Exact match. Not showing close matches.
PICList
Thread
'[EE] Capacitor went Poof! ... why?'
2004\11\24@113626
by
PicDude
|
Hi all,
Late last night, while building a 3-output power supply, a capacitor burned
out (it just made a poof sound and the electrolyte leaked out) and it's
totally baffling me.
I built the 12V stage, tested it under a load, and it worked fine. Then I
built the 5V stage, and that worked fine under load. When I built the final,
3.3V stage and was testing it, the input capacitor for the 5V stage died.
This time though, the 5V stage was NOT under load. I immediately thought
that the output capacitor must've died since it was not under load and
perhaps the output voltage went too high, but it was actually the input
capacitor (Nichicon 330uf, 35V, low-ESR electrolytic).
In more detail, the input voltage (which in this case was 14V from a switching
power supply) is split in 3 directions, each going to a separate SB540
schottky diode (for reverse-voltage protection), then to the switching
circuitry for each stage. The 5V switcher is an OnSemi MC33167, with a 330uf
35V input capacitor. Because there are reverse-protection diodes on there, I
can't see how anything happening on either of the other stages could cause a
reverse voltage to the 330uf cap. And since my input is 14V, that cap was
receiving a supply of ~13.5V (after the reverse-protection diode).
I've not re-connected it for testing as I'd like to know what I should
check/test for. So far, I've verified that that cap and other components
were soldered in with the correct polarity, etc. Also, can/should I test
with the existing (blown) cap in place, or should I replace it first? I
assume the latter. I know that switchers need a minimum load, and I did make
provisions for this on the PCB, but do not have a min load resistor on there
yet. How would I determine the min load required? The datasheet does not
indicate how to determine this. Let me know if you need other info/circuit
diagrams, etc.
Much thanks,
-Neil.
____________________________________________
2004\11\24@132859
by
olin_piclist
PicDude wrote:
> Late last night, while building a 3-output power supply, a capacitor
> burned out (it just made a poof sound and the electrolyte leaked out)
> and it's totally baffling me.
Most likely possibility: The capacitor was in backwards. They do take it
for a while, then BOOM. I know you said you checked this. Check it again.
I still think it's the most likely explanation.
Second possibility: The ripple current was way too high for the cap, which
heated it up and eventually caused it to explode.
> Also,
> can/should I test with the existing (blown) cap in place, or should I
> replace it first?
Definitely replace it. It could look like an open, a short, or a toxic fume
generator.
> I know that switchers need a minimum load,
Not true. Some commercial switching power supplies do indeed need a minimum
load, but this is not inherent to switchers, only particular
implementations.
> How would I determine the min load required? The datasheet does
> not indicate how to determine this.
If it has a minimum load, this should be in the data sheet. If the data
sheet doesn't mention it, then I would assume it has no minimum load
requirement.
*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com
____________________________________________
2004\11\24@141503
by
Herbert Graf
|
On Wed, 2004-11-24 at 10:36 -0600, PicDude wrote:
> I built the 12V stage, tested it under a load, and it worked fine. Then I
> built the 5V stage, and that worked fine under load. When I built the final,
> 3.3V stage and was testing it, the input capacitor for the 5V stage died.
> This time though, the 5V stage was NOT under load. I immediately thought
> that the output capacitor must've died since it was not under load and
> perhaps the output voltage went too high, but it was actually the input
> capacitor (Nichicon 330uf, 35V, low-ESR electrolytic).
>
> In more detail, the input voltage (which in this case was 14V from a switching
> power supply) is split in 3 directions, each going to a separate SB540
> schottky diode (for reverse-voltage protection), then to the switching
> circuitry for each stage. The 5V switcher is an OnSemi MC33167, with a 330uf
> 35V input capacitor. Because there are reverse-protection diodes on there, I
> can't see how anything happening on either of the other stages could cause a
> reverse voltage to the 330uf cap. And since my input is 14V, that cap was
> receiving a supply of ~13.5V (after the reverse-protection diode).
Well it looks like you've covered the bases, so it's simply possible you
got a bad cap. I've seen that from time to time.
FWIW electrolytic's can be both cranky and very forgiving at the same
time, and just because one blows at a certain time doesn't mean the
event that caused it to blow happened WHEN it blew.
For example, at work I was talking to a colleague about a few things
when out of the blue a cap when boom. Upon closer inspection it was
revealed that they had used a 6.3V cap on a 12V line, simple right?
Wrong. The board had been powered up for DAYS! :) It just took that long
for the cap to go, which surprised us. I guess that's a case where
buying a LOWER quality cap is a good idea, it'll withstand abuse for
less time! :)
I say replace the cap and keep going, if you're absolutely sure about
the voltages and polarity involved. TTYL
-----------------------------
Herbert's PIC Stuff:
http://repatch.dyndns.org:8383/pic_stuff/
____________________________________________
2004\11\24@142007
by
Dave VanHorn
At 01:28 PM 11/24/2004, Olin Lathrop wrote:
>PicDude wrote:
> > Late last night, while building a 3-output power supply, a capacitor
> > burned out (it just made a poof sound and the electrolyte leaked out)
> > and it's totally baffling me.
>
>Most likely possibility: The capacitor was in backwards. They do take it
>for a while, then BOOM. I know you said you checked this. Check it again.
>I still think it's the most likely explanation.
How is that output regulated?
I've seen flyback switchers do this where the regulated output has a heavy
load, and the unregged outputs voltage went higher than the designer
anticipated.
____________________________________________
2004\11\24@143820
by
Robert Rolf
|
Too high ripple current?? I have seen switchers which
parallel several lower value caps to get ESR down and
handle high ripple currents.
Could it have been one of the 'bad electrolyte' caps
made in the 2001 that caused so much grief with mother
board makers (and caused 100ks of board failures in the field)?
Look for "'bad capacitor' electrolyte motherboard" on google for
the details. e.g.
http://www.spectrum.ieee.org/WEBONLY/resource/feb03/ncap.html
"It has all the elements of a good thriller: a stolen secret formula,
bungled corporate espionage, untraceable goods, and lone wolves saving
the little guy from the misdeeds of multinational corporations. In this
case, a mistake in the stolen formulation of the electrolyte in a
capacitor has wrecked hundreds of PCs and may wreck still more in what
is an industrywide problem.
....
Most of the leaking capacitors pulled from bad boards in the United
States, according to repair people, were labeled Tayeh, not a brand
affiliated with known capacitor makers. Many others were unmarked.
"
======================
This is exactly what has happened to several of my 2002 vintage
Gigabyte boards.
http://www.spectrum.ieee.org/WEBONLY/resource/feb03/ncap01.jpg
R
Dave VanHorn wrote:
{Quote hidden}> At 01:28 PM 11/24/2004, Olin Lathrop wrote:
>
>> PicDude wrote:
>> > Late last night, while building a 3-output power supply, a capacitor
>> > burned out (it just made a poof sound and the electrolyte leaked out)
>> > and it's totally baffling me.
>>
>> Most likely possibility: The capacitor was in backwards. They do
>> take it
>> for a while, then BOOM. I know you said you checked this. Check it
>> again.
>> I still think it's the most likely explanation.
>
>
> How is that output regulated?
>
> I've seen flyback switchers do this where the regulated output has a
> heavy load, and the unregged outputs voltage went higher than the
> designer anticipated.
>
> ______________________________________________
2004\11\24@145426
by
alan smith
I'd go with ripple current, especially since it worked
before just fine, but adding more stages that might be
switching in sync, will increase the ripple current
quite abit. Thats why when doing mulitple slaves, if
you can run on different frequencies, you can reduce
the input ripple. Might need to look at increaseing
the number of caps on the front end as well.
__________________________________
Do you Yahoo!?
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com
____________________________________________
2004\11\24@161703
by
John Ferrell
Sounds like there might have been some unwanted oscillation. Try it again
with a bypass cap across the electrolytic.
John Ferrell
My Competition is not my enemy!
http://DixieNC.US
{Original Message removed}
2004\11\24@183011
by
Support - KF4HAZ
> How would I determine the min load required? The datasheet does
> not indicate how to determine this.
Usually the data sheet will specify the min. load if there is one, but
connecting a scope or at least a voltmeter with no-load should tell you if one is required and if it is you can start loading it till the voltage begins to come close to spec.
KF4HAZ - Lonnie
____________________________________________
2004\11\25@092952
by
PicDude
On Wednesday 24 November 2004 01:54 pm, alan smith scribbled:
> I'd go with ripple current, especially since it worked
> before just fine, but adding more stages that might be
> switching in sync, will increase the ripple current
> quite abit. Thats why when doing mulitple slaves, if
> you can run on different frequencies, you can reduce
> the input ripple. Might need to look at increaseing
> the number of caps on the front end as well.
These are all running at different frequencies, since they're pretty much 3
independent circuits -- 72khz, 52khz, and 260khz.
Cheers,
-Neil.
____________________________________________
2004\11\25@092955
by
PicDude
On Wednesday 24 November 2004 01:19 pm, Dave VanHorn scribbled:
> How is that output regulated?
Not sure what you're asking here. This is a basic buck converter (MC33167),
regulated internally by the chip using it's feedback input.
> I've seen flyback switchers do this where the regulated output has a heavy
> load, and the unregged outputs voltage went higher than the designer
> anticipated.
Hmmm... but this output, and all outputs are regulated. This particular cap
is also rated at 35V, but the voltage at this point is 13.5V.
Cheers
-Neil.
____________________________________________
2004\11\25@092955
by
PicDude
On Wednesday 24 November 2004 01:38 pm, Robert Rolf scribbled:
> Too high ripple current?? I have seen switchers which
> parallel several lower value caps to get ESR down and
> handle high ripple currents.
I'm going to check for that when I replace the cap. This is a low-ESR cap. I
noticed that higher-voltage caps also have less ESR, and that's why I used a
35V cap here.
> Could it have been one of the 'bad electrolyte' caps
> made in the 2001 that caused so much grief with mother
> ...
Ugh! Earlier this year, someone gave me a couple scrap motherboards that had
died because of bad caps, and I found this odd. It's all making sense now.
This cap was brand spanking new, so hopefully it's not one of those, but then
again, if there is a problem with a component, I'll usually find out within
the first few hours of use.
Cheers,
-Neil.
____________________________________________
2004\11\25@093012
by
PicDude
On Wednesday 24 November 2004 01:15 pm, Herbert Graf scribbled:
> Well it looks like you've covered the bases, so it's simply possible you
> got a bad cap. I've seen that from time to time.
>
I'm really hoping this is all there is to it. But I have to wait until I can
order another low-ESR cap.
> FWIW electrolytic's can be both cranky and very forgiving at the same
> ...
> I say replace the cap and keep going, if you're absolutely sure about
> the voltages and polarity involved. TTYL
I'll get a scope on it the next time to see if anything unexpected is
happening.
Cheers,
-Neil.
____________________________________________
2004\11\25@093053
by
PicDude
|
|
On Wednesday 24 November 2004 12:28 pm, Olin Lathrop scribbled:
> Most likely possibility: The capacitor was in backwards. They do take it
> for a while, then BOOM. I know you said you checked this. Check it again.
> I still think it's the most likely explanation.
Re-verified, and it is correct. I would've loved this to be the problem, as
I'd have a simple fix. :-(
> Second possibility: The ripple current was way too high for the cap, which
> heated it up and eventually caused it to explode.
>
> Definitely replace it. It could look like an open, a short, or a toxic
> fume generator.
I'll replace and get a scope on it to see how much ripple is actually there.
I was powering it off of an older switcher, which could have bad ripple.
> Not true. Some commercial switching power supplies do indeed need a
> minimum load, but this is not inherent to switchers, only particular
> implementations.
>
> If it has a minimum load, this should be in the data sheet. If the data
> sheet doesn't mention it, then I would assume it has no minimum load
> requirement.
Okay ... assumption on my part.
Cheers,
-Neil.
____________________________________________
2004\11\25@100734
by
madscientist
|
this is true of most parts, particularly about not necessarily blowing
immediately when "abused". components often fail some time after some
initial abuse or repeated abuse. this is very common in the case of
voltage surges and lightning etc. i've repaired a lightning damaged
answering machine, had more parts fail within hours, replaced those and
had others fail the next day before i gave up and realized it had taken
a big enough hit to marginalize most or all of the parts. i've seen
this with computers as well when only part of it (say a modem port)
seems to have failed at first and then more problems crop up weeks or
months later.
it's also painfully easy to make the same mistake over and over when
inserting parts, i've done this once or twice, in one case it took a
month for me to figure out why all of the apd modules were "bad", when
in fact i kept putting in a wrong resistor on 3 different days over a 2
week period, and i had built plenty of these boards before. it simply
did not occur to me that i was using too small a resistor to terminate a
line and that was why the signal was weak, no one else spotted it
either, i was most embarrassed a month later when i figured it out,
fortunately we had not thrown the "bad" units away and they were easy to
fix (since the apd modules were about $300 each..). it's possible for a
board to be marked incorrectly or confusingly as well.
when in doubt use brute force trouble shooting, i.e. use a meter and a
scope and don't assume anything is "right" (and look at the connections
on the board and the signal/power flow to make sure they make sense if
you designed it). i've seen commercial boards with incorrect polarity
markings before as well in consumer gear where they just put the part in
correctly even though the marking was wrong so it's important to check
before replacing parts that the board actually agrees with how things
are inserted and/or that it is correct in the first place.
most equipment failures i've seen come from components that are slightly
over stressed continuously or intermittently and just take years to
finally fail. tv repair shops usually know which parts "always" fail on
a given chassis, i did this for awhile and it is amazingly accurate. certain parts just wind up being slightly over stressed because things
weren't properly calculated or because there are "surge" or startup or
abnormal but common operating conditions that weren't considered when
the parts were speced out. i've seen slightly abused resistors that
burned in half but still looked nearly normal and the ends would touch,
most of the time creating an intermittent. parts do sometimes fail in
atypical ways. i've seen a ceramic cap fail under very modest operating
conditions with no obvious physical problem until i swapped it out in
desperation (i was swapping all the parts of that stage at that point,
if it made sense or not, and ceramic caps nearly never fail). sometimes
brute force is necessary, particularly if it's something you just
designed. sometimes other people won't even be able to find mistakes
that will make you feel terribly silly when you do finally figure it
out, don't feel bad, we're human, we make random mistakes sometimes, all
of us.
Herbert Graf wrote:
-------
> FWIW electrolytic's can be both cranky and very forgiving at the same
> time, and just because one blows at a certain time doesn't mean the
> event that caused it to blow happened WHEN it blew.
---------
-- “Cowardice asks the question: is it safe? Expediency asks the question:
is it politic? Vanity asks the question: is it popular? But conscience
asks the question: is it right? And there comes a time when one must
take a position that is neither safe, nor politic, nor popular- but one
must take it simply because it is right.” : Martin Luther King Jr.
1929-1968 <http://www.guardian.co.uk/worldlatest/story/0,1280,-4614717,00.html>
___________________________________________
2004\11\25@103357
by
madscientist
|
|
just be very careful with the grounding, it is very hard to get good
scope traces in a switching power supply, especially with multiple
switchers running because there is a lot of noise. most of the switcher
data sheets/app notes i've read go to great lengths to explain this and
show what is necessary to get a meaningful scope trace in a single
switcher, with 3 switchers running it's worse. did you design the
circuit board? i know layout is somewhat critical with switchers.
PicDude wrote:
------
> I'll get a scope on it the next time to see if anything unexpected is
> happening.
--------
-- “Cowardice asks the question: is it safe? Expediency asks the question:
is it politic? Vanity asks the question: is it popular? But conscience
asks the question: is it right? And there comes a time when one must
take a position that is neither safe, nor politic, nor popular- but one
must take it simply because it is right.” : Martin Luther King Jr.
1929-1968 <http://www.guardian.co.uk/worldlatest/story/0,1280,-4614717,00.html>
___________________________________________
2004\11\25@104316
by
Dave VanHorn
At 09:36 AM 11/25/2004, the madscientist wrote:
>just be very careful with the grounding, it is very hard to get good
>scope traces in a switching power supply, especially with multiple
>switchers running because there is a lot of noise. most of the switcher
>data sheets/app notes i've read go to great lengths to explain this and
>show what is necessary to get a meaningful scope trace in a single
>switcher, with 3 switchers running it's worse. did you design the
>circuit board? i know layout is somewhat critical with switchers.
A current probe, even a minimal one, is very handy with switchers.
It's one thing to calculate ripple current in a cap, quite another
sometimes to actually measure it.
____________________________________________
2004\11\25@122939
by
PicDude
|
|
FWIW, I was thinking that I wouldn't be sleep properly for a week or so until
I can order/receive a new low-ESR cap (330uf), so I dug up a 1000uf 50V
generic-ESR electrolytic and soldered it in place. The PS works beautifully
now. And the cap is oriented (polarity) the same as the blown cap. I will
replace with the correct cap when I get it though.
I still need to see what sort of ripple I get on it (my scope is not here
currently), heatsink the chips, and test will all outputs under
significant/max load, but at this point all looks good so far.
BTW, on the subject of minimum loads, the datasheet for National's LM2678 does
not state anything about minimum load, but I noticed that the output voltage
was off (~4.2V) without a load, but came back within spec with a 1k load
resistor in place. The other outputs have the proper voltages w/o a load.
Much thanks,
-Neil.
On Thursday 25 November 2004 08:36 am, the madscientist scribbled:
{Quote hidden}> just be very careful with the grounding, it is very hard to get good
> scope traces in a switching power supply, especially with multiple
> switchers running because there is a lot of noise. most of the switcher
> data sheets/app notes i've read go to great lengths to explain this and
> show what is necessary to get a meaningful scope trace in a single
> switcher, with 3 switchers running it's worse. did you design the
> circuit board? i know layout is somewhat critical with switchers.
>
> PicDude wrote:
> ------
>
> > I'll get a scope on it the next time to see if anything unexpected is
> > happening.
>
> --------
____________________________________________
2004\11\25@130835
by
Alan B. Pearce
>BTW, on the subject of minimum loads, the datasheet for
>National's LM2678 does not state anything about minimum
>load, but I noticed that the output voltage was off
>(~4.2V) without a load, but came back within spec with
>a 1k load resistor in place. The other outputs have
>the proper voltages w/o a load.
depending what output current you designed for, it may have gone into a
discontinuous mode, and that may give strange currents in the input cap,
which could have caused it to blow despite being low ESR.
On the other hand it may just have been a faulty cap, possibly even labelled
with the wrong polarity.
____________________________________________
2004\11\25@135007
by
Richard.Prosser
And then there's parts that are incorrectly marked.
I've seen this with both electrolytics and diodes. Not counting counterfeit
parts that are overstressed or just incorrect.
RP
this is true of most parts, particularly about not necessarily blowing
immediately when "abused". components often fail some time after some
initial abuse or repeated abuse. this is very common in the case of
voltage surges and lightning etc. i've repaired a lightning damaged
answering machine, had more parts fail within hours, replaced those and
had others fail the next day before i gave up and realized it had taken
a big enough hit to marginalize most or all of the parts. i've seen
this with computers as well when only part of it (say a modem port)
seems to have failed at first and then more problems crop up weeks or
months later.
>snip ......
____________________________________________
2004\11\25@152256
by
Sascha
You may have a grounding problem and due to noise in the ground line to get
more than 35V on the cap. I had the same problem last week on a 100W 13V to
200V push-pull converter running at 250kHz. In my case the smps controller
blew up although it is rated for 40V and in my circuit it gets max. 18V. But
due to a bed grounding strategy the noise vas greater than 40V and in a
matter of 1-2 hours the controller got burned. Provide separate return paths
for ground currents to the lowest impedance point and as large traces as you
can afford. This worked for me.
Sascha
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.289 / Virus Database: 265.4.2 - Release Date: 11/24/2004
____________________________________________
2004\11\25@202512
by
Martin Klingensmith
|
|
PicDude wrote:
{Quote hidden}>Hi all,
>
>Late last night, while building a 3-output power supply, a capacitor burned
>out (it just made a poof sound and the electrolyte leaked out) and it's
>totally baffling me.
>
>I built the 12V stage, tested it under a load, and it worked fine. Then I
>built the 5V stage, and that worked fine under load. When I built the final,
>3.3V stage and was testing it, the input capacitor for the 5V stage died.
>This time though, the 5V stage was NOT under load. I immediately thought
>that the output capacitor must've died since it was not under load and
>perhaps the output voltage went too high, but it was actually the input
>capacitor (Nichicon 330uf, 35V, low-ESR electrolytic).
>
>In more detail, the input voltage (which in this case was 14V from a switching
>power supply) is split in 3 directions, each going to a separate SB540
>schottky diode (for reverse-voltage protection), then to the switching
>circuitry for each stage. The 5V switcher is an OnSemi MC33167, with a 330uf
>35V input capacitor. Because there are reverse-protection diodes on there, I
>can't see how anything happening on either of the other stages could cause a
>reverse voltage to the 330uf cap. And since my input is 14V, that cap was
>receiving a supply of ~13.5V (after the reverse-protection diode).
>
>
I'll assume your 5v switcher is a buck topology - what frequency is it
and what size inductor did you use before the cap?
--
Martin Klingensmith
____________________________________________
2004\11\27@031209
by
PicDude
|
|
On Thursday 25 November 2004 02:07 pm, Sascha scribbled:
> You may have a grounding problem and due to noise in the ground line to get
> more than 35V on the cap. I had the same problem last week on a 100W 13V to
> 200V push-pull converter running at 250kHz. In my case the smps controller
> blew up although it is rated for 40V and in my circuit it gets max. 18V.
> But due to a bed grounding strategy the noise vas greater than 40V and in a
> matter of 1-2 hours the controller got burned. Provide separate return
> paths for ground currents to the lowest impedance point and as large traces
> as you can afford. This worked for me.
>
> Sascha
I did follow as many of the PCB design rules as possible (large traces, ground
plane, short runs, etc).
I had also thoroughly inspected the board after (and before) assembly and did
not find any bad joints. Yes, it was only visual, but since changing the
cap, it's been working fine. I'm currently installing an OS, with the PS
powering the mobo, small drive and CD.
Cheers,
-Neil.
____________________________________________
2004\11\27@031424
by
PicDude
On Thursday 25 November 2004 07:25 pm, Martin Klingensmith scribbled:
>
> I'll assume your 5v switcher is a buck topology - what frequency is it
> and what size inductor did you use before the cap?
It's an OnSemi MC33167 chip running at 72khz. There is no inductor *before*
the cap -- this was the input capacitor that failed.
Cheers,
-Neil.
____________________________________________
2004\11\27@114001
by
Martin Klingensmith
PicDude wrote:
>On Thursday 25 November 2004 07:25 pm, Martin Klingensmith scribbled:
>
>
>>I'll assume your 5v switcher is a buck topology - what frequency is it
>>and what size inductor did you use before the cap?
>>
>>
>
>It's an OnSemi MC33167 chip running at 72khz. There is no inductor *before*
>the cap -- this was the input capacitor that failed.
>
>Cheers,
>-Neil.
>
>
I thought it was the output capacitor that failed, I was just making
sure you were using an inductor on the output. (you never know!)
--
Martin K
____________________________________________
2004\11\27@142818
by
Dave VanHorn
>I thought it was the output capacitor that failed, I was just making sure
>you were using an inductor on the output. (you never know!)
The input side has plenty ripple current as well.
____________________________________________
More... (looser matching)
- Last day of these posts
- In 2004
, 2005 only
- Today
- New search...
