Al electro's vs polymer?

[bitrex]

On 8/27/19 11:27 PM, Sjouke Burry wrote:

On 28.08.19 4:53, Phil Allison wrote:
whitless the 3rd wrote:

------------------------


Why not just stick with aluminum electrolytic?
The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing
with
age, just use two or three.


**  Huh ?

  Doing that makes them last longer ??

  On which planet?


On this one. Ripple current is spread across the 3 caps,
resulting in lower temperature and longer life.

I think using three caps can be evidence of Muntzing-design, not
necessarily reliability-design. Use three cheapo 47uF caps in parallel
where you need 150uF that also handily combine correctly to provide the
ESR the control loop needs.

now you also have three points of failure instead of one (costly) 150u
with a better ESR and temp spec

 

[bitrex]

On 8/27/19 11:57 PM, Phil Allison wrote:

Sjouke Burry wrote:

Phil Allison wrote:

whitless the 3rd wrote:

------------------------


Why not just stick with aluminum electrolytic?
The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing with
age, just use two or three.


** Huh ?

Doing that makes them last longer ??

On which planet?


On this one.


** Well, that begs the question ....


Ripple current is spread across the 3 caps,

** It's too small to matter.

resulting in lower temperature and longer life.


** Nope.

FYI: EVERY case of early electro failure I have seen, the cause was high ambient temp, not ripple current.

Dopey PCB designers placing them too close to hot power semis and resistors.



.... Phil

a number of the mobo cap failures on my on PC equipment I've seen were
definitely due to heat. machine runs constantly, air intake starts
getting dust clogged, temperature goes up, cap temperature goes up,
regulation gets worse, the buck switches start dissipating more heat
into the caps that are right next to them, etc.

On cheap mobos it's a toss up as to whether a capacitor pops or a buck
MOSFET pops first. thermal epoxying some little copper heat sinks to the
buck transistors and some aluminum fins on the cap bank helps a lot. The
more costly motherboards do this from the factory

 

[bitrex]

On 8/27/19 11:27 PM, Sjouke Burry wrote:

On 28.08.19 4:53, Phil Allison wrote:
whitless the 3rd wrote:

------------------------


Why not just stick with aluminum electrolytic?
The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing
with
age, just use two or three.


**  Huh ?

  Doing that makes them last longer ??

  On which planet?


On this one. Ripple current is spread across the 3 caps,
resulting in lower temperature and longer life.

Meh, in consumer equipment they use three 100u where they need a 330,
they all run hot, one starts going much higher ESR sooner rather than
later and the supply dies.

 

[bitrex]

On 8/28/19 1:58 AM, whit3rd wrote:

On Tuesday, August 27, 2019 at 7:53:55 PM UTC-7, Phil Allison wrote:
whitless the 3rd wrote:

The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing with
age, just use two or three.


** Huh ?

Doing that makes them last longer ??

On which planet?

If you need 500 uF, just use three 330 uF units. The usual failure mode
being electrolyte loss/drying out, would leave the failed unit an open circuit,
so the others cover for it.

I've done some troubleshooting on old machines (Mac IIci, if anyone recalls those)
where seal failure could take out about a third of the distributed filter caps
without anyone noticing (unless the electrolyte goo ate through a trace).

Don't think they had any choice in the IIci era but to distribute the
caps because, in the vernacular, even the plain-Jane Al electros
available at the time sucked balls compared to what's available now.

 

[bitrex]

On 8/28/19 1:58 AM, whit3rd wrote:

On Tuesday, August 27, 2019 at 7:53:55 PM UTC-7, Phil Allison wrote:
whitless the 3rd wrote:

The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing with
age, just use two or three.


** Huh ?

Doing that makes them last longer ??

On which planet?

If you need 500 uF, just use three 330 uF units. The usual failure mode
being electrolyte loss/drying out, would leave the failed unit an open circuit,
so the others cover for it.

You can do that in your own projects or if you work for the
defense/military contractor (or those Xtreme Fat1tly $1000 PC PSUs for
gamers) but nobody is ever gonna use three 330uF caps for a 500 in a
general consumer price-point oriented product, though.

I've done some troubleshooting on old machines (Mac IIci, if anyone recalls those)
where seal failure could take out about a third of the distributed filter caps
without anyone noticing (unless the electrolyte goo ate through a trace).

 

[whit3rd]

On Wednesday, August 28, 2019 at 10:30:44 AM UTC-7, bitrex wrote:

Meh, in consumer equipment they use three 100u where they need a 330,
they all run hot, one starts going much higher ESR sooner rather than
later and the supply dies.

Tolerances on capacitors can be +50, -20 %; if you 'need' 330u,
three 100u won't do it. Having tolerance if one fails high-ESR
would have been a good bet during the capacitor plague (which
cost warranty centers many hundreds of megabucks).

Component failing isn't the bad thing; functional failure of the system is.
Avoid it with component quality (uncertain: refer to the capacitor plague),
or with redundancy and overdesign, or... run into functioinal failure.

 

[bitrex]

On 8/28/19 3:48 PM, whit3rd wrote:

On Wednesday, August 28, 2019 at 10:30:44 AM UTC-7, bitrex wrote:

Meh, in consumer equipment they use three 100u where they need a 330,
they all run hot, one starts going much higher ESR sooner rather than
later and the supply dies.

Tolerances on capacitors can be +50, -20 %; if you 'need' 330u,
three 100u won't do it. Having tolerance if one fails high-ESR
would have been a good bet during the capacitor plague (which
cost warranty centers many hundreds of megabucks).

+50% vs - 20% and the maybe the distribution skews to high and you get
three spins on the wheel? sounds like an excellent opportunity to save
20 cent on each unit with a 99% confidence each one end up with just
about what it needs, perhaps

Component failing isn't the bad thing; functional failure of the system is.
Avoid it with component quality (uncertain: refer to the capacitor plague),
or with redundancy and overdesign, or... run into functioinal failure.

 

bitrex <user@example.net> wrote in news:Ipy9F.223047$on8.146436
@fx46.iad:

The
more costly motherboards do this from the factory

The PS feeding the MOBO matters too. Those local supplies work
better if the DC feeding them is robust and clean.

So if you went and bought a POS cheap PC power supply, you are part
of the problem.

Especially if you also run a good high end vid card, and that means
a good PS is 100% required.

I say this because my 12 core server runs fine at all CPU tax
situations. I see no heat and even after high end benchmarks get
run, much less doing fractals and 3D renderings, etc.

So if you are one of those 'just enough to make it work' power
supply guys, you cause many of your own problems.

 

[whit3rd]

On Thursday, August 29, 2019 at 4:09:06 AM UTC-7, DecadentLinux...@decadence.org wrote:

The PS feeding the MOBO matters too. Those local supplies work
better if the DC feeding them is robust and clean.

So if you went and bought a POS cheap PC power supply, you are part
of the problem.

Yeah, but engineering doesn't require half the box to do all the
job. Cray-1 power supplies were brute-force unregulated DC; the differential
logic didn;t modulate the power any more than a resistor load would have.

 

On Wednesday, 28 August 2019 18:17:45 UTC+1, bitrex wrote:

On 8/28/19 1:58 AM, whit3rd wrote:
On Tuesday, August 27, 2019 at 7:53:55 PM UTC-7, Phil Allison wrote:
whitless the 3rd wrote:

The regular old not-so-low-ESR
caps are fine for lots of purposes, and if you don't like 'em failing with
age, just use two or three.


** Huh ?

Doing that makes them last longer ??

On which planet?

If you need 500 uF, just use three 330 uF units. The usual failure mode
being electrolyte loss/drying out, would leave the failed unit an open circuit,
so the others cover for it.

You can do that in your own projects or if you work for the
defense/military contractor (or those Xtreme Fat1tly $1000 PC PSUs for
gamers) but nobody is ever gonna use three 330uF caps for a 500 in a
general consumer price-point oriented product, though.

This pc uses a bank of 1000uF caps in the onboard volage converter. I know cos eventually I replaced em all. It's not unusual for PCs to do that. It didn't fail to work until almost all that capacity was gone, so it's just done for reliability.


NT

I've done some troubleshooting on old machines (Mac IIci, if anyone recalls those)
where seal failure could take out about a third of the distributed filter caps
without anyone noticing (unless the electrolyte goo ate through a trace).

 

tabbypurr@gmail.com wrote in
news:03d7816e-2e54-406f-8581-e92aee92715f@googlegroups.com:

This pc uses a bank of 1000uF caps in the onboard volage
converter. I know cos eventually I replaced em all. It's not
unusual for PCs to do that. It didn't fail to work until almost
all that capacity was gone, so it's just done for reliability.

You sound like SkyTard Farting does. Except he is way too dumb to
actually repair the stuff he breaks.

You replace a lot of EL caps in your crap? What do you buy old
used gear?

Do you bake your stuff in an oven so often you have to replace the
caps?

 

whit3rd <whit3rd@gmail.com> wrote in
news:2aa08070-b595-495c-9c18-3a655e9cefb7@googlegroups.com:

On Thursday, August 29, 2019 at 4:09:06 AM UTC-7,
DecadentLinux...@decadence.org wrote:

The PS feeding the MOBO matters too. Those local supplies work
better if the DC feeding them is robust and clean.

So if you went and bought a POS cheap PC power supply, you are
part
of the problem.

Yeah, but engineering doesn't require half the box to do all the
job. Cray-1 power supplies were brute-force unregulated DC; the
differential logic didn;t modulate the power any more than a
resistor load would have.

Their systems were not powered the way these motherboards are with
immeiately proximal dc-dc conversion being part of the heat product.

They operate better and more efficiently with a clean DC in feed.
So less heat would be on those parts doing their jobs just a slight
bit easier.

DC to DC supplies are unhappey when the DC IN component is full of
PARD, ripple, lulls in regulation, etc.

 

On Friday, 30 August 2019 16:10:20 UTC+1, DecadentLinux...@decadence.org wrote:

tabbypurr wrote in
news:03d7816e-2e54-406f-8581-e92aee92715f@googlegroups.com:

This pc uses a bank of 1000uF caps in the onboard volage
converter. I know cos eventually I replaced em all. It's not
unusual for PCs to do that. It didn't fail to work until almost
all that capacity was gone, so it's just done for reliability.



You sound like SkyTard Farting does. Except he is way too dumb to
actually repair the stuff he breaks.

You replace a lot of EL caps in your crap? What do you buy old
used gear?

Do you bake your stuff in an oven so often you have to replace the
caps?

If you had half a brain you could answer those questions yourself.
I've had this thing running for years, it was a top of the line system & still doesn't need replacing. It's still quicker than many new machines. It's done well.


NT