Electrolytic caps in series

[Glenn Baddeley]

Hi Phil,

OK, you've tempted me to post again. You are quite entitled to
your personal opinion, and I to mine. Most of your points are
quite valid and arguable, but I do see my point of view as
still plausible. Would anyone else care to throw in their
personal opinion or experiences?

Glenn.

Phil Allison wrote:

"Glenn Baddeley ** STOP TOP POSTING !!!!!!!

Phil Allison

has a DC leakage current which is much greater than the other
caps, >> > it will put a high voltage stress on the other caps, and
they may >> > go BANG.

** For a cap to do as you suggest, it must be a faulty cap - ie
one >> that is not able to meet maker's specs for max voltage or
leakage at >> rated voltage.

Maybe,


** What does "maybe " mean ???

Do you see the point or not ??

I'll post it again: " If neither elector cap is faulty - then
nothing bad can happen when wired in series and used at around 70% of
rated voltage. "

So far - you have simply not addressed this at all.

You have yet to post any sort of case.



but there is a low % of failure rate that will be encountered
eventually, especially after 20+ years. I would rather "insure" for
that possibility that risk a catastrophic failure and expensive
repair.


** Totally begs the question.


If one cap has 1mA leakage and another cap has 2mA leakage,
the additional voltage across one of the caps may lead to its
premature failure.


** That is a whole bunch of "ifs" - there.

The caps in question have tested at less than 10 uA of leakage at 70
% of rated voltage.

You are IGNORING facts and just making up numbers that suit you.



Ballast resistors are no help if a cap is faulty.

It depends on the type of fault.


** I carefully described the type of fault.

Go read my post again !!!!!!



** Wrong - WW resistors are only readily available in values up
to >> about 10 kohms - far too low a value for use in parallel with
high >> voltage electros.

eg. Design for 450 V across each cap, using 10W resistors.

R = E^2 / P = 450 x 450 / 7.5 = 27 KOhm. These are available
in 10W, or use two 12K or 15K 5W in series.
I = E / R = 17mA approx. This would protect an excess leakage
fault of 5 - 10 mA.


** Shame how the 47 uF caps in question only have 2 to 10 uA of
leakage.

Shame how 17 mA of extra current is not acceptable in a valve amp or
many other situations where such electros are used.

Shame how the 7.5 watts of heat from the resistors would damage the
electros if placed near them.

Shame how high value WW resistors are notorious for failing open
when subjected to continuous, high DC voltages.

Shame what a load of crapology you post.


The point is simple - you are better off without ballast resistors
across electros unless they are 100 % reliable.

Sorry, I don't agree.


** Huh - so it is OK by you if one resistor fails open and the
other then forces an electro overvoltage ???

Do you realise you just totally contradicted yourself ???


Every text book and article I have ever seen uses
resistors.


** Ok - so you rely on rote learning and the superstitions of
others.

How intelligent - NOT !!



(If you personally insult me again you will not receive any further
replies from me,


** You are posting mindless tripe - Glenn.

Worse, you INSIST on top posting - the method only used by fools
and wankers.


effectively halting this interesting objective discussion,


** Confusing the matter with irrational drivel is not my idea of
"objective".




........... Phil

 

[Terry Given]

Glenn Baddeley wrote:

Hi Phil,

OK, you've tempted me to post again. You are quite entitled to
your personal opinion, and I to mine. Most of your points are
quite valid and arguable, but I do see my point of view as
still plausible. Would anyone else care to throw in their
personal opinion or experiences?

Glenn.

Hi Glenn,

I have personally seen capacitors explode violently when overvoltaged,
due to failure of caps in a series/parallel string. Two 2200uF 450V caps
in series, paralleled 18 times (36 caps). DC bus voltage ranged from
560Vdc to 850Vdc. The cap bank was made of 6 identical sections, and
used 450mm x 600mm x 2mm Aluminium sheets as +Vdc and -Vdc, with 1mm
lexan sheet as insulation/spacing. Each individual section had a
separate Aluminium plate to join the center points of all 6 caps, and
each of these 6 plates were connected with 1mm^2 hookup wire. Each
section also had a pair of (IIRC) 27k "balancing" resistors across the
series/parallel array.

The capacitors had been rendered faulty by the injudicious application
of Methyl Bromide, which ate the cathode terminals. Within a 6-cap
array, one cap going open circuit shifts the voltage division from 50/50
to 60/40, which at max bus volts is 510V across one side, enough to
break the caps. When two paralleled caps go open, one side carries 75%
of the DC voltage. The resultant explosions tore the 2mm DC bus like
paper, and in one case blew the drive doors open (mangling the cheesy
door lock in the process). In all cases the 1mm^2 hookup wire fried.

The "balancing" resistors had absolutely no effect, as their impedance
was huge compared with the capacitor impedance when load current was
being drawn. They only affect the voltage balancing at DC, which with a
rectifier cap only occurs at NO LOAD. This was confirmed by measurement.
Cap leakage current specs also showed the 27k resistors could not be
relied upon to do any real DC balancing either, as the leakage spec was
quite high - IIRC we needed to throw away many tens of W to get proper
DC balancing. And the AC balancing is really the important part (hint:
what does the input current of a rectifier/capacitor filter look like).

OTOH I normally provide a discharge path of some description, usually
R's and LEDs to provide a "look out" indicator. To tell the truth its
more like a "for fucks sake dont stick a scope probe here, its live"
indicator

Cheers
Terry

 

[Phil Allison]

"Terry Given" <my_name@ieee.org> wrote in message
news:Llqwe.11187$U4.1428455@news.xtra.co.nz...

I have personally seen capacitors explode violently when overvoltaged, due
to failure of caps in a series/parallel string.........


( snip drivel)

The capacitors had been rendered faulty by the injudicious application of
Methyl Bromide, which ate the cathode terminals.

( snip more crazy stuff)


** So they don't just do weird things to sheep in NZ ........




.............. Phil

 

[Phil Allison]

"Glenn Baddeley" = still bloody top posting !!!


** STOP !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

And the AC balancing
is really the important part (hint: what does the input current of a
rectifier/capacitor filter look like).

A good point, I've never seen this aspect mentioned anywhere before.

** So how do you know it really is a good point ??





........... Phil

 

[Glenn Baddeley]

Two 2200uF 450V caps in series, paralleled 18 times (36 caps).

Wow, that's a pretty big cap bank.

Cap leakage current specs also showed the
27k resistors could not be relied upon to do any real DC balancing
either, as the leakage spec was quite high - IIRC we needed to throw
away many tens of W to get proper DC balancing.

.... apparently a lot higher than the 10uA example quoted by Phil.
Some HAM transmitters do throw away 10's watts to getter better
load / no load regulation from a 1 - 2 KVA tranny. It concerns
me that all that heat is generated near the caps.

And the AC balancing
is really the important part (hint: what does the input current of a
rectifier/capacitor filter look like).

A good point, I've never seen this aspect mentioned anywhere before.

Cheers,
Glenn.

Terry Given wrote:

I have personally seen capacitors explode violently when
overvoltaged, due to failure of caps in a series/parallel string. Two
2200uF 450V caps in series, paralleled 18 times (36 caps). DC bus
voltage ranged from 560Vdc to 850Vdc. The cap bank was made of 6
identical sections, and used 450mm x 600mm x 2mm Aluminium sheets as
+Vdc and -Vdc, with 1mm lexan sheet as insulation/spacing. Each
individual section had a separate Aluminium plate to join the center
points of all 6 caps, and each of these 6 plates were connected with
1mm^2 hookup wire. Each section also had a pair of (IIRC) 27k
"balancing" resistors across the series/parallel array.

The capacitors had been rendered faulty by the injudicious
application of Methyl Bromide, which ate the cathode terminals.
Within a 6-cap array, one cap going open circuit shifts the voltage
division from 50/50 to 60/40, which at max bus volts is 510V across
one side, enough to break the caps. When two paralleled caps go open,
one side carries 75% of the DC voltage. The resultant explosions tore
the 2mm DC bus like paper, and in one case blew the drive doors open
(mangling the cheesy door lock in the process). In all cases the
1mm^2 hookup wire fried.

The "balancing" resistors had absolutely no effect, as their
impedance was huge compared with the capacitor impedance when load
current was being drawn. They only affect the voltage balancing at
DC, which with a rectifier cap only occurs at NO LOAD. This was
confirmed by measurement. Cap leakage current specs also showed the
27k resistors could not be relied upon to do any real DC balancing
either, as the leakage spec was quite high - IIRC we needed to throw
away many tens of W to get proper DC balancing. And the AC balancing
is really the important part (hint: what does the input current of a
rectifier/capacitor filter look like).

OTOH I normally provide a discharge path of some description, usually
R's and LEDs to provide a "look out" indicator. To tell the truth its
more like a "for fucks sake dont stick a scope probe here, its live"
indicator

Cheers
Terry

 

[Terry Given]

Glenn Baddeley wrote:

Two 2200uF 450V caps in series, paralleled 18 times (36 caps).


Wow, that's a pretty big cap bank.

for a 660A 400Vac motor controller. One of the exploding cap banks was
right beside me - thta was one of three paralleled 660A drives (a 1MW
motor controller).

AC motor controller design is all about capacitor lifetime - caps are
expensive and bulky, and the wear-out mechanism effectively sets the
operational life of the drive.

The 660A drive range went from 100A - 1750A. When we designed the
replacement 140A product, we moved the electrolytics out of the heatsink
hot air path. This reduced cap temeprature enough to drop one of the 4
pairs of caps, a significant saving ($50 each).

Its easy to measure cap core temperature too - just have Hitachi make
you a few dozen caps with inbuilt thermocouples. Seriously, peel back
the top of the can (where the score lines are to control rupturing) and
bung a thermocouple in there.

That particular little Methyl Bromide problem killed well over
$1,000,000 worth of product, and required some 1,700 caps to be swapped
out in-situ, over a 5-day period. oops.

Cap leakage current specs also showed the
27k resistors could not be relied upon to do any real DC balancing
either, as the leakage spec was quite high - IIRC we needed to throw
away many tens of W to get proper DC balancing.


... apparently a lot higher than the 10uA example quoted by Phil.

IMO never use a part whos specs you cannot obtain. All caps are most
certainly *not* created equal.

that being said, mass manufacture is a very different beast from low
volume manufacture, where its perfectly feasible to select-on-test,
match parts etc.

If a datasheet gives a min & max range, its often a good idea to allow
for it. Sometimes the ranges are due to tester limitations (eg opamp
input bias current), other times due to process/batch variations.

These particular caps had a +/- 1mm spec on the terminal hole locations.
And it occurred in practice, too.

Some HAM transmitters do throw away 10's watts to getter better
load / no load regulation from a 1 - 2 KVA tranny. It concerns
me that all that heat is generated near the caps.

that should be a concern. Look up a decent datasheet for a WW resistor -
quite often these can have a 200 - 300C temperature rise at full power.

That sounds like a rather poor approach. I'd rather save the watts, and
spend the money on a regulator. that being said, if the power
consumption and temperature can be dealt with, then why not.

And the AC balancing
is really the important part (hint: what does the input current of a
rectifier/capacitor filter look like).


A good point, I've never seen this aspect mentioned anywhere before.

Cheers,
Glenn.

Neither had I, but its pretty obvious really.

Ultimately it doesnt really matter - cap leakage can pretty much be
ignored and balancing resistors are pointless. If you make a
series/parallel array, tie all the center points together. That way the
(hopefully random) variations tend to cancel each other out. And leave
yourself plenty of voltage headroom on the caps - Phils 30% spare is a
good choice.

But the OP wanted to know how to calculate the balancing resistors.....

Cheers
Terry

 

[Uncle-Fester]

Terry Given wrote:

That particular little Methyl Bromide problem killed well over
$1,000,000 worth of product, and required some 1,700 caps to be swapped
out in-situ, over a 5-day period. oops.

I assume thats a typo ? ($1,000,000)



Cheers
Terry

 

[Terry Given]

Uncle-Fester wrote:

Terry Given wrote:

That particular little Methyl Bromide problem killed well over
$1,000,000 worth of product, and required some 1,700 caps to be
swapped out in-situ, over a 5-day period. oops.

I assume thats a typo ? ($1,000,000)

nope. a megabuck worth of gear. Of course once we figured out what the
problem was and why, the actual cost to repair was much lower, but the
total cost was still pretty high - around $100k.

thats tiny compared to the potential cost of the problem. It was a
$3,000,000 order, and the customer (rightly) demanded the problem be
solved (and proven so) or the equipment removed and replaced (with a
competitors) at our expense. The follow-on effects could have been
horrendous.

But a careful analysis of all (as in the last 30 years) product
failures, along with dissection of every item of the order, showed no
correlations at all - size, loading, environment, date manufactured,
assembler, pcb batch numbers etc. The conclusion drawn from this
analysis was that something happened to one third of this order. So we
got the shipping records, and WHAM - every item within one container was
faulty, all other items were fine. Five minutes later Methyl Bromide
fumigation was identified as the culprit and the problem was solved to
the customers satisfaction (although the rework continued for several
more days).

electron microscopy was used by the customer to examine the failed caps.
Surprisingly the metallurgists who looked at them didnt identify
corrosion or its byproducts, which would have helped.

despite having sent senior engineers to placate the customer (who was
spending half a billion dollars with them on another job), several days
later the cap manufacturer admitted another customer had the exact same
problem several years earlier. They had forgotten about it.

Cheers
Terry

 

[The Real Andy]

On Wed, 22 Jun 2005 21:47:41 +1000, "Phil Allison"
<philallison@tpg.com.au> wrote:

"Rheilly Phoull"
G'day All
What are the rules or method to calculate the values of balancing
resistors
when using electros in series ??


** Forget it - just use caps that have a large margin in excess of the
needed voltage.

Eg - two 350 volt types applied to a 500 volt supply.

The caps will very soon reach a mutual, acceptable agreement on what
precise voltage suits their individual taste !!

in this case a resistor is good for a bleeder thats about all i

ever use them for

 

[Phil Allison]

"The Real Andy"
Phil Allison

** Forget it - just use caps that have a large margin in excess of the
needed voltage.

Eg - two 350 volt types applied to a 500 volt supply.

The caps will very soon reach a mutual, acceptable agreement on what
precise voltage suits their individual taste !!


in this case a resistor is good for a bleeder thats about all i
ever use them for

** What type do you use that can stand 250 volt DC indefinitely ??

High value resistors, of all types, seem to have a very high failure (
failing open) rate after a few years when subjected to continuous DC above
about 100 volts.




........... Phil

 

[Terry Given]

Phil Allison wrote:

"The Real Andy"
Phil Allison

** Forget it - just use caps that have a large margin in excess of the
needed voltage.

Eg - two 350 volt types applied to a 500 volt supply.

The caps will very soon reach a mutual, acceptable agreement on what
precise voltage suits their individual taste !!


in this case a resistor is good for a bleeder thats about all i
ever use them for



** What type do you use that can stand 250 volt DC indefinitely ??

High value resistors, of all types, seem to have a very high failure (
failing open) rate after a few years when subjected to continuous DC above
about 100 volts.


.......... Phil

any ideas as to the mechanism? come across any decent app notes?

I presume that as high value Rs the peak power dissipation would still
be pretty low - I've seen lots of low-value resistors fail after being
abused in this way, eg 0603 10R 12V gatedrive resistor.

what about electromigration? wouldnt that tend to bridge out the
serpentine cuts, thus reducing R

what about VR37s, they are designed for very high voltage pulses. we
used pairs of 470k VR37 resistors in series for DC bus and line voltage
sensing, +/-300V - 500V about earth so 150V-250V per resistor. never had
a single failure over perhaps 10 years and 40,000 products.

We also used 100k PR02's across each half of the DC bus cap bank, so
about 0.9W continuous at 600V/2 and 1.6W at 800V/2(short term during
regen). dont know about failures there though, as power electronics
failures tend to trash the entire area, so its quite possible. the
dissipation is a bit high at 800V, but luckily the nearby electrolytics
will provide some cooling

At one stage I tried damping an EMI filter with 1R PR02s in series with
the 1uF caps. If turned on at an appropriately high line voltage (which
one phase always is in 3-phase systems) the PR02 emitted a bright flash
of light and went open circuit. probably because (400V*1.4)^2/1R = 320kW
peak pulse power. We found that by observing the flash...wtf?! A carbon
composition resistor ate the peak pulse though.

Cheers
Terry

 

[Phil Allison]

"Terry Given"

Phil Allison wrote:
"The Real Andy"
Phil Allison

** Forget it - just use caps that have a large margin in excess of
the
needed voltage.

Eg - two 350 volt types applied to a 500 volt supply.

The caps will very soon reach a mutual, acceptable agreement on what
precise voltage suits their individual taste !!


in this case a resistor is good for a bleeder thats about all i
ever use them for



** What type do you use that can stand 250 volt DC indefinitely ??

High value resistors, of all types, seem to have a very high failure (
failing open) rate after a few years when subjected to continuous DC
above about 100 volts.


any ideas as to the mechanism? come across any decent app notes?

** The resistors that fail are operating within published specs.

Ergo - the "mechanism" is bad manufacture.

Not much mention of that in any "app notes'.

what about electromigration?

** What - without the right Visa ??

what about VR37s, they are designed for very high voltage pulses.

** Values start at 1M ohm - useless.





............ Phil

 

[Terry Given]

Phil Allison wrote:

"Terry Given"

Phil Allison wrote:

"The Real Andy"
Phil Allison


** Forget it - just use caps that have a large margin in excess of
the
needed voltage.

Eg - two 350 volt types applied to a 500 volt supply.

The caps will very soon reach a mutual, acceptable agreement on what
precise voltage suits their individual taste !!


in this case a resistor is good for a bleeder thats about all i
ever use them for



** What type do you use that can stand 250 volt DC indefinitely ??

High value resistors, of all types, seem to have a very high failure (
failing open) rate after a few years when subjected to continuous DC
above about 100 volts.


any ideas as to the mechanism? come across any decent app notes?



** The resistors that fail are operating within published specs.

Ergo - the "mechanism" is bad manufacture.

Not much mention of that in any "app notes'.

I too suspect this is largely the reason.

I once had a batch of 0.1% resistors from vishay that were off by up to
25%. it was a ups DC balancing circuit, and it didnt like it at all.
Only 20-30 units were affected.

vishay were great, wrote me a nice report. basically contaminated
material wrecked part of a batch, but the effect was triggered by
high(ish) temperatures so they all measured ok at the factory, and
through our genrad pcb tester, but once the unit heated up to 50C or so,
they drifted like crazy.

in general my experience is that nice cheap components are quite often
unreliable shit, and the only way to tell is to wait a year or two. e.g.
every company I've ever worked for has had a cheap ceramic bypass
capacitor horror story to tell.

what about electromigration?



** What - without the right Visa ??

lol. other than its existence, I dont know much about electromigration,
but it appears these guys do:

http://www.theo-phys.uni-essen.de/tp/forsch/krug.html

whaddya know, its kind of analogous to crack propagation, so certainly
could cause open circuits. AIUI its an issue at IC level because the
small dimensions result in high E field strengths. The same would occur
in a serpentine resistor at high voltage.

what about VR37s, they are designed for very high voltage pulses.



** Values start at 1M ohm - useless.

I have 100pcs 470k sitting in front of me.

1998 PA08B databook:

VR25 100k - 15M 1,5,10% E12/E24 1,600Vdc peak
VR37 100k - 33M 1,5% E24/E96 3,500Vdc peak
VR68 100k - 68M 1,5% E24/E96 10,000Vdc peak
PR02 0R33 - 1M 1,5% E24/E96 500Vdc peak

........... Phil

Cheers
Terry

 

[Phil Allison]

"Terry Given"
Phil Allison wrote:

what about VR37s, they are designed for very high voltage pulses.



** Values start at 1M ohm - useless.

I have 100pcs 470k sitting in front of me.

** My sources

1. Farnell in One cat.

2. Philips General cat 1978.

1998 PA08B databook:

** I believe you - but that is not the simple answer.


The simple one is to use a few cheap 0.5 or 0.75 watt MF resistors in
series - so each cops less than 100 volts and runs cool.




............ Phil

 

[Terry Given]

Phil Allison wrote:

"Terry Given"
Phil Allison wrote:



what about VR37s, they are designed for very high voltage pulses.



** Values start at 1M ohm - useless.

I have 100pcs 470k sitting in front of me.




** My sources

1. Farnell in One cat.

2. Philips General cat 1978.

I "inherited" a full set of philips 2000 databooks (many are 1998). I
regret throwing out so many databooks in the past

1998 PA08B databook:



** I believe you - but that is not the simple answer.


The simple one is to use a few cheap 0.5 or 0.75 watt MF resistors in
series - so each cops less than 100 volts and runs cool.

........... Phil

absolutely, and with smt its trivial to put down as many as required
(although per-part power dissipation also drops). The overall footprint
is not dissimilar, but the inductance is a whole lot lower, as is the
capacitance. designing to cope with single component failures is a lot
easier (eg 6 aint much more than 5, but 2 is a lot more than 1) too.

Cheers
Terry

 

[David L. Jones]

Terry Given wrote:

Phil Allison wrote:
"Terry Given"
Phil Allison wrote:



what about VR37s, they are designed for very high voltage pulses.



** Values start at 1M ohm - useless.

I have 100pcs 470k sitting in front of me.




** My sources

1. Farnell in One cat.

2. Philips General cat 1978.


I "inherited" a full set of philips 2000 databooks (many are 1998). I
regret throwing out so many databooks in the past




1998 PA08B databook:



** I believe you - but that is not the simple answer.


The simple one is to use a few cheap 0.5 or 0.75 watt MF resistors in
series - so each cops less than 100 volts and runs cool.

........... Phil

absolutely, and with smt its trivial to put down as many as required
(although per-part power dissipation also drops). The overall footprint
is not dissimilar, but the inductance is a whole lot lower, as is the
capacitance. designing to cope with single component failures is a lot
easier (eg 6 aint much more than 5, but 2 is a lot more than 1) too.

Did the same thing with a 500V DC distributed supply system. Was much
easier to simply use a few SMD resistors in series as the ballast
resistors, rather than having to source high voltage ones. Too bad
there wasn't the same choice on the caps, volume realestate was at a
premium (more so than PCB realestate) so we had to run tight margins
and had little selection available when it came to the cap vs voltage
vs volume vs height tradeoff.

Dave

 

[Uncle-Fester]

Phil Allison wrote:

** Fuck off -




.......... Phil

Make you insipid weak little toaster boy .