TEC failure mode?
Samuel M. Goldwasser
(what of it there is originally!)? Overheating causing the junctions
to change somehow?
I've seen several cases where there is no physical damage and the resistance
is still low, but cooling no longer has enough efficiency to be useful. The
thing simply gets hot on both sides. This even with the same heatsinking
and drive current as a good one that can freeze icecubes on the cold side.
--
sam | Sci.Electronics.Repair FAQ: http://www.repairfaq.org/
Repair | Main Table of Contents: http://www.repairfaq.org/REPAIR/
+Lasers | Sam's Laser FAQ: http://www.repairfaq.org/sam/lasersam.htm
| Mirror Sites: http://www.repairfaq.org/REPAIR/F_mirror.html
Important: Anything sent to the email address in the message header above is
ignored unless my full name AND either lasers or electronics is included in the
subject line. Or, you can contact me via the Feedback Form in the FAQs.
Lostgallifreyan
news:3a7p6n...@repairfaq.org:
> What would cause a TEC (i.e., Peltier device) to lose efficiency
> (what of it there is originally!)? Overheating causing the junctions
> to change somehow?
>
> I've seen several cases where there is no physical damage and the
> resistance is still low, but cooling no longer has enough efficiency to
> be useful. The thing simply gets hot on both sides. This even with the
> same heatsinking and drive current as a good one that can freeze
> icecubes on the cold side.
>
They don't like water (condensation). It causes migration of atoms across the
divide between elements, I think. I forget the details but it can and does
erode performance in some similar way given time. Overheating also might
cause it but I don't know how, usually it just melts the solder and they
collapse if not mounted with even compression. If thermal cycling was the
cause of embrittlement the loss of conductivity would show up.
James Sweet
> What would cause a TEC (i.e., Peltier device) to lose efficiency
> (what of it there is originally!)? Overheating causing the junctions
> to change somehow?
>
> I've seen several cases where there is no physical damage and the resistance
> is still low, but cooling no longer has enough efficiency to be useful. The
> thing simply gets hot on both sides. This even with the same heatsinking
> and drive current as a good one that can freeze icecubes on the cold side.
>
> --
>
They are semiconductors aren't they? I suspect the individual junctions
could fail as a short circuit just as diodes and other semiconductors
can, that will certainly make it get hot.
Samuel M. Goldwasser
True, but we're talking about TECs with many dozens or over 100 junctions.
For them all to fail would mean some kind of gross overcurrent, which
doesn't seem likely given the way they are driven. If a few failed, that
would partially short circuit the thermal isolation between the two
surfaces. So, that could be it. How many though?
Samuel M. Goldwasser
I doubt these had condesnation. In fact, some of them were coated with
RTV silicone. I have seen the collaps failure mode. :( :)
Lostgallifreyan
> I doubt these had condesnation. In fact, some of them were coated with
> RTV silicone.
Which is itself hygroscopic, isn't it? I know some silicone sealants require
water (usually from atmospheric vapour) to cure, not sure about that one
though, or any remaining porosity afterwards. I think some of them outgas
some nasty fumes on a small scale during cure too. Maybe chlorine based? That
could change the chemistry a bit if it happened in a sealed volume. (Trying
not to get confused with cyanoacrylates here..) Anyway, I'm thinking someone
might have used the wrong kind of silicone sealant.
Lostgallifreyan
> True, but we're talking about TECs with many dozens or over 100 junctions.
> For them all to fail would mean some kind of gross overcurrent,
Indeed, and on a current controlled drive (maybe it isn't but bear with me..)
either the shorts would result in lowering resistance, or a failure of
cooling power would cause the driver to excell itself and overgeat the TEC to
collapse, or at least alerting the user in some other fairly obvious way.
I still think it's a chemical attach thing, given the fairly large area that
so many couples offer to their surroundings, but either way it's a
semiconductor so would it help to try low current noise readings on it and
comparing with others from same or different source, same as with other kinds
of semiconductor junction? I have no idea how revealing that might be, it's
just a passing idea..
Phil Hobbs
> James Sweet <james...@gmail.com> writes:
>
>> Samuel M. Goldwasser wrote:
>>> What would cause a TEC (i.e., Peltier device) to lose efficiency
>>> (what of it there is originally!)? Overheating causing the junctions
>>> to change somehow?
>>> I've seen several cases where there is no physical damage and the
>>> resistance
>>> is still low, but cooling no longer has enough efficiency to be useful. The
>>> thing simply gets hot on both sides. This even with the same heatsinking
>>> and drive current as a good one that can freeze icecubes on the cold side.
>>> --
>> They are semiconductors aren't they? I suspect the individual
>> junctions could fail as a short circuit just as diodes and other
>> semiconductors can, that will certainly make it get hot.
>
> True, but we're talking about TECs with many dozens or over 100 junctions.
> For them all to fail would mean some kind of gross overcurrent, which
> doesn't seem likely given the way they are driven. If a few failed, that
> would partially short circuit the thermal isolation between the two
> surfaces. So, that could be it. How many though?
>
TECs are generally wired with all the junctions in series, believe it or
not. I suspect that you've got cracks due to thermal cycling, which
lead to excess I**2R dissipation in the thin spots. Ferrotec, Marlow,
and (iirc) Tellurex have a hard-solder option that improves the
temperature cycling behaviour enormously. Melcors are great right near
room temperature (e.g. for diode stabilization) but 100 full range
temperature cycles can kill one.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
Lostgallifreyan
news:4I2dne_hCoAB_RbX...@supernews.com:
> TECs are generally wired with all the junctions in series, believe it or
> not. I suspect that you've got cracks due to thermal cycling, which
> lead to excess I**2R dissipation in the thin spots.
Wouldn't that lead to a proportional increase in total resistance though?
I'm starting to google for this cos it's interesting. Nothing yet, save this
awesome irrelevancy:
"MICROSTRUCTURAL FAILURE MECHANISMS IN COOKED AND AGED CARROTS".
Thought that was far too good not to give it an honourable mention.
http://doi.wiley.com/10.1111/j.1745-4603.2000.tb00301.x
Don't know where it leads, it gave me a cookie error but the title alone is
worth the price of admission.
James Sweet
> James Sweet <james...@gmail.com> writes:
>
>> Samuel M. Goldwasser wrote:
>>> What would cause a TEC (i.e., Peltier device) to lose efficiency
>>> (what of it there is originally!)? Overheating causing the junctions
>>> to change somehow?
>>> I've seen several cases where there is no physical damage and the
>>> resistance
>>> is still low, but cooling no longer has enough efficiency to be useful. The
>>> thing simply gets hot on both sides. This even with the same heatsinking
>>> and drive current as a good one that can freeze icecubes on the cold side.
>>> --
>> They are semiconductors aren't they? I suspect the individual
>> junctions could fail as a short circuit just as diodes and other
>> semiconductors can, that will certainly make it get hot.
>
> True, but we're talking about TECs with many dozens or over 100 junctions.
> For them all to fail would mean some kind of gross overcurrent, which
> doesn't seem likely given the way they are driven. If a few failed, that
> would partially short circuit the thermal isolation between the two
> surfaces. So, that could be it. How many though?
>
> --
>
Are they sensitive to ESD? What about overheating? If the fan or bond to
the heatsink on the hot side failed a TEC would get very hot very fast.
Christoph Bollig
Hi Sam and everyone else,
I'm not an expert, but I believe the heat sink plays a big role.
Apparently it can be a problem, if only a few of the cells are not
heatinked properly, e.g. if the heat sink is a bit too small, has a
hole somewhere or a bubble in the heat sink compount or glue.
I believe the solder they use has a relatively low melting point. If
the hot side gets too hot, it might melt and cause damage.
The other problem we have experienced in the past is a runaway effect
if the temperature difference becomes too large. In that case, the TEC
has to work harder and causes more heat. Eventually the heat sink
cannot cope any more and the whole thing heats up.
BTW, are these really semiconductors? I was under the impression that
they use two different metals. But that's just what I have picked up
somewhere long ago...
Christoph
Samuel M. Goldwasser
You're thinking of thermocouples.
TECs appear to be blocks of N and P type semiconductor (no junction) soldered
between the two plates. Current into the N-type semcidonctor heats the
same side and cools the opposite side. Current into the P-type semiconductor
does the opposite.
Phil Hobbs
The resistance would go up some, but the reduced Seebeck effect voltage
drop (due to the degraded delta-T) would mask it for some considerable
time.
Time to try bending a few carrots.
Lostgallifreyan
> The resistance would go up some, but the reduced Seebeck effect voltage
> drop (due to the degraded delta-T) would mask it for some considerable
> time.
>
Interesting. Not thought of that.
> Time to try bending a few carrots.
>
I'm on a hat-trick here, first cherries, now carrots... next? I'll think of
something. :)
Lostgallifreyan
> What would cause a TEC (i.e., Peltier device) to lose efficiency
> (what of it there is originally!)? Overheating causing the junctions
> to change somehow?
>
> I've seen several cases where there is no physical damage and the
> resistance is still low, but cooling no longer has enough efficiency to
> be useful. The thing simply gets hot on both sides. This even with the
> same heatsinking and drive current as a good one that can freeze
> icecubes on the cold side.
>
Sam, I looked in a Melcor data book to help answer a general TEC question on
sci.electronic.components, and I read this:
"At temperatures above 80�C two phenomena seriously reduce useful life:
Above 80�C copper diffusion into the thermoelements occurs due to increasing
solid solubility in the thermoelectric material and increasing diffusion
rate. At 100-110�C the combined solubility and diffucion rate could result in
approximately 25% loss of device performance within 100 hours.
Above 85�C in the soldering process (using bismuth-tin alloy) small amounts
of selenium, tellurium, antimony and nickel are inherently dissolved into the
bismuth-tin solder. Although the melting point of the base solder is 136�C,
the combined mixture of all elements results in either a minute eutectic
phase or a highly effective solid state reaction occuring at above 85�C that
starts to delaminate the ends of the thermoelements by physical penetration
between cleavage planes in the thermoelectric material. This results in a
mechanical failure of the interface."
(That was written, not a copy/paste, but I checked it carefully).
Samuel M. Goldwasser
> s...@repairfaq.org (Samuel M. Goldwasser) wrote in
> news:3a7p6n...@repairfaq.org:
>
> > What would cause a TEC (i.e., Peltier device) to lose efficiency
> > (what of it there is originally!)? Overheating causing the junctions
> > to change somehow?
> >
> > I've seen several cases where there is no physical damage and the
> > resistance is still low, but cooling no longer has enough efficiency to
> > be useful. The thing simply gets hot on both sides. This even with the
> > same heatsinking and drive current as a good one that can freeze
> > icecubes on the cold side.
> >
>
> Sam, I looked in a Melcor data book to help answer a general TEC question on
> sci.electronic.components, and I read this:
>
> "At temperatures above 80�C two phenomena seriously reduce useful life:
>
> Above 80�C copper diffusion into the thermoelements occurs due to increasing
> solid solubility in the thermoelectric material and increasing diffusion
> rate. At 100-110�C the combined solubility and diffucion rate could result in
> approximately 25% loss of device performance within 100 hours.
I suppose that if the laser was run without proper external cooling
unattended, 100 hours (or 200 or 300 hours) isn't that long. And the
behavior is a loss of efficiency, not a total failure (although we've
all see that no doubt as well).
--
sam | Sci.Electronics.Repair FAQ: http://www.repairfaq.org/
Repair | Main Table of Contents: http://www.repairfaq.org/REPAIR/
+Lasers | Sam's Laser FAQ: http://www.repairfaq.org/sam/lasersam.htm
| Mirror Sites: http://www.repairfaq.org/REPAIR/F_mirror.html
Important: Anything sent to the email address in the message header above is
ignored unless my full name AND either lasers or electronics is included in the
subject line. Or, you can contact me via the Feedback Form in the FAQs.
> Above 85�C in the soldering process (using bismuth-tin alloy) small amounts
Lostgallifreyan
> Lostgallifreyan <no-...@nowhere.net> writes:
>
>> s...@repairfaq.org (Samuel M. Goldwasser) wrote in
>> news:3a7p6n...@repairfaq.org:
>>
>> > What would cause a TEC (i.e., Peltier device) to lose efficiency
>> > (what of it there is originally!)? Overheating causing the junctions
>> > to change somehow?
>> >
>> > I've seen several cases where there is no physical damage and the
>> > resistance is still low, but cooling no longer has enough efficiency
>> > to be useful. The thing simply gets hot on both sides. This even
>> > with the same heatsinking and drive current as a good one that can
>> > freeze icecubes on the cold side.
>> >
>>
>> Sam, I looked in a Melcor data book to help answer a general TEC
>> question on sci.electronic.components, and I read this:
>>
>> "At temperatures above 80�C two phenomena seriously reduce useful life:
>>
>> Above 80�C copper diffusion into the thermoelements occurs due to
>> increasing solid solubility in the thermoelectric material and
>> increasing diffusion rate. At 100-110�C the combined solubility and
>> diffucion rate could result in approximately 25% loss of device
>> performance within 100 hours.
>
> I suppose that if the laser was run without proper external cooling
> unattended, 100 hours (or 200 or 300 hours) isn't that long. And the
> behavior is a loss of efficiency, not a total failure (although we've
> all see that no doubt as well).
>
That 'unattended' is the killer. TEC circuits are usually meant to be self-
maintaining so people will tend to unattend them.. I like their description,
it seems to fit what you described. My earlier assumption was that water or
other chemical played a part in the degradation, and I think their assumption
is that water was excluded. I think you said it had been excluded too, but I
doubt it takes much of a leak, if it can get in, it WILL collect, because
condensation forms a pressure difference that will draw in more. On the other
hand it's easy to prove if it was there, because it will sit around as
evidence.
heruursciences
> s...@repairfaq.org (Samuel M. Goldwasser) wrote innews:1vm95e...@repairfaq.org:
The most common failure mode in TEC's I've seen, especially higher
power level is migration of the solder alloying with the PbTe diodes
that swaps the pd out for sn and in. over time it causes embrittlement
and fracture at the junction and the TEC stops working.
Lostgallifreyan
209490...@o21g2000vbl.googlegroups.com:
> On Sep 15, 6:00�am, Lostgallifreyan <no-...@nowhere.net> wrote:
>> s...@repairfaq.org (Samuel M. Goldwasser) wrote innews:1vm95e64.fsf@repai
Any idea if the cause is water presence, or thermal cycling? I imagine it
doesn't just happen, or TEC's wouldn't be much use in space where they can't
maintain them easily.
Phil Hobbs
Space TECs are made with hard solder, and are very much more durable
mechanically. They survive thousands of thermal cycles, whereas
soft-solder ones (e.g. Melcors) die after about 100. Marlow, Tellurex,
and Ferrotec make hard solder TECs, but iirc Marlow's are the only
space-qualified ones, and have prices to match. I don't know about the
intermetallic problem.
Some TECs come stuffed with closed-cell foam, which reduces convective
loss inside the TEC structure. I'm not sure if it's a win or a loss as
far as condensation goes--the water would have a harder time getting in,
and a harder time getting out again.
Lostgallifreyan
> Space TECs are made with hard solder, and are very much more durable
> mechanically. They survive thousands of thermal cycles, whereas
> soft-solder ones (e.g. Melcors) die after about 100.
A tad low, surely? If it were so, we could tune into Google on the 'Dead
peltier' waveband and hear the overclockers screaming, no? :)
Phil Hobbs
They don't do as much cycling as you might think--the big idea there is
to keep the chip near room temperature, and let the HS get hotter.
That's much gentler than a full temperature cycle.
Lostgallifreyan
> Lostgallifreyan wrote:
>> Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote in
>> news:Wq6dnS7ZLLQCblDX...@supernews.com:
>>
>>> Space TECs are made with hard solder, and are very much more durable
>>> mechanically. They survive thousands of thermal cycles, whereas
>>> soft-solder ones (e.g. Melcors) die after about 100.
>>
>> A tad low, surely? If it were so, we could tune into Google on the 'Dead
>> peltier' waveband and hear the overclockers screaming, no? :)
>
> They don't do as much cycling as you might think--the big idea there is
> to keep the chip near room temperature, and let the HS get hotter.
> That's much gentler than a full temperature cycle.
>
That's sort of what I was getting at though, it seems unlikely that full
cycle lifetime is a useful guide to lifetime as it's unlikely to happen. I
imagine a full thermal cycle would have to be defined as a rise to full
deltaT and back to equality, and possibly falling through to minus deltaT and
returning to equality for a literal full cycle, like those of a periodic
waveform. Even the half-wave case is only likely to occur for sensors in
conditions where some access for replacement ws planned for.
Anyone know whether thermopile laser power metres fail much from this kind of
problem? Do people with a few tens of watts have to replace a sensor within a
couple of hundred readings or less because of TEC failure as opposed to other
reasons like damaging of coatings?
Phil Hobbs
No, of course not. But if you don't believe me about soft solder vs
hard solder TECs, try looking at Marlow's or Ferrotec's literature.
Melcor TECs are really horrible for applications with lots of
cycling--wide range laser tuning or CCD cooling, for instance.
Lostgallifreyan
No doubts of hard solders being better so long as nothing gets stressed
enough to take advantage of any brittleness in the device. I just thought (or
hoped) that all but the cheapest soft soldered TECs used an alloy that was
ductile enough not to fatigue too quickly.
What is curious is that difference between Marlow and Melcor, considering
that they seem to be rivals in the same markets for lots of devices.
Joerg
Maybe I was darn lucky but I've yanked dozens of TECs back and forth
between 0C (sometimes less) and 40C or so constantly, during system
tests. Definitely many hundred times. Those were not the expensive
space-rated ones yet none died on me or became sluggish. They were all
in PID loops but I'd have noticed if one would have started to go.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Phil Hobbs
The hard solder ones will easily handle that, just not the cheap ones.