capacitor failure mode etc ?

[N_Cook]

2x 250V , 2.2uF yellow epoxy sealed polyester? caps used as intermediary
high DC storage in an electric fence unit. Failed as one of these caps is
fairly consistently ohmic at 370 ohm.
Now repaired 195V dc over them ,peak. Used in the wettest part of the UK ,
Cumbria, so assumed due to condensation getting inside the cap, no direct
rain ingress traces seen.
But removing the cap casing and pulling back the leads , no trace of
green/corrossion. Heating , with low heat of a hot air gun , ohms drop to
350 and then back up again on cooling.
Failure due to damp or HV punch through? .
This is PbF solder but is RoHS a consideration inside such caps , ie
dendrite formation? If punch through then does the oscillator need knocking
back a bit, 2 presets on the board. Replaced both of these caps and another
one used in the pulse shaping drive to the "pulse Tx" , the usual backwards
mains Tx. Thinking perhaps if unit is in direct sunlight perhaps the pump
circuit gives out 300V say. Schematic not seen but HV zeners in that area
270V + 16V plus another , voltage not seen but probably 16V.
I was inclined to drop the value of the 270V one but I assume these are
fudge values, "selected on test" for nA leakage at less than combined knee
voltage of about 300V and has been working for 5 years , then presumably ok
to leave as is , as these zeners DVM diode test at least as normal.

 

[N_Cook]

nesesu <neil_sutcliffe@telus.net> wrote in message
news:edbd6ce2-b688-43d6-8566-9491811c3ca1@googlegroups.com...
On Thursday, December 20, 2012 2:50:30 AM UTC-8, N_Cook wrote:

2x 250V , 2.2uF yellow epoxy sealed polyester? caps used as intermediary
high DC storage in an electric fence unit. Failed as one of these caps is

fairly consistently ohmic at 370 ohm. Now repaired 195V dc over them ,peak.
Used in the wettest part of the UK , Cumbria, so assumed due to condensation
getting inside the cap, no direct rain ingress traces seen. But removing the
cap casing and pulling back the leads , no trace of green/corrossion.
Heating , with low heat of a hot air gun , ohms drop to 350 and then back up
again on cooling. Failure due to damp or HV punch through? . This is PbF
solder but is RoHS a consideration inside such caps , ie dendrite formation?
If punch through then does the oscillator need knocking back a bit, 2
presets on the board. Replaced both of these caps and another one used in
the pulse shaping drive to the "pulse Tx" , the usual backwards mains Tx.
Thinking perhaps if unit is in direct sunlight perhaps the pump circuit
gives out 300V say. Schematic not seen but HV zeners in that area 270V + 16V
plus another , voltage not seen but probably 16V. I was inclined to drop the
value of the 270V one but I assume these are fudge values, "selected on
test" for nA leakage at less than combined knee voltage of about 300V and
has been working for 5 years , then presumably ok to leave as is , as these
zeners DVM diode test at least as normal.

As you say, there are several factors involved here. Indeed it could be
moisture ingress or it could be punch through. The punch through could be
from a spike or just a weak point in the dielectric that failed under long
term stress somewhat below rating. Since this is in a fence charger I would
also consider lightning. I have repaired a couple of fence chargers that
suffered from induced transients from nearby lightning, and in both cases
the caps had failed to high leakage. One unit was an electromechanical type
where the cap across the points had gone leaky, the other was similar to
what you have and the storage cap was blown.

Neil S.

++++


I've come across a near lightning struck one before and it knocked out a
custom PIC/ASIC with no work around. I would gave expected that more likely
than high V caps failing from spikes. This 2006 made unit oddly could have
be made in 1976 all discrete , assuming a Programmable Unijunction Transitor
for the SCR was available in 1976 . I had asked the owner about lightning
in area and he reckoned not.
I wonder what other test to determine if punch through or metalisation
creep/dendrite ? only for professional interest reasons, I'm reasonably
content to return, as is , without any mods. As lightning is "act of God"
cannot sensibly mitigate against that. One advantage of old circuits -
nearly always repeairable

 

[nesesu]

On Thursday, December 20, 2012 2:50:30 AM UTC-8, N_Cook wrote:

2x 250V , 2.2uF yellow epoxy sealed polyester? caps used as intermediary high DC storage in an electric fence unit. Failed as one of these caps is fairly consistently ohmic at 370 ohm. Now repaired 195V dc over them ,peak. Used in the wettest part of the UK , Cumbria, so assumed due to condensation getting inside the cap, no direct rain ingress traces seen. But removing the cap casing and pulling back the leads , no trace of green/corrossion. Heating , with low heat of a hot air gun , ohms drop to 350 and then back up again on cooling. Failure due to damp or HV punch through? . This is PbF solder but is RoHS a consideration inside such caps , ie dendrite formation? If punch through then does the oscillator need knocking back a bit, 2 presets on the board. Replaced both of these caps and another one used in the pulse shaping drive to the "pulse Tx" , the usual backwards mains Tx. Thinking perhaps if unit is in direct sunlight perhaps the pump circuit gives out 300V say. Schematic not seen but HV zeners in that area 270V + 16V plus another , voltage not seen but probably 16V. I was inclined to drop the value of the 270V one but I assume these are fudge values, "selected on test" for nA leakage at less than combined knee voltage of about 300V and has been working for 5 years , then presumably ok to leave as is , as these zeners DVM diode test at least as normal.

As you say, there are several factors involved here. Indeed it could be moisture ingress or it could be punch through. The punch through could be from a spike or just a weak point in the dielectric that failed under long term stress somewhat below rating. Since this is in a fence charger I would also consider lightning. I have repaired a couple of fence chargers that suffered from induced transients from nearby lightning, and in both cases the caps had failed to high leakage. One unit was an electromechanical type where the cap across the points had gone leaky, the other was similar to what you have and the storage cap was blown.

Neil S.

 

Lead free solder should have nothing to do with hermetic sealing really, unless you are talking the tinning of the leads. Even that is pretty far fetched I think. The tinning is very thin and the thermal expansion coeffeicient of the base wire underneath should be the major factor.

If you want my highly respected carefully considered professional opinion - I think the cheap ass thing just went bad.

 

"Unijunction Transitor
for the SCR was available in 1976"

Oh definitely. I think every Tektronix manual since about 1965 had a tutorial on them. (and I STILL coudn't design a circuit using one they are too wierd lol) Actually UJT triggering was one of the things about good scopes, they would lock up really fast. I had a 2465 for a time and it's sync sucked.. I mean it DID actually sync, but it took too damn long. I was spoiled by older scopes that would lock up faster than the shutdown could shut down the TV. I could cop a collector waveform off the HOT (or LOT for you over there) fast enogh to see it, and then start unconnecting things. It was quite useful.

Years ago there was a time when if I didn't have a shutdown problem figured out in about five minutes, scrapping the set was considered. I had my routine and it worked. I'm now working on a new routine of course.

Of couirse it is all useless knowledge now. I feel like a fucking trainee now.

 

[N_Cook]

I've started unwinding this and it would seem to be wide scale
electrochemical breakdown. Each turn I unwind the ohmage goes up each time I
measure it, now 700 ohms or so. The unwound film looks like a piece of 35 mm
projection film of white image and no sprocket holes. Milky clear patches,
not clear polyester, between "silvering" that is wavy at the edges,
periodicity of the half the length per turn, and a repeat strip like 35 mm
film strip, twice per turn. The silvering of each large flat face seems to
have disappeared leaving it at the tight bends, twice per turn
Of course I don't know what it should look like and will have to break into
a new, different make one, to check and also one of the other 2 of the same
batch as the failed one , both measure near enough 2.2uF.
And try taking pics of it all

 

[N_Cook]

Here is a pic.
http://www.diverse.4mg.com/duff_cap.jpg
The leftmost is the bad cap, centre an older but unused Philips also 2.2uF
250V, and the third is from the same batch as the bad, not ohmic up to
30Mohm anyway

The 2 red >s are "frames" 10&11 that show the curving and the graph paper
behind where all the "silvering" is absent, so film colours consist of
silver, grey colour , hazy white and clear .
Centre is good "silver" and crystal clear polyester , only colours, with
alternate gutters for the edge connection, metal lead masses retained so
ragged edges from my unwinding.
Right one shows the same effect on the outer 4 frames but beyond that the
metal foil is entire, but going home as the other.
Yellow case is the bad TC make and blue is 3 sides of the Philips puter
case. The philips actual capacitor volume is much less than the TC so epoxy
filling , inside the block casing is complete, but both the others you can
see the active film in a number of places , after removing the block plastic
casing. This epoxy? inner encapsulation was much easier to break away than
the Philips. I suspect the Philips used 100% epoxy and the TC had fillered
epoxy so maybe making porous. No green corrossion products seen but I assume
capillary migration of condensation through the poor epoxy from the open end
with the leads or along the leads which were bright shiney in all of them.
So this will be a generic fault with Electric Shepherd ESB55 , first
becoming apparent in the wettest county of Cumberland in 5 years of use and
then other counties later presumably

 

[Cydrome Leader]

N_Cook <diverse@tcp.co.uk> wrote:

Here is a pic.
http://www.diverse.4mg.com/duff_cap.jpg
The leftmost is the bad cap, centre an older but unused Philips also 2.2uF
250V, and the third is from the same batch as the bad, not ohmic up to
30Mohm anyway

The 2 red >s are "frames" 10&11 that show the curving and the graph paper
behind where all the "silvering" is absent, so film colours consist of
silver, grey colour , hazy white and clear .
Centre is good "silver" and crystal clear polyester , only colours, with
alternate gutters for the edge connection, metal lead masses retained so
ragged edges from my unwinding.
Right one shows the same effect on the outer 4 frames but beyond that the
metal foil is entire, but going home as the other.
Yellow case is the bad TC make and blue is 3 sides of the Philips puter
case. The philips actual capacitor volume is much less than the TC so epoxy
filling , inside the block casing is complete, but both the others you can
see the active film in a number of places , after removing the block plastic
casing. This epoxy? inner encapsulation was much easier to break away than
the Philips. I suspect the Philips used 100% epoxy and the TC had fillered
epoxy so maybe making porous. No green corrossion products seen but I assume
capillary migration of condensation through the poor epoxy from the open end
with the leads or along the leads which were bright shiney in all of them.
So this will be a generic fault with Electric Shepherd ESB55 , first
becoming apparent in the wettest county of Cumberland in 5 years of use and
then other counties later presumably

I can't tell anything from the tiny photo, but a lack of metallization
sounds like the cap had the elctrodes burn from constant overvoltage or
breakdown/overload/ With such a low resistance, it may not be possible to
measure
the actual capacitance of the bad caps. It will drop as "self healing"
caps start to wear out. Something around 5% loss is considered end of life
for many applications.

Are you able to burn test the dielectric to see what it really is?

 

[N_Cook]

Cydrome Leader <presence@MUNGEpanix.com> wrote in message
news:kb2e12$ltf$1@reader1.panix.com...

N_Cook <diverse@tcp.co.uk> wrote:
Here is a pic.
http://www.diverse.4mg.com/duff_cap.jpg
The leftmost is the bad cap, centre an older but unused Philips also
2.2uF
250V, and the third is from the same batch as the bad, not ohmic up to
30Mohm anyway

The 2 red >s are "frames" 10&11 that show the curving and the graph
paper
behind where all the "silvering" is absent, so film colours consist of
silver, grey colour , hazy white and clear .
Centre is good "silver" and crystal clear polyester , only colours, with
alternate gutters for the edge connection, metal lead masses retained so
ragged edges from my unwinding.
Right one shows the same effect on the outer 4 frames but beyond that
the
metal foil is entire, but going home as the other.
Yellow case is the bad TC make and blue is 3 sides of the Philips puter
case. The philips actual capacitor volume is much less than the TC so
epoxy
filling , inside the block casing is complete, but both the others you
can
see the active film in a number of places , after removing the block
plastic
casing. This epoxy? inner encapsulation was much easier to break away
than
the Philips. I suspect the Philips used 100% epoxy and the TC had
fillered
epoxy so maybe making porous. No green corrossion products seen but I
assume
capillary migration of condensation through the poor epoxy from the open
end
with the leads or along the leads which were bright shiney in all of
them.
So this will be a generic fault with Electric Shepherd ESB55 , first
becoming apparent in the wettest county of Cumberland in 5 years of use
and
then other counties later presumably



I can't tell anything from the tiny photo, but a lack of metallization
sounds like the cap had the elctrodes burn from constant overvoltage or
breakdown/overload/ With such a low resistance, it may not be possible to
measure
the actual capacitance of the bad caps. It will drop as "self healing"
caps start to wear out. Something around 5% loss is considered end of life
for many applications.

Are you able to burn test the dielectric to see what it really is?

I'm assuming, although no water trails seen, that it starts with capilliary
action bringing dampness between the outer casing and the capacitor proper
as in Philips and TC the epoxy fill does not bond to the outer case , it
pulls away easily.
Then something to do with the 200V, perhaps the curving edges , maximum
width at the middle of the flat faces is something to do with charge
distribution. As powdered aluminium is grey I'm assuming the grey colour of
the depleted metal is finely divided Al and the white is Aluminium oxide.
So you reckon millions of point discharges have erroded away the metal ?
sounds logical but would not the plastic show signs of roughening, I will
have a look under a microscope. The discharges have to be through the
plastic, would dampness enhance that, considering the effect starts at the
outer layers , not the inner, it would seem
I would not know what to smell for / test for, with burning
or do you mean melting and then
The melting point of HDPE (High Density Polyethelyne) is about 130 şC
The melting point of LDPE (Low Density Polyethelyne) is about 110 şC
The melting point of PET (Polyethylene terphthalate) is about 250-260 şC
The melting point of PP (Polypropylene) is about 160-170 şC
The melting point of PS (Polystyrene) is about 70-115 şC
The melting point of PVC (Polyvinyl Chloride) is about 75-90 şC

 

[N_Cook]

Cydrome Leader <presence@MUNGEpanix.com> wrote in message
news:kb2e12$ltf$1@reader1.panix.com...

N_Cook <diverse@tcp.co.uk> wrote:
Here is a pic.
http://www.diverse.4mg.com/duff_cap.jpg
The leftmost is the bad cap, centre an older but unused Philips also
2.2uF
250V, and the third is from the same batch as the bad, not ohmic up to
30Mohm anyway

The 2 red >s are "frames" 10&11 that show the curving and the graph
paper
behind where all the "silvering" is absent, so film colours consist of
silver, grey colour , hazy white and clear .
Centre is good "silver" and crystal clear polyester , only colours, with
alternate gutters for the edge connection, metal lead masses retained so
ragged edges from my unwinding.
Right one shows the same effect on the outer 4 frames but beyond that
the
metal foil is entire, but going home as the other.
Yellow case is the bad TC make and blue is 3 sides of the Philips puter
case. The philips actual capacitor volume is much less than the TC so
epoxy
filling , inside the block casing is complete, but both the others you
can
see the active film in a number of places , after removing the block
plastic
casing. This epoxy? inner encapsulation was much easier to break away
than
the Philips. I suspect the Philips used 100% epoxy and the TC had
fillered
epoxy so maybe making porous. No green corrossion products seen but I
assume
capillary migration of condensation through the poor epoxy from the open
end
with the leads or along the leads which were bright shiney in all of
them.
So this will be a generic fault with Electric Shepherd ESB55 , first
becoming apparent in the wettest county of Cumberland in 5 years of use
and
then other counties later presumably



I can't tell anything from the tiny photo, but a lack of metallization
sounds like the cap had the elctrodes burn from constant overvoltage or
breakdown/overload/ With such a low resistance, it may not be possible to
measure
the actual capacitance of the bad caps. It will drop as "self healing"
caps start to wear out. Something around 5% loss is considered end of life
for many applications.

Are you able to burn test the dielectric to see what it really is?

by burn test , this sort of discrimination?
http://www.boedeker.com/burntest.htm

 

[N_Cook]

Along the way I seem to have misinterpreted the meaning of "self-healing" as
the plastic is punctured and then melts to reseal. But seemingly it is the
aluminium that "spark erodes" away to something, what material ? and then
cannot discharge to that point again and termed healing.
Via microscope lots of random tiny holes in the aluminium in the bad
sections, no holes seen in the plastic.
Next thing is to identify the insulator film as I assume some plastics are
porous or whatever the term is under the presence of water/water vapour to
reduce the electrostatic breakdown potential.

 

[N_Cook]

revised pic on that URL
the Philips 373 MKT was polyester as was the yellow TC probably, by that
flame test.
Yellow flame , sustained after flame removal, no drips, slow
progress,brittle remains, although HCl not smelled.
This film 15 microns, WVTR permeability of 25 micron films in g/m^2/day
polyester 4
Polythene 20
Polycarbonate 170, I didn;t realise this was so "porous"
From the boating industry epoxy has much better permeability figure than
polyester.

Now just to find the quantative deleterious effect of water/vapour on
dielectric strength.

 

[Cydrome Leader]

N_Cook <diverse@tcp.co.uk> wrote:

revised pic on that URL
the Philips 373 MKT was polyester as was the yellow TC probably, by that
flame test.
Yellow flame , sustained after flame removal, no drips, slow
progress,brittle remains, although HCl not smelled.
This film 15 microns, WVTR permeability of 25 micron films in g/m^2/day
polyester 4
Polythene 20
Polycarbonate 170, I didn;t realise this was so "porous"
From the boating industry epoxy has much better permeability figure than
polyester.

Now just to find the quantative deleterious effect of water/vapour on
dielectric strength.

polyster is cheap, so that's usually what you'll come across first for a
DC capacitor.

Tape and fill caps aren't considered hermetically sealed. If you want
that you need a metal can, but those cost more money, so moisture ingress
isn't really too surprising- if that's the problem.

in the self healing caps, once the dielecric punches though, the super
thin aluminum coating burns up like a fuse, isolating that are with the
pinhole though the plastic. Everytime this happens, you lose more and more
plate area and capacitance drops.

the low resistance reading is strange though. Even damp nylon probably
won't conduct at much, and it really lik to absorb water.

 

[Cydrome Leader]

N_Cook <diverse@tcp.co.uk> wrote:

Cydrome Leader <presence@MUNGEpanix.com> wrote in message
news:kb2e12$ltf$1@reader1.panix.com...
N_Cook <diverse@tcp.co.uk> wrote:
Here is a pic.
http://www.diverse.4mg.com/duff_cap.jpg
The leftmost is the bad cap, centre an older but unused Philips also
2.2uF
250V, and the third is from the same batch as the bad, not ohmic up to
30Mohm anyway

The 2 red >s are "frames" 10&11 that show the curving and the graph
paper
behind where all the "silvering" is absent, so film colours consist of
silver, grey colour , hazy white and clear .
Centre is good "silver" and crystal clear polyester , only colours, with
alternate gutters for the edge connection, metal lead masses retained so
ragged edges from my unwinding.
Right one shows the same effect on the outer 4 frames but beyond that
the
metal foil is entire, but going home as the other.
Yellow case is the bad TC make and blue is 3 sides of the Philips puter
case. The philips actual capacitor volume is much less than the TC so
epoxy
filling , inside the block casing is complete, but both the others you
can
see the active film in a number of places , after removing the block
plastic
casing. This epoxy? inner encapsulation was much easier to break away
than
the Philips. I suspect the Philips used 100% epoxy and the TC had
fillered
epoxy so maybe making porous. No green corrossion products seen but I
assume
capillary migration of condensation through the poor epoxy from the open
end
with the leads or along the leads which were bright shiney in all of
them.
So this will be a generic fault with Electric Shepherd ESB55 , first
becoming apparent in the wettest county of Cumberland in 5 years of use
and
then other counties later presumably



I can't tell anything from the tiny photo, but a lack of metallization
sounds like the cap had the elctrodes burn from constant overvoltage or
breakdown/overload/ With such a low resistance, it may not be possible to
measure
the actual capacitance of the bad caps. It will drop as "self healing"
caps start to wear out. Something around 5% loss is considered end of life
for many applications.

Are you able to burn test the dielectric to see what it really is?




by burn test , this sort of discrimination?
http://www.boedeker.com/burntest.htm

you can do that or see how the film melts, tears or handles stretching
(changes color etc).

A DC filter-ish cap would likely be polyester, but they have other
plastics now these days in addition to the plain old polycarbonate and
polypropylene. Polystyrene caps are no longer made last I heard.

 

[N_Cook]

Cydrome Leader <presence@MUNGEpanix.com> wrote in message
news:kb8367$e6b$1@reader1.panix.com...

N_Cook <diverse@tcp.co.uk> wrote:
revised pic on that URL
the Philips 373 MKT was polyester as was the yellow TC probably, by that
flame test.
Yellow flame , sustained after flame removal, no drips, slow
progress,brittle remains, although HCl not smelled.
This film 15 microns, WVTR permeability of 25 micron films in g/m^2/day
polyester 4
Polythene 20
Polycarbonate 170, I didn;t realise this was so "porous"
From the boating industry epoxy has much better permeability figure than
polyester.

Now just to find the quantative deleterious effect of water/vapour on
dielectric strength.

polyster is cheap, so that's usually what you'll come across first for a
DC capacitor.

Tape and fill caps aren't considered hermetically sealed. If you want
that you need a metal can, but those cost more money, so moisture ingress
isn't really too surprising- if that's the problem.

in the self healing caps, once the dielecric punches though, the super
thin aluminum coating burns up like a fuse, isolating that are with the
pinhole though the plastic. Everytime this happens, you lose more and more
plate area and capacitance drops.

the low resistance reading is strange though. Even damp nylon probably
won't conduct at much, and it really lik to absorb water.

What happens to the "evaporated" aluminium from all those "35mm frames" ?
perhaps some of it migrates and settles at the edges as a slightly
conductive coating. It was interesting that as I unwound each layer of 2
"frames" the resistance incremented higher each time on a regular basis.
The real fault of these TC is the capacitor proper had to be squashed in the
outer shell and hence the moisture path in at the centre of those "frames"
as the potting was absent at those points , the silvering remains at the
tight curving sections of each turn . I'm amazed at how spectacularly
cleanly all the Al has erroded from those frames
, the remnant silvering is finely punctured though , looking via microscope.
Once an island of silvering emerges with a ring of eroded around it , what
mechanism ensures the erosion of the islands?

 

[Michael A. Terrell]

N_Cook wrote:

What happens to the "evaporated" aluminium from all those "35mm frames" ?

Ask the OEM. They understand the technology. You waste a lot of
time on useless things, and this is one of them, unless you're going to
manufacture a few million plastic caps every year.

The damned cap is closed, so it is redeposited near the puncture.
Where else could it go? In most applications for self healing
capacitors, it won't matter.