Tantalum cap failures

[[email protected]]

Hello Mike

Unfortunately, the requirement is worse than that - it's 10 years
shelf-life AND 25 years operational.

However, I accept the principal in what you suggest and we do have
items on the aircraft that have scheduled maintenance, though we avoid
it like the plague if we can.

Part of the problem is that we can't get at some of these caps to
replace them once the unit is fully assembled, due to a partial
encapsulation which is applied for mechanical reasons. This
effectively means we would have to bin the whole PCB, which is rather
expensive.

With regard to failures in use, things are not as bad as they first
seem. We have comprehensive BIT, (Built In Test), that monitors the
circuit at all times and shuts down motor drive in the event of a
problem. If the fault took out the circuit breaker, the loss of power
to the equipment would have exactly the same effect as a shutdown by
BIT.

Under these circumstances, whilst the pilot would have lost some
functionality, he would still have control and would be able to get
home without undue dificulty.

Ted

 

[[email protected]]

The supply to our equipment is nominally 28V DC, derived from AC
generators and 3-phase rectification, so, whilst it is subject to
variations, it is essentially clean of any high frequency noise.

The PWM takes place inside our equipment and comprises a three-phase
motor winding, with each phase connected to a totem-pole power mosfet
drive, connected across the 28V. The windings are therefore switched
between 28V and 0V, with current drawn from and then dumped back into
the 28V supply.

The array of tantalum capacitors is there to source/sink the PWM
currents and limit voltage ripple on the 28V to an acceptable level. A
series inductor between our internal 28V and the aircraft 28V then
limits the exported current ripple to a prescribed figure.

We could have gone for linear control of the motor, (there are days
when I wish we had!), but we would have simply traded all these
problems for that of getting rid of all the heat generated.

Ted

 

[[email protected]]

Quite - I've had difficulty getting this point through to a number of
people.

I have posted a follow-up to this thread stating that we are going to
use MLCCs. However, since then, a colleague from a different site has
told me that he has experienced a lot of reliability problems with
multi-layer ceramics, caused by defects in manufacture that can lead to
combustion during use.

I'm currently digging around on google trying to get more info on this
- I certainly don't want to trade one unreliable component for another.

If anyone has any specific knowledge/experience of MLCCs, I would be
most grateful for some feedback.

Ted

 

[Pooh Bear]

The supply to our equipment is nominally 28V DC, derived from AC
generators and 3-phase rectification, so, whilst it is subject to
variations, it is essentially clean of any high frequency noise.

The PWM takes place inside our equipment and comprises a three-phase
motor winding, with each phase connected to a totem-pole power mosfet
drive, connected across the 28V. The windings are therefore switched
between 28V and 0V, with current drawn from and then dumped back into
the 28V supply.

Ahhh !

So when you switch off a motor winding there's no appreciable limit to the
flyback pulse other than the motor winding's own inductance ?

Any snubbing ?

The array of tantalum capacitors is there to source/sink the PWM
currents and limit voltage ripple on the 28V to an acceptable level. A
series inductor between our internal 28V and the aircraft 28V then
limits the exported current ripple to a prescribed figure.

We could have gone for linear control of the motor, (there are days
when I wish we had!), but we would have simply traded all these
problems for that of getting rid of all the heat generated.

Ok - got it.

Graham

 

[Pooh Bear]

Quite - I've had difficulty getting this point through to a number of
people.

I have posted a follow-up to this thread stating that we are going to
use MLCCs. However, since then, a colleague from a different site has
told me that he has experienced a lot of reliability problems with
multi-layer ceramics, caused by defects in manufacture that can lead to
combustion during use.

I've seen standard MLCs burn up just like tantalums. Same mechanism. High
pulse current. You can even do it with plastic film caps if you try hard
enough !


Graham

 

[Mike Monett]

Hello Mike

Unfortunately, the requirement is worse than that - it's 10 years
shelf-life AND 25 years operational.

However, I accept the principal in what you suggest and we do have
items on the aircraft that have scheduled maintenance, though we
avoid it like the plague if we can.

Part of the problem is that we can't get at some of these caps to
replace them once the unit is fully assembled, due to a partial
encapsulation which is applied for mechanical reasons. This
effectively means we would have to bin the whole PCB, which is
rather expensive.

With regard to failures in use, things are not as bad as they
first seem. We have comprehensive BIT, (Built In Test), that
monitors the circuit at all times and shuts down motor drive in
the event of a problem. If the fault took out the circuit breaker,
the loss of power to the equipment would have exactly the same
effect as a shutdown by BIT.

Under these circumstances, whilst the pilot would have lost some
functionality, he would still have control and would be able to
get home without undue dificulty.

Ted

Well, since the pcb is encapsulated, when the tantalums and MLCC
fail catastrophically, you have to replace the pcb anyway.

It looks like the only capacitor type you have left is solid
polymer.

If you use solid polymer, you only need to check the ripple voltage
during scheduled maintenance. If the voltage looks like it will go
out of spec between now and the next scheduled maintenance cycle,
junk the pcb. Otherwise log the ripple value and put it back in
service.

If the pcb has been on the shelf for 10 years, obviously you want to
test it on the bench before installing it in the a/c. Include the
ripple voltage as part of the checkout procedure.

Since the degradation in ripple voltage is gradual, this will give
plenty of advance warning of a system failure. This is infinitely
better than an unexpected and catastrophic failure at night, close
to the ground, in the rain.

One of the biggest problems in flying is handling multiple failures.
Pilots who can easily cope with one or two problems may become
saturated with one more failure and crash the plane.

Anything you can do to avoid an unexpected catastrophic failure may
end up saving the plane and the pilot.

Mike Monett

 

[[email protected]]

Our PWM frequency is nominally 80kHz, and this is a reasonable
compromise between switching losses and supply capacitance requirements
for our application.

I've re-visited my design notes and, with the SFE tants, the I*esr
product dominates capacitor ripple voltage, being some five times
larger than the dV/dt component. I've since been looking at MLCCs,
which have a quoted esr of 2m-ohms, and using these for the supply
reservoir, I could drop the total capacitance to around 100uF and still
meet the ripple voltage requirement.

However, there are doubts regarding the reliability of MLCCs, so I'm
reviewing things at the moment.

A couple of people had suggested using cap multipliers, but
unfortunately, as John Larkin points out, active parts don't store
energy.

Ted

 

[Paul Mathews]

Perhaps this wasn't clear from my earlier post: Many capacitor failures
can be traced to solder process and mechanical stresses. For example,
MLCCs don't usually fail due to current or voltage stress alone,
instead, they fail after being damaged by mechanical stresses resulting
from soldering, board handling, and temperature cycling. Many methods
have been devised to minimize such effects, including these:
1. Use the smallest possible package sizes and parallel multiple small
caps.
2. Take steps to minimize PCB flexing for SMT components and/or use
packages with leads.
3. Avoid encapsulation, particularly with hard encapsulating materials.
4. Avoid all processing steps that cause mechanical stresses, e.g.,
lead bending, handling with pliers, dispensing loose parts into bins,
etc.

Another poster mentioned snubbing. Perhaps you already replied to
that, but it's worth mentioning that a combination of snubbing and
shunt filtering (parallel caps with no series resistor) can be much
more effective than shunt filtering alone.

Paul Mathews

 

[John Larkin]

Our PWM frequency is nominally 80kHz, and this is a reasonable
compromise between switching losses and supply capacitance requirements
for our application.

I've re-visited my design notes and, with the SFE tants, the I*esr
product dominates capacitor ripple voltage, being some five times
larger than the dV/dt component. I've since been looking at MLCCs,
which have a quoted esr of 2m-ohms, and using these for the supply
reservoir, I could drop the total capacitance to around 100uF and still
meet the ripple voltage requirement.

However, there are doubts regarding the reliability of MLCCs, so I'm
reviewing things at the moment.

A couple of people had suggested using cap multipliers, but
unfortunately, as John Larkin points out, active parts don't store
energy.

Ted

Look into polymer tantalums. Super-low esr and no ignition mechanism.
If you're esr limited, the polymers would allow you to go with less uF
and higher voltage ratings, for more reliability.

John

 

[[email protected]]

Hello Paul

Your original post was clear enough and I took it to mean that MLCC
failures could occur due to misshandling at some stage during assembly.

However, someone has since told me that many failures occur in MLCCs
due to problems in the manufacturing process - e.g. cavities and loss
of metalisation - but that the manufacturers point to handling problems
as a smoke-screen for these failures. That's what I've been Googling
to find more info on. Problems that can be prevented by appropriate
build control is one thing, but I've no wish to swap one unreliable
component for another.

With regards to snubbing, I considered fitting snubbers early in the
design and decided we didn't need them. I need to do some digging to
check my reasoning.

Since we're dealing with what's essentially a current source here, all
a snubber would do is slow down the rate of change of current in the
caps. Are you saying that's an issue?

Ted

 

[Pooh Bear]

Hello Paul

Your original post was clear enough and I took it to mean that MLCC
failures could occur due to misshandling at some stage during assembly.

However, someone has since told me that many failures occur in MLCCs
due to problems in the manufacturing process - e.g. cavities and loss
of metalisation - but that the manufacturers point to handling problems
as a smoke-screen for these failures. That's what I've been Googling
to find more info on. Problems that can be prevented by appropriate
build control is one thing, but I've no wish to swap one unreliable
component for another.

With regards to snubbing, I considered fitting snubbers early in the
design and decided we didn't need them. I need to do some digging to
check my reasoning.

Since we're dealing with what's essentially a current source here, all
a snubber would do is slow down the rate of change of current in the
caps. Are you saying that's an issue?

I most certainly would !

Graham

 

[Mike Monett]

I have posted a follow-up to this thread stating that we are going
to use MLCCs. However, since then, a colleague from a different
site has told me that he has experienced a lot of reliability
problems with multi-layer ceramics, caused by defects in
manufacture that can lead to combustion during use.

I'm currently digging around on google trying to get more info on
this I certainly don't want to trade one unreliable component for
another.

If anyone has any specific knowledge/experience of MLCCs, I would
be most grateful for some feedback.

Ted

Yes, here's more confirmation of the failure mechanism of
multi-layer ceramics, and a possible solution using Metallized
Plastic capacitors (MLP):

--------------------------------------------------------------------
"Optimizing Output Filters Using Multilayer Polymer Capacitors in
High Power Density Low-Voltage Converters"

Bruce Carsten
ITW Paktron
Lynchburg, Virginia

Surface-mounted MLC capacitors have found some use at frequencies of
1MHz, where their impedance can be less than that of electrolytics.
Price has been something of a limiting factor in commercial
applications, but the major technological problem has been the
construction of semi-stable capacitors larger than luF, which can be
surface-mounted on printed circuit boards. Even if they survive the
stress of soldering, temperature cycles tend to crack larger ceramic
capacitors due to the coefficient of thermal expansion mismatch
between the capacitor and substrate.

Metallized Plastic capacitors have similar electrical properties to
MLCs without the differential expansion problem, but until recently,
they have been larger than MLCs of the same capacity. This has
changed with thinner dielectrics and new construction techniques;
they are now volumetrically competitive with ceramic capacitors and
will become more so in the future.

[...]

MLP "Plastic" Capacitors. These are the lowest cost capacitors of
the low HF impedance capacitor types. No expensive raw materials are
used, so costs can be expected to hold or decline with improvements
in film manufacturing technology.

The self-healing capability of metallized plastic capacitors make
them quite reliable, with relatively graceful and typically gradual
failure with over-voltage.

[I once successfully (and somewhat dramatically) demonstrated the
robustness of this self-healing property to a skeptical fellow
engineer by cutting the corner off of a capacitor with a hacksaw,
and then driving a nail through a drilled hole in the middle. About
25 percent of the capacity was lost, but it would still withstand
rated voltage.]

ESR is very low, typically 3 milliohms at 1MHz for a 10uF, 50V
capacitor, and can be lowered further if needed with heavier
metallization (which will have minimal effect on capacitor volume).
RMS current capability is correspondingly high, on the order of
1A/uF. Voltage and current ratings apply to 85 degrees C, where ESR
is at a broad minimum with polyester dielectric; operating
temperatures to 125 degrees C are possible with derating.

Here's the url's:

411k PDF:

http://www.powerpulse.net/powerpulse/archive/pdf/aa_021901b.pdf

HTML:

http://www.powerpulse.net/powerpulse/archive/aa_021901b1.stm

Mike Monett

 

[Mike Monett]

Ted: There's a lot more information on MLP capacitors at

http://www.paktron.com/techarticles/main.html

Another interesting comment on MLCC failures : they are piezoelectric and
may crack with high inrush currents.

Mike Monett

 

[[email protected]]

Well, if parasitics in the caps led to high dI/dt causing excessive
voltage, I could see the problem, but that's not the case here - I've
done extensive measurement using a 100MHz DSO during the high current
events and there are no troublesome voltage spikes on the caps and
voltage remains comfortably within rated operation throughout.

(We have an array of 14x470nF sm ceramics in parallel with the tants to
help deal with the HF content and this seems to work well in that
respect).

If there's another problem mechanism in tantalums relating to high
dI/dt, I'm very interested to hear about it.

Ted

 

[Paul Mathews]

Re snubbing: I was assuming that the main purpose of the caps was to
reduce conducted and radiated EMI, much of which is presumably in the
form of ringing waveforms. Damping with critical value can result in
less EMI overall with less capacitance required.
Paul Mathews

 

[[email protected]]

Hello again John

I've been looking at these but I've not been able to find anything with
a higher working voltage than 25V, (searching not helped by the fact
that Google keeps hanging up whilst trying to access data for no
apparent reason).

I'm coming round to the view that we might be best using solid polimer
aluminiums, with a ripple-voltage monitor linked into BIT. They quote
a life of more than 10 years, but when you look at the curves of
projected life against temperature and RH, things look much more
optimistic - greater than 100 years at 75C and 45%RH. (How true these
forecasts are is another matter, but I won't be around to worry about
it).

Ted

 

[[email protected]]

Hello once again Mike

Having gone all around the houses, I've come to the conclusion that
solid polymers are the way to go, with a ripple-voltage monitor linked
into BIT.

As you say, they forecast very long life under more benign conditions
than the +125C/85% RH extreme - in practice, I reckon they would meet
both the shelf-life and operational life requirements in our
environment.

Ted

 

[[email protected]]

Hello Paul

The main source of EMC emissions in this design is the bi-directional,
rectangular supply current waveform and that's what the tantalums are
there to deal with. Yes, there are also EMC issues relating to the PWM
motor drive waveforms and we had to introduce specific measures to deal
with this, in order to meet our EMC emmissions spec.

However, we don't fit snubbers to the PWM outputs as these would
require a peak pulse current rating of 20A at a PRF of nominally
160kHz.

Ted

 

[John Larkin]

Hello again John

I've been looking at these but I've not been able to find anything with
a higher working voltage than 25V, (searching not helped by the fact
that Google keeps hanging up whilst trying to access data for no
apparent reason).

I'm coming round to the view that we might be best using solid polimer
aluminiums, with a ripple-voltage monitor linked into BIT. They quote
a life of more than 10 years, but when you look at the curves of
projected life against temperature and RH, things look much more
optimistic - greater than 100 years at 75C and 45%RH. (How true these
forecasts are is another matter, but I won't be around to worry about
it).

Ted

Regular aluminums fail because the water leaks out, and polymer elecs
fail because the water leaks in!

John

 

[Robert Baer]

Regular aluminums fail because the water leaks out, and polymer elecs
fail because the water leaks in!

John

From Digikey:
1) Kemet high temperature wet slug Tantalum caps; 125V@85C is highest
voltage i could find, de-rated to 62V@200C - values 1.7uF to 82uF.
2) BC high temperature Aluminum caps; 200V@125C is highest voltage i
could find, "useable" to 150C - values 2.2uF to 100uF.