Electronic components aging

On Tue, 15 Oct 2013 13:49:07 -0400, Spehro Pefhany
<speffSNIP@interlogDOTyou.knowwhat> wrote:

>On Tue, 15 Oct 2013 09:44:18 -0700, John Larkin
><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>>
>>>Don't know. The niobium oxide ones are claimed to use the new "no
>>>burn" technology so presumably they'll spontaneously burst into flames
>>>less frequently. 
>>
>>MnO2 tantalums are fine as long as peak current (ie, dv/dt) is limited. So,
>>don't use them to bypass power rails.
>
>They're okay for bypassing power rails if you have a local regulator
>that limits current (and if you derate the voltage). 

Derating by at least 2:1, better 3:1, seems safe. I have blown them up
downstream of an LM1117, which can peak at about 1.2 amps.


-- 

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

Piotr Wyderski wrote:
> Speaking of high reliability... I think that it is often
> a somewhat neglected issue, so I start this thread as a
> mean to collect *practical* observations for people who
> care about long MTBF. In other words, "if I had to build
> a device which should last 50 years, I would... what?"
>
> Resistors (if not overloaded): immortal
* Only if leaded and mounted according to NASA - leads having graceful 
hook-like "loops" to relieve stress. Then wrapped around turret for 
mechanical reliability, and as a final step soldered for electrical 
reliability.
   Oh, yes..100 percent test values BEFORE using as you might find the 
one in a million that is almost shorted or crystallized to high value, 
or the one in ten thousand (or less) that was improperly binned.

>
> Ceramic capacitors: as above
* Same gotchas.

> Tantalum/nobium caps: ?
* Same gotchas. Except Tants have a poor life that gets worse as 
temperature goes up. I would not recommend them for 50 year service; 
maybe 10 year service..

>
> Electrolytic caps: disaster area
* I know of the old wet can 'lytics that lasted on the average of 10-15 
years,with maybe 10 percent lasting to 20 years WITHOUT electrolyte 
replacement. And this is in "typical" home environment,some areas in the 
US hotter than others.
   The problem with the "dry" 'lytics is there is no liquid electrolyte 
that can be replaced.
   The Sprague TE series seems to be the exception, lasting over 20 
years with no apparent change (typical home environment).

>
> Transistors, diodes and ICs: the silicon die should not
> degrade, but how about the endurance of the resin?
* Excellent; have seen no problems up to and including 200C.
   Around 210C or so, then something goes irrevocably sour.
   This info is not to be confused with usability,only as an indicator 
of possible reliability.
* In the 70's or so, at (the original) Fairchild, they were making power 
transistors in the TO-3 package for hi-rel, and the Army told them to 
STOP plugging them into a test fixture as that decreased the reliability 
by at least an order of magnitude.
   The problem was, "merely" pluggint them in added stress to the 
leads,creating microcracks in the glass-to-kovar interface.
   So,extend that to anything that adds stress in lead=to=package interface.

> At least some early Polish ICs had problems here:
> the thermal coefficient of the casing was not well-matched
> and power cycling finally broke the bonding wires.
> There were some moisture absorbtion problems, too.
> Is it still an issue?
* *Temperature* cycling causes problems, most especially in vias, 
metal-glass-conductor (metal or doped silicon); due exclusively to the 
large difference in the coefficient of expansion.
* There were attempts to glop silicone on top of a chip before 
encapsulation to reduce bondwire stresses..successful in that regard,but 
too much damn moisture got encapsulated with the glop.

>
> BGA: it can be expected that thermal cycling will
> eventually destroy the balls, as there are no "springs"
> to absorb thermal stresses. Gull wings are much better here.
* Agree. Know noting about BGA issues.

>
> FR4: ?
* X,Y,Z coefficients of expansion are all different and are grossly 
different that any parts you may encounter.
   Here is where the NASA stress-relief techniques come in handy.
   PCB vias should, in my opinion, be as large as one can get away with; 
none of this 14 mil-in-thick PCB. 20 mil is OK even at 200C.

>
> Soldering: the EU has done a lot in order to make
> the newer devices not very reliable as a consequence
> of the RoHS directive. I see nothing wrong with the
> old SnPb joints, the old boards look healthy.
* I may have been the only vender that was RoHS compatible at least a 
year before the EU  made an un-avoidable rule.
   One has no choice if one wishes to guarantee (essentially) PCBs up to 
204C.. silver bearing solder is the lowest temperature must.
   I am totally against the SAC crap, despite the reasons behind the use.
   Multicore makes Savbit, a tin-lead solder alloyed with a small amount 
of copper - to eliminate the pitting of soldering bits (irons to us 
USians) due to copper in the production bits alloying out and into the 
solder.
   There is NO SUCH "CONCERN" WRT reflow soldering!
   The added copper decreases the reliability of the solder connection.
   With regard to silver-bearing solder, the same can be said about SAC 
crap.

>
> Conformal coating: ?
* Maybe..with great care and atmospheric controls (NO moisture), baking 
of units beforehand, etc.

>
> Wires: ?
* Review NASA specs WRT nicking of wire during stripping.

>
> Please add your comments.
>
> Best regards, Piotr

RobertMacy wrote:

> On Tue, 15 Oct 2013 09:14:23 -0700, <edward.ming.lee@gmail.com> wrote:
> 
>>
>>>
>>> Avoid all-silicon current paths, and don't make direct connection
>>> between
>>> silicon and the outside world.
>>
>> That means one thing: Plain Old Relays.  How long can sealed relays be
>> expected to last, if we only fire it up occasionally?  Perhaps one in
>> weeks.
> 
> 
> Would designers of those Jupiter/Staurn satellites, etc jump in here? Give
> back to the community. Their stuff works ten years and upwords of 25
> years. Be great if they wrote a little history of the 'battles' one must
> embark on and the design philosophy required to overcome THOSE reliability
> obstacles.

I think they may the first to admit that they did not think the Voyagers 
would be operating for so long. It is one thing for the systems to last a 
long time, but a completely different thing if you are asked to make sure it 
will keep going.

-- 
********************************************************************
Paul E. Bennett IEng MIET.....<email://Paul_E.Bennett@topmail.co.uk>
Forth based HIDECS Consultancy.............<http://www.hidecs.co.uk>
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-510979
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

On Tuesday, October 15, 2013 6:42:34 PM UTC+2, Spehro Pefhany wrote:
> On Tue, 15 Oct 2013 12:50:39 +0200, Piotr Wyderski
>=20
> <peter.pan@neverland.mil> wrote:
>=20
>=20
>=20
> >Speaking of high reliability... I think that it is often
>=20
> >a somewhat neglected issue, so I start this thread as a
>=20
> >mean to collect *practical* observations for people who
>=20
> >care about long MTBF. In other words, "if I had to build
>=20
> >a device which should last 50 years, I would... what?"
>=20
>=20
>=20
> Be careful about using terms like MTBF unless you know the actual
>=20
> definitions. Often it applies only to the bottom of the bathtub curve,
>=20
> so infantile failures and wear-out are not considered. IOW, a part
>=20
> with a 20 year MTBF could legitimately wear out in a few years (if it
>=20
> even lasts the first month). =20
>=20
>=20
>=20
> >Resistors (if not overloaded): immortal
>=20
>=20
>=20
> Not my experience. Resistors that get warm and resistors that are
>=20
> stressed by high voltages (especially large DC voltage) often die
>=20
> early. Surge damage (often limits are not specified) can occur on
>=20
> stuff connected to the outside world (eg. induced currents from
>=20
> lightning strikes).=20
>=20
>=20
>=20
> Resistors run at < 10% of rating and low voltage can pretty much be
>=20
> ignored, IME.=20
>=20
>=20
>=20
> Trimpots and trimcaps are pretty good too (unless abused).
>=20
>=20
>=20
> >Ceramic capacitors: as above
>=20
>=20
>=20
> Pretty reliable, not as good as unstressed resistors.=20
>=20
>=20
>=20
> >Tantalum/nobium caps: ?
>=20
>=20
>=20
> Don't know. The niobium oxide ones are claimed to use the new "no
>=20
> burn" technology so presumably they'll spontaneously burst into flames
>=20
> less frequently.=20
>=20
>=20
>=20
> >Electrolytic caps: disaster area
>=20
>=20
>=20
> Expect to replace them after 5 years to 50 years, depending on how
>=20
> much heat they see (internal and external) and other factors. But
>=20
> they're quite _reliable_, they just have a limited and fairly
>=20
> predictable life, like electromechanical relays. It's pretty much
>=20
> impossible to make a mains-powered device of any usefulness without
>=20
> electrolytic caps, and there is a lot of experience with their
>=20
> reliability- eg. my HP 333A distortion meter is loaded with them
>=20
> (maybe 50+) and it still works fine after maybe 40-45 years.=20
>=20
>=20
>=20
> Lack of electrolytic caps can lead to extreme design choices that may
>=20
> negatively affect reliability.=20
>=20

Indeed. Some people are afraid of electrolytic capacitors, but the wear-out=
 mechanisms are well known and the lifetime can be very long if you treat i=
t as you should. The thermal cycle effects which is one of the major root c=
auses for failures are low due to mechanical reliefs in the packages used

Ceramic capacitors, that is used instead of electrolytic capacitors since t=
hey are thought to have better reliability, have horrible thermal cycling s=
pecs (at least for SMD types)

Cheers

Klaus

On Tuesday, October 15, 2013 6:22:26 PM UTC+2, Robert Macy wrote:
> On Tue, 15 Oct 2013 09:14:23 -0700, <edward.ming.lee@gmail.com> wrote:
>=20
>=20
>=20
> >
>=20
> >>
>=20
> >> Avoid all-silicon current paths, and don't make direct connection =20
>=20
> >> between
>=20
> >> silicon and the outside world.
>=20
> >
>=20
> > That means one thing: Plain Old Relays.  How long can sealed relays be =
=20
>=20
> > expected to last, if we only fire it up occasionally?  Perhaps one in =
=20
>=20
> > weeks.
>=20
>=20
>=20
>=20
>=20
> Would designers of those Jupiter/Staurn satellites, etc jump in here? Giv=
e =20
>=20
> back to the community. Their stuff works ten years and upwords of 25 =20
>=20
> years. Be great if they wrote a little history of the 'battles' one must =
=20
>=20
> embark on and the design philosophy required to overcome THOSE reliabilit=
y =20
>=20
> obstacles.

I did design for ESA in an earlier job.

We used the JPL Derating guidelines (Jet Propulsion Laboratory, which later=
 became a subdivision of NASA).

The idea of course is to keep the subjected stimuli well below the ratings =
of the device. The stimuli is calculated from what is expected and any shor=
t time overload condition. Normally maximum 60% of the ratings for voltage,=
 current and power.

Reduced length document:

http://engineer.jpl.nasa.gov/practices/1201.pdf


When the basics is covered, the surroundings of the device is handled. In p=
rinciple any device is scrutinized for any failure and a DFMEA analysis is =
done to make sure what kind of propagation a single error has. The errors i=
s not just from components FIT, emvironment influences etc, but also from c=
osmic raditation. So a SEU (Single Event Upset) is also analyzed. For examp=
le, for ICs which has not passed the nessesary radiation limnits, the devic=
e is surround with components to protect it from the SEU and to reset it af=
ter the SEU

Any device has a FIT number, not only the Space rated parts, which can be f=
ound for most devices digging deeper in the manufactor tests reports. The F=
IT is normally deduces from accelerated thermal cycling tests.

As for the special RAD hard devices, that is used for Space designs, the di=
es are AFAIR special dies, used only for Space, with extensive testing on a=
ll important parameters

So any device is considered to be failure prone, and to increase availabili=
ty figures, redundant circuits are used, either fully redundant or partiall=
y redundant with voting systems

The trend recently has been to try to use non-RAD HARD components for space=
 flight, to see if it is possible to save costs with possibly lower reliabi=
lity.

Cheers

Klaus

On Wednesday, October 16, 2013 12:15:50 AM UTC+2, Klaus Kragelund wrote:
> On Tuesday, October 15, 2013 6:22:26 PM UTC+2, Robert Macy wrote:
>=20
> > On Tue, 15 Oct 2013 09:14:23 -0700, <edward.ming.lee@gmail.com> wrote:
>=20
> >=20
>=20
> >=20
>=20
> >=20
>=20
> > >
>=20
> >=20
>=20
> > >>
>=20
> >=20
>=20
> > >> Avoid all-silicon current paths, and don't make direct connection =
=20
>=20
> >=20
>=20
> > >> between
>=20
> >=20
>=20
> > >> silicon and the outside world.
>=20
> >=20
>=20
> > >
>=20
> >=20
>=20
> > > That means one thing: Plain Old Relays.  How long can sealed relays b=
e =20
>=20
> >=20
>=20
> > > expected to last, if we only fire it up occasionally?  Perhaps one in=
 =20
>=20
> >=20
>=20
> > > weeks.
>=20
> >=20
>=20
> >=20
>=20
> >=20
>=20
> >=20
>=20
> >=20
>=20
> > Would designers of those Jupiter/Staurn satellites, etc jump in here? G=
ive =20
>=20
> >=20
>=20
> > back to the community. Their stuff works ten years and upwords of 25 =
=20
>=20
> >=20
>=20
> > years. Be great if they wrote a little history of the 'battles' one mus=
t =20
>=20
> >=20
>=20
> > embark on and the design philosophy required to overcome THOSE reliabil=
ity =20
>=20
> >=20
>=20
> > obstacles.
>=20
>=20
>=20
> I did design for ESA in an earlier job.
>=20
>=20
>=20
> We used the JPL Derating guidelines (Jet Propulsion Laboratory, which lat=
er became a subdivision of NASA).
>=20
>=20
>=20
> The idea of course is to keep the subjected stimuli well below the rating=
s of the device. The stimuli is calculated from what is expected and any sh=
ort time overload condition. Normally maximum 60% of the ratings for voltag=
e, current and power.
>=20
>=20
>=20
> Reduced length document:
>=20
>=20
>=20
> http://engineer.jpl.nasa.gov/practices/1201.pdf
>=20
>=20
>=20
>=20
>=20
> When the basics is covered, the surroundings of the device is handled. In=
 principle any device is scrutinized for any failure and a DFMEA analysis i=
s done to make sure what kind of propagation a single error has. The errors=
 is not just from components FIT, emvironment influences etc, but also from=
 cosmic raditation. So a SEU (Single Event Upset) is also analyzed. For exa=
mple, for ICs which has not passed the nessesary radiation limnits, the dev=
ice is surround with components to protect it from the SEU and to reset it =
after the SEU
>=20
>=20
>=20
> Any device has a FIT number, not only the Space rated parts, which can be=
 found for most devices digging deeper in the manufactor tests reports. The=
 FIT is normally deduces from accelerated thermal cycling tests.
>=20
>=20
>=20
> As for the special RAD hard devices, that is used for Space designs, the =
dies are AFAIR special dies, used only for Space, with extensive testing on=
 all important parameters
>=20
>=20
>=20
> So any device is considered to be failure prone, and to increase availabi=
lity figures, redundant circuits are used, either fully redundant or partia=
lly redundant with voting systems
>=20
>=20
>=20
> The trend recently has been to try to use non-RAD HARD components for spa=
ce flight, to see if it is possible to save costs with possibly lower relia=
bility.
>=20

For practical design, important items to consider apart from other mentione=
d in the thread is the difference in thermal expansion coefficients from co=
mponents to the PCB. So for long reliability designs, placing ceramics on a=
 FR4 substrate is not always optimal

Leadfree solder may also not be optimal, although I think the discussion of=
 reliability of leadfree versus leaded designs have not ended

Cheers

Klaus

On Tue, 15 Oct 2013 08:39:37 -0700, John Larkin
<jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

>On Tue, 15 Oct 2013 12:50:39 +0200, Piotr Wyderski <peter.pan@neverland.mil>
>wrote:
>
>>Speaking of high reliability... I think that it is often
>>a somewhat neglected issue, so I start this thread as a
>>mean to collect *practical* observations for people who
>>care about long MTBF. In other words, "if I had to build
>>a device which should last 50 years, I would... what?"
>
>
>Keep power dissipation and currents low in semiconductors. Current causes
>electromigration. Heat wrecks everything. Use copper pours to keep temps down.
>
<snip>

You mean stay within a components rating?  Novel idea.

Cheers

On Tue, 15 Oct 2013 19:50:40 -0400, Martin Riddle
<martin_rid@verizon.net> wrote:

>On Tue, 15 Oct 2013 08:39:37 -0700, John Larkin
><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>>On Tue, 15 Oct 2013 12:50:39 +0200, Piotr Wyderski <peter.pan@neverland.mil>
>>wrote:
>>
>>>Speaking of high reliability... I think that it is often
>>>a somewhat neglected issue, so I start this thread as a
>>>mean to collect *practical* observations for people who
>>>care about long MTBF. In other words, "if I had to build
>>>a device which should last 50 years, I would... what?"
>>
>>
>>Keep power dissipation and currents low in semiconductors. Current causes
>>electromigration. Heat wrecks everything. Use copper pours to keep temps down.
>>
><snip>
>
>You mean stay within a components rating?  Novel idea.
>
>Cheers

I mean to keep junction temperatures as low as possible if you want
max MTBF. Operating "within ratings" (at max specified Tj) isn't good
for reliability.




-- 

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

On 16/10/2013 9:15 AM, Klaus Kragelund wrote:

> http://engineer.jpl.nasa.gov/practices/1201.pdf
>

The graph at the end indicates that reducing the stress ratio (as 
defined there), provides an increase in reliability, but not a dramatic 
one. One could suspect that what that graph is really showing is that 
the the higher the stress ratio, the more quickly a latent fault becomes 
manifest, rather than that the stress itself causes the fault.

Sylvia.

Paul E Bennett wrote:


> With the best will in the world, there is no electronics based technology
> that will last that long. It is hard enough doing designs that are
> required to have a life of 20 or 25 years with minimal (no) maintenance.
> After that time the equipment is replaced in major refurbishment
> programmes.
I have a digital clock I made 40 years ago.  It has a flaky connection
somewhere, but still works.  I have a stereo amp I made even longer ago
that still works.

Jon