company/brand for a while OK? If so, then I would say a manufacturing defect. Contact them. They'll appreciate the feedback to fix the issue and will likely credit your account.
If it's a new supplier, let me share with you my similar experience - don't cheap out on electrolytics run near capacity (excuse the pun).... the result is invariably *KABOOM!*
Good luck!
What about heat dissipation problems with these narrow-spaced caps ? What if (possible) different path resistance exagerates this problem by uneven distribution of ripple current ?
Regards
Charles
Acrotronics 1000 uF/ 35 V have been used a couple of years and the numbers are 30000-40000 I should guess and they are 14 in parallel in the application but only 15 volts max but 250 A behind. There has been no failures in those products.
The problem here accurs at a couple of mA in charging current and always in steady state which is 8/246 mA (= quite small leakage).
What happens if a cap is connected to a power supply of 35 V with 1000 A and only a couple nH in inductance?
Can anyone see a possible reason why this is unsuitable?
At the moment (before we have tested Panasonic) I think this is to 90% a quality problem and to 10% a not a good idea to connect some many caps in this way.
Charles, the problem is at 25 deg C and at steady state, only leakage current.
We have a contact with the manufacturer but they have not yet got any samples from us (today maybe) but the answers so far tells us nothing really. One solution is to use 50 V caps but why does a 35 V pop at 28 V?
The problem here accurs at a couple of mA in charging current and always in steady state which is 8/246 mA (= quite small leakage).
What happens if a cap is connected to a power supply of 35 V with 1000 A and only a couple nH in inductance?
Can anyone see a possible reason why this is unsuitable?
At the moment (before we have tested Panasonic) I think this is to 90% a quality problem and to 10% a not a good idea to connect some many caps in this way.
Charles, the problem is at 25 deg C and at steady state, only leakage current.
We have a contact with the manufacturer but they have not yet got any samples from us (today maybe) but the answers so far tells us nothing really. One solution is to use 50 V caps but why does a 35 V pop at 28 V?
"1 in 500 caps" That's 0.2%, which might be expected for a subtle cap defect, or might even be normal for a given type. Also check for seal problems and exposure to anything that would cause the cap to deteriorate if the seal were compromised. I've heard about LC issues on digital boards with many bypass caps, and think we even had that problem where I once worked, but you'll have to do some searching on that one.
Hi P-A,
I've had some experiences (nasty!) with large numbers of parallel-connected capacitors, both electrolytic and non electrolytic.
The problems are exacerbated by this parallel connection: when defects are cleared, in the case of self-healing for plastics and oxide cracks filling-up for electrolytics, there are normally small variations of voltage across the capacitor; when the capacitor is connected to a very "stiff" source of tension, this minute variation causes huge currents which stress further the (moderately) faulty capacitor, and precipitate a catastrophic failure.
This means that capacitors used in this way have to be screened with maniacal care to avoid problems.
A way to do that is to measure the noise on the charging current: even if the average leakage is very low, in the µA range, this current may be partly made up of relatively short and large spikes, which means the capacitor is unsuitable for // connection.
Manufacturers use a similar process to assess the quality of film capacitors. I remember C Bateman has written a paper around the subject.
LV
I've had some experiences (nasty!) with large numbers of parallel-connected capacitors, both electrolytic and non electrolytic.
The problems are exacerbated by this parallel connection: when defects are cleared, in the case of self-healing for plastics and oxide cracks filling-up for electrolytics, there are normally small variations of voltage across the capacitor; when the capacitor is connected to a very "stiff" source of tension, this minute variation causes huge currents which stress further the (moderately) faulty capacitor, and precipitate a catastrophic failure.
This means that capacitors used in this way have to be screened with maniacal care to avoid problems.
A way to do that is to measure the noise on the charging current: even if the average leakage is very low, in the µA range, this current may be partly made up of relatively short and large spikes, which means the capacitor is unsuitable for // connection.
Manufacturers use a similar process to assess the quality of film capacitors. I remember C Bateman has written a paper around the subject.
LV
1 of 500 is not acceptable, not at rated voltage. We have 3000 caps in a cabinet so the probability to have 100% working is zero. I'll guess I'll have to wait until tomorrow or so to here what Acrotronics have to say.
Elvee, your conclusions seem not to be unlikely. Do you have possibly any more info. How did you solve the problem. My theory is that some caps are a bit weaker but not too weak in a normal application but my application is critical for this poor guy.
Can't say but a hint, where do you need 13000 A at low voltage? Input power is a switched PS of 21 kW, volume about 10 liters.
Al electros can degrade on the shelf - the oxide layer slowly dissolves without polarizing V
reforming instructions ramp V over Hrs to rated V with current limited supply, low impedance of many parallel caps may be destroying the weak dielectric cap as mentioned above - I guess you'd just have to reform assy very slowly with a prog V source - leakage monitoring probably not helpful when normal variance of 100s of "good" caps will hide the "bad" cap's excess leakage
reforming instructions ramp V over Hrs to rated V with current limited supply, low impedance of many parallel caps may be destroying the weak dielectric cap as mentioned above - I guess you'd just have to reform assy very slowly with a prog V source - leakage monitoring probably not helpful when normal variance of 100s of "good" caps will hide the "bad" cap's excess leakage
jcx, I tested 100 caps and I set the level at 100 uA leakage and max voltage at 48 V. Most caps have a leakage of 50-100 uA at 48 V but one cap had 40 V, one 41 V and a couple at 44, 45 V. I'll guess the 40 and 41 V caps could be possible victims. In other terms, those 35 V will take about 45 V without a problem but not all. I'll guess a possible solution is to test the caps first at low voltage and then at real working voltage which is 30 volts. Putting 35 V is asking for trouble. One other solution is to use 50 or 63 V caps but they are wider = less capacitance at a fixed area.
I'd try ramping completed cap assy(s) to working V slowly, perhaps overnight, as a step in the manufacturing flow in hope of reforming any weak cap's dielectric layer without the catastrophic failure before puting them into normal operation
also, I've seen one ref claim Electros live longest at 70-80% of rated V, so maybe higher V rating might help infant mortality at cost of shorter wearout life
also, I've seen one ref claim Electros live longest at 70-80% of rated V, so maybe higher V rating might help infant mortality at cost of shorter wearout life
Have you considered the possibility of testing the caps before they are soldered to the board? With all those caps in parallel it's hard to find the problem part(s). Individual testing is sort of extreme, but you're apparently making a sort of extreme device and it is calling for extreme measures.
I'm guessing the cap bank is for a class A amp driving a bass ribbon or a fan-type subwoofer.
I_F
I'm guessing the cap bank is for a class A amp driving a bass ribbon or a fan-type subwoofer.
I_F
5000000 caps, don't think individual testing is something we want.I_Forgot said:
I_Forgot said:
Yes a very nice classs A amp 
with 21 kW power supply with 48 cables with 95 square millimeters cross section. Which AWG will this be 
djQUAN said:
imix500 said:
ΔV/ΔT of electro's would be too crummy for either 😉
Well, you MIGHT get 15 or 20 spot welds before they overheated and melted
I solved the problem by tightening the screening to weed out anything remotely suspect. This required individual testing.peranders said:
But of course, I only had to deal with hundreds of parts, not millions.
With such large numbers, you could ask the manufacturer for a special qualification procedure, a sort of pre-shipment burn-in.
It wouldn't be free, but it would be the easiest solution.
LV
How did you build the board, similar mine?
Have many caps did you have in parallel? Type? Working voltage?
Have many caps did you have in parallel? Type? Working voltage?
- Status
- Not open for further replies.