开云体育

The MANUFACTURER'S DATASHEETS I read specifically call out a small (10 ohm for example) series resistor to mitigate the tantalum's known inability to handle surges. Tek followed these guidelines.

The OTHER problems tantalums have with a short lifetime and a catastrophic failure mechanism, that we now know so well, probably did not become fully understood until years later after it was too late. That was not something Tek would become aware of until it was too late.

When tantalums first became available they had some great advantages. From an engineering perspective their size makes it possible to squeeze more functionality into a smaller space. Their cost makes them economically attractive and lowers instrument costs. What would you have done when presented with this new technology?

Dennis

On Wed, 10 Oct 2012, Dennis Tillman wrote:

I've just finished "shotgun approach" rebuild of older TG501 -- almost all
74xx replaced with 74HCT/74LS, lots of resistors removed, some added (it's
been done in old days when it was considered OK to leave TTL inputs floating
for logical 1 so some pullups were added.) Then it came to the capacitors.
The thing has a lot of dipped tantalums (no other electrolytics,
everything's tantalum.) I started pulling them one by one and testing with
my HP 4274A. The result was stunning -- _ALL_ of them were as good as new
and even better. If you don't know almost all of modern capacitors with e.g.
+/-20% tolerance come out very close to the -20% margin. Yes, they are
within specs but they are ALL at the lower end. It might be there are some
manufacturers that don't squeeze the last penny but everything I tested is
that way. So if you have a 100uF capacitor of recent manufacturing you can
be pretty sure you'll get something like 82uF. Or 88uF if you are lucky. It
is NEVER 100uF not even mentioning higher than that.

ALL tantalums in that TG501 (40 years old) were at least their stated
capacitance and most of them were higher. There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.) 330uF 6V 40year old tantalum had 360uF capacitance with less
than 100mOhms ESR. Only polymer aluminum types can beat that and you can be
pretty sure none of 330uF ones will have 360uF capacitance.

And those old tantalums never leak or dry out. If they survived that many
years you can be sure they will last forever and there is absolutely no
reason for replacing them. There is almost nothing better than those these
days and for their longevity nothing comes even close to them. Ceramics may
compete on longevity but it is not that easy to replace a 330uF capacitor
with a ceramic one. And there is absolutely no advantages it would've got.

Yes, tantalums do fail. But that failure rate is orders of magnitude lower
than aluminum electrolytics that leak, dry out, loose their capacitance etc.
Yes, aluminums fail in different way -- they just become mockups -- and
tantalums fail with a short but, IMHO, it is better because you do replace
them as they fail unlike aluminums that are almost dead but instrument still
works somehow...

The MANUFACTURER'S DATASHEETS I read specifically call out a small (10 ohm
for example) series resistor to mitigate the tantalum's known inability to
handle surges. Tek followed these guidelines.

The OTHER problems tantalums have with a short lifetime and a catastrophic
failure mechanism, that we now know so well, probably did not become fully
understood until years later after it was too late. That was not
something Tek would become aware of until it was too late.

When tantalums first became available they had some great advantages. From
an engineering perspective their size makes it possible to squeeze more
functionality into a smaller space. Their cost makes them economically
attractive and lowers instrument costs. What would you have done when
presented with this new technology?

Dennis

-----Original Message-----
From: Chuck Harris [mailto:cfharris@...]
Sent: Wednesday, October 10, 2012 5:20 AM

I suspect that tek engineers didn't put the resistors and inductors before the tantalum capacitors to protect from surges, but rather to act as an input resistor to a simple RC filter.... the network was there because they wanted to clean up the power going to subassemblies.
The fact that the small resistors and inductors acted as surge current limiters, and fuses was just a happy side benefit.

Tek engineers used the tantalums just like everyone else did during that era, as small sized large value storage capacitors that had good HF characteristics. It wasn't until much later that the capacitor manufacturers started seeing surge related failures, and began to provide warnings about limiting surge currents. They never told you to limit working voltage to 1/2 the capacitor's rated voltage... that is something that NASA learned as it tried to qualify tantalums for space use. NASA learned about the voltage issues because one of their standard ways of qualifying parts was to run them at 2x their rated voltage, which in the case of a tantalum was a disaster. Even if the part survived the testing, it was damaged and set to fail under normal operation.

-Chuck Harris

Miroslav Pokorni wrote:

Hello Dennis,
You said: "Tek engineers know this and design for it."

If that is so, how come that there are massive failures of tantalums
in 465, 7854 and other scopes.

On 10/9/2012 6:49 AM, Dennis Tillman wrote:

Tek engineers
know this and design for it.

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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*

On Wed, Oct 10, 2012 at 6:59 PM, Sergey Kubushyn <ksi@...> wrote:

There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.)

At what frequency? The ceramic should have well below 10mR at 100kHz
according to specification, and I just don't see how it could be
otherwise considering the construction.

Regarding the generally low capacitance value on electrolytics, I have
extensive test data for Rubycon and Nippon Chemicon at work and will
check if they are consistently at the lower capacitance tolerance.

ST

On Wed, 10 Oct 2012, Stefan Trethan wrote:

On Wed, Oct 10, 2012 at 6:59 PM, Sergey Kubushyn <ksi@...> wrote:

There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.)

At what frequency? The ceramic should have well below 10mR at 100kHz
according to specification, and I just don't see how it could be
otherwise considering the construction.

1KHz. This is actual measured values. As for the specifications majority of
manufacturers don't give any values on ESR in their datasheets. It is
"common knowledge" they are "very low ESR" but reality check shows totally
different picture. If you look at e.g. NIC NMC High CV Series datasheet (one
of the few who give data on ESR) you'll see a ESR vs Frequency chart. It
clearly shows that MLCCs are better when going into MHz range but they are
nothing to write home about at lower frequencies.

Regarding the generally low capacitance value on electrolytics, I have
extensive test data for Rubycon and Nippon Chemicon at work and will
check if they are consistently at the lower capacitance tolerance.

Once again, I do NOT have very extensive data, just several hundred
capacitors. It might be not a representative sample but I could not find a
SINGLE one of recent manufacturing that was 100% of stated capacity or
higher. On the old ones I could not find more than couple of pieces that
were under 100%.

ST


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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*

At 1kHz that's just about spot on specification for the taiyo yuden part.
The impedance at this frequency will be what, 15, 16 Ohms? Not sure
why you'd worry....

At what frequency did you measure capacitance? Just so that we don't
waste both our time again.

ST

On Wed, 10 Oct 2012, Stefan Trethan wrote:

At 1kHz that's just about spot on specification for the taiyo yuden part.
The impedance at this frequency will be what, 15, 16 Ohms? Not sure
why you'd worry....

I do not worry. It is not about using it at a particular frequency.

At what frequency did you measure capacitance? Just so that we don't
waste both our time again.

1KHz. And I don't have time to waste too.

ST

On Wed, Oct 10, 2012 at 9:22 PM, Sergey Kubushyn <ksi@...> wrote:

On Wed, 10 Oct 2012, Stefan Trethan wrote:

On Wed, Oct 10, 2012 at 6:59 PM, Sergey Kubushyn <ksi@...> wrote:

There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.)

At what frequency? The ceramic should have well below 10mR at 100kHz
according to specification, and I just don't see how it could be
otherwise considering the construction.

1KHz. This is actual measured values. As for the specifications majority of
manufacturers don't give any values on ESR in their datasheets. It is
"common knowledge" they are "very low ESR" but reality check shows totally
different picture. If you look at e.g. NIC NMC High CV Series datasheet (one
of the few who give data on ESR) you'll see a ESR vs Frequency chart. It
clearly shows that MLCCs are better when going into MHz range but they are
nothing to write home about at lower frequencies.

Regarding the generally low capacitance value on electrolytics, I have
extensive test data for Rubycon and Nippon Chemicon at work and will
check if they are consistently at the lower capacitance tolerance.

Once again, I do NOT have very extensive data, just several hundred
capacitors. It might be not a representative sample but I could not find a
SINGLE one of recent manufacturing that was 100% of stated capacity or
higher. On the old ones I could not find more than couple of pieces that
were under 100%.

ST


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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*


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

Yahoo! Groups Links

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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*

That explains that then, they measure a few percent lower at 1kHz.
It's unfair to accuse a manufacturer of penny pinching if you don't
meet their specified testing conditions (usually 120Hz).

Saving a few percent of foil is probably not worth the hassle of
having to meet tighter tolerances caused by leaning towards the lower
end. More likely they used to go above nominal with the old caps
because tolerances were asymmetrical back then.
I'll check out the test data to see if the distribution is offset when
I get the chance.

ST

On Wed, 10 Oct 2012, Stefan Trethan wrote:

That explains that then, they measure a few percent lower at 1kHz.
It's unfair to accuse a manufacturer of penny pinching if you don't
meet their specified testing conditions (usually 120Hz).

It gives same readings at 120Hz for capacitance if it can measure it
(>100uF). For 10-100uF range the most accurate results are at 400 Hz. 1KHz
is a convenient test frequency giving wide range with relatively good
accuracy. It might give up to 3% error at the boundary F/C ranges where it
barely measures capacitance (too high F for big C or too low F for small C)
but it is never 20%. If you want to get most of it, sure you have to choose
a proper frequency. But for 3 digits 1KHz is good for measuring almost
anything. Also capacitance itself does not vary that much with frequency.
ESR varies orders of magnitude more.

Saving a few percent of foil is probably not worth the hassle of
having to meet tighter tolerances caused by leaning towards the lower
end. More likely they used to go above nominal with the old caps

It does. And almost all big manufacturers has shown tendency towards saving
a penny risking a buck. It doesn't make sense in a long run but who thinks
farther than a quarter? Does it make sense to lay off the best (those with
highest salaries) to make a nice profit figure? Can you name a company that
did not fell to a temptation of using cheaper chinese capacitors in high
frequency high ripple applications and thus avoided massive failures because
of blown capacitors? Remember, profit of saving a penny is now and this is
when the higher management gets their bonuses for making profit. Failures
are later and they are not punished for those. And no one of them is going
to work for a particular company forever -- they will go to a different one
for making that one more profitable too once they are done with their
current one (i.e. once it is destroyed by making it more profitable.)

because tolerances were asymmetrical back then.

Nope. Old ones usually fall in the middle of their tolerance range. If it
was +80/-20% the majority of 1,000uF capacitors had 1,300uF capacitance. And
even +/-20% ones rarely went under 100%. It was also understood that it is
better to overshoot than undershoot as far as capacitance goes.

I'll check out the test data to see if the distribution is offset when
I get the chance.

ST

On Wed, Oct 10, 2012 at 10:23 PM, Sergey Kubushyn <ksi@...> wrote:

On Wed, 10 Oct 2012, Stefan Trethan wrote:

At 1kHz that's just about spot on specification for the taiyo yuden part.
The impedance at this frequency will be what, 15, 16 Ohms? Not sure
why you'd worry....

I do not worry. It is not about using it at a particular frequency.

At what frequency did you measure capacitance? Just so that we don't
waste both our time again.

1KHz. And I don't have time to waste too.

ST

On Wed, Oct 10, 2012 at 9:22 PM, Sergey Kubushyn <ksi@...> wrote:

On Wed, 10 Oct 2012, Stefan Trethan wrote:

On Wed, Oct 10, 2012 at 6:59 PM, Sergey Kubushyn <ksi@...> wrote:

There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.)

At what frequency? The ceramic should have well below 10mR at 100kHz
according to specification, and I just don't see how it could be
otherwise considering the construction.

1KHz. This is actual measured values. As for the specifications majority of
manufacturers don't give any values on ESR in their datasheets. It is
"common knowledge" they are "very low ESR" but reality check shows totally
different picture. If you look at e.g. NIC NMC High CV Series datasheet (one
of the few who give data on ESR) you'll see a ESR vs Frequency chart. It
clearly shows that MLCCs are better when going into MHz range but they are
nothing to write home about at lower frequencies.

Regarding the generally low capacitance value on electrolytics, I have
extensive test data for Rubycon and Nippon Chemicon at work and will
check if they are consistently at the lower capacitance tolerance.

Once again, I do NOT have very extensive data, just several hundred
capacitors. It might be not a representative sample but I could not find a
SINGLE one of recent manufacturing that was 100% of stated capacity or
higher. On the old ones I could not find more than couple of pieces that
were under 100%.

ST


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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*


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

Yahoo! Groups Links

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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*


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

Yahoo! Groups Links

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

Yahoo! Groups Links

---
*
* KSI@home KOI8 Net < > The impossible we do immediately. *
* Las Vegas NV, USA < > Miracles require 24-hour notice. *
*

I have found the same thing with old solid tantalums. Up until the
time they short out, and some of the ones I have tested had already
had other units of the same type fail, they test good for capacitance,
D, and leakage. I have yet to find a marginal one.

For electrolytics I usually find two different failure modes. The
ones operating at high ripple current will fail completely open one at
a time so I usually change them all at once. The others will be some
fraction of their original capacitance. In that case, usually all of
the others of the same type are similarly worn out.

On Wed, 10 Oct 2012 09:59:21 -0700 (PDT), Sergey Kubushyn
<ksi@...> wrote:

On Wed, 10 Oct 2012, Dennis Tillman wrote:

I've just finished "shotgun approach" rebuild of older TG501 -- almost all
74xx replaced with 74HCT/74LS, lots of resistors removed, some added (it's
been done in old days when it was considered OK to leave TTL inputs floating
for logical 1 so some pullups were added.) Then it came to the capacitors.
The thing has a lot of dipped tantalums (no other electrolytics,
everything's tantalum.) I started pulling them one by one and testing with
my HP 4274A. The result was stunning -- _ALL_ of them were as good as new
and even better. If you don't know almost all of modern capacitors with e.g.
+/-20% tolerance come out very close to the -20% margin. Yes, they are
within specs but they are ALL at the lower end. It might be there are some
manufacturers that don't squeeze the last penny but everything I tested is
that way. So if you have a 100uF capacitor of recent manufacturing you can
be pretty sure you'll get something like 82uF. Or 88uF if you are lucky. It
is NEVER 100uF not even mentioning higher than that.

ALL tantalums in that TG501 (40 years old) were at least their stated
capacitance and most of them were higher. There were bunch of 10uF 25V ones
that I was going to replace with 1206 SMD X5R ceramics but to my surprise
those dipped tantalums had something like 300mOhm ESR that is actually
BETTER than ceramics (Taiyo Yuden 10uF 25V 1206 X5R ceramics of recent
manufacturing that I had several reels in my parts bins have something like
350mOhms ESR.) 330uF 6V 40year old tantalum had 360uF capacitance with less
than 100mOhms ESR. Only polymer aluminum types can beat that and you can be
pretty sure none of 330uF ones will have 360uF capacitance.

And those old tantalums never leak or dry out. If they survived that many
years you can be sure they will last forever and there is absolutely no
reason for replacing them. There is almost nothing better than those these
days and for their longevity nothing comes even close to them. Ceramics may
compete on longevity but it is not that easy to replace a 330uF capacitor
with a ceramic one. And there is absolutely no advantages it would've got.

Yes, tantalums do fail. But that failure rate is orders of magnitude lower
than aluminum electrolytics that leak, dry out, loose their capacitance etc.
Yes, aluminums fail in different way -- they just become mockups -- and
tantalums fail with a short but, IMHO, it is better because you do replace
them as they fail unlike aluminums that are almost dead but instrument still
works somehow...

There is no lean towards the lower tolerance for Rubycon or Nippon
Chemicon (some Panasonic) if correctly measured:
<>

At 1k you will measure a lower value, I got 5% lower for a 470u.

Believe whatever you want, I've pointed out what I thought to be
imprecise statements, I don't intend to do any more than that.

ST