Durability/reliability of components from various manufacturers

[Perry Babin]

Hello all,


Has anyone done head to head tests of identical parts from various
manufacurers?

I'm specifically interested in power BJTs. In comparing TIP35C/TIP36C
transistors from two different manufacturers (both very well respected),
I've found that one repeatedly failed before the other (both were driven
until they failed). I did this with different batches (different date
codes) and the results were the same. I'd be particularly interested in
any information about poor quality components from Fairchild, On-Semi
and ST Microelectronics (all of which are top quality brands).

As a side note, there was one manufacturer whose transistors failed in
virtually 100% of the amplifiers that they were used in (not one of the
manufacturers listed above).

Any information about manufacturers who make poor quality or marginal
quality parts would be greatly appreciated.

Thanks in advance
Perry Babin

 

[Active8]

[snip]

I've come across situations where parts from different manufacturers, but
with the same part numbers, have different specs. Take, for example, the
BDV67A. These are power Darlingtons. I've been using the ones made by
SEMELAB, which are rated for a total of 200W continuous power dissipation.
Now, if you check the ones made by MOSTEK (again, BDV67A), they are rated
for only 125W !!! That's a BIG difference for power designs. Comparing SOA
graphs shows the MOSTEK ones can take around HALF the current of the SEMELAB
ones at the same Vce. Of course the differences in this case are documented,
but it just goes to show that a part number is sometimes not enough, but the
manufacturer must be specified too and data sheets checked carefully.

thanks. it's more work, but it could keep me from getting bitten.

mike

 

[Spehro Pefhany]

Hello all,


Has anyone done head to head tests of identical parts from various
manufacurers?

I'm specifically interested in power BJTs. In comparing TIP35C/TIP36C
transistors from two different manufacturers (both very well respected),
I've found that one repeatedly failed before the other (both were driven
until they failed). I did this with different batches (different date
codes) and the results were the same. I'd be particularly interested in
any information about poor quality components from Fairchild, On-Semi
and ST Microelectronics (all of which are top quality brands).

As a side note, there was one manufacturer whose transistors failed in
virtually 100% of the amplifiers that they were used in (not one of the
manufacturers listed above).

Any information about manufacturers who make poor quality or marginal
quality parts would be greatly appreciated.

Sorry, I don't see how you could possibly draw any useful conclusions
about quality (conformance to specifications) from such tests.

Best regards,
Spehro Pefhany

 

[Tony Williams]

Has anyone done head to head tests of identical parts from various
manufacurers? [snip]
Any information about manufacturers who make poor quality or
marginal quality parts would be greatly appreciated.

I've seen similar poor quality assembly in TVS devices.

Trouble is, you don't know who makes what these days.

Are they actually making the whole devices themselves?

Are they buying the chips in from a foundry and just
doing the final packaging and testing?

Are they even buying-in the finished product and marking
them up as their own devices?

 

[Kevin Kilzer]

Sorry, I don't see how you could possibly draw any useful conclusions
about quality (conformance to specifications) from such tests.

Quality has nothing to do with the specification. A specification can
be written to enhance quality, but "quality" itself is a subjective
measurement. Quality can be affected by everything in the
manufacturing process, from the accuracy of construction to how a
device is handled. Most quality issues do not show up until the
devices have been in the field for many years.

For example, electronic components can be damaged by static
electricity, and that damage cannot be immediately detected by
electrical measurements, but affects the life-expectancy of the part.

Variations between manufacturers have been around for a long time, and
are a fact of life. Large electronics companies have entire
departments that inspect factories before buying from those companies,
and invent tests for every lot when the devices arrive. This may
include running many samples at high temperature for weeks or months,
before using that lot of parts.

Kevin

 

[Zak]

Quality has nothing to do with the specification. A specification can
be written to enhance quality, but "quality" itself is a subjective
measurement. Quality can be affected by everything in the
manufacturing process, from the accuracy of construction to how a
device is handled. Most quality issues do not show up until the
devices have been in the field for many years.

For example, electronic components can be damaged by static
electricity, and that damage cannot be immediately detected by
electrical measurements, but affects the life-expectancy of the part.

So... the distributor may be the cause of the 'duds'.

Variations between manufacturers have been around for a long time, and
are a fact of life.

And it is not certain that if the max current for one manufacturer is 10
amps, and you test different manufacturer's at 10 amps, and one
manufacturer fails, that that would point at worse unreliability at say
8 amps. Especially not in the alternate manufacturer only claims 7 amps.


Thomas

 

[Spehro Pefhany]

Turns out they were a much larger and better run operation.

I meant to say they (the one we picked) was the better one. ;-)

Best regards,
Spehro Pefhany

 

[Reg Edwards]

Do a Google on "Electronic Component Reliability" FITS.

Don't forget fits.

 

[Tim Shoppa]

Hello all,


Has anyone done head to head tests of identical parts from various
manufacurers?

I'm specifically interested in power BJTs. In comparing TIP35C/TIP36C
transistors from two different manufacturers (both very well respected),
I've found that one repeatedly failed before the other (both were driven
until they failed). I did this with different batches (different date
codes) and the results were the same. I'd be particularly interested in
any information about poor quality components from Fairchild, On-Semi
and ST Microelectronics (all of which are top quality brands).

As a side note, there was one manufacturer whose transistors failed in
virtually 100% of the amplifiers that they were used in (not one of the
manufacturers listed above).

My first thought is that the design or the test, not the transistors,
is faulty

It doesn't sound like your test was against the transistor specification.
Testing past the specification isn't completely useless, but I'm not sure
that what you've done is useful.

For power components, packaging and heat-sinking issues come in as well,
and these are not necessarily as well specified. I've often found that
when part x in manufacturer's y assembly failed repeatedly, it was due to
extra stress on the leads due to poor mechanical design of the assembly...

Tim.

 

[Klaus Bahner]

The transistors are used in audio amplifiers. The amplifiers are all
burned in at full power (~400 RMS per channel into 1 ohm) until they
thermal (at ~80C). Then they are allowed to come back on (after they
cool down enough for the protection circuit to disengage) under full
load and full power. There is VERY rarely a failure when they are tested
this way. The problem comes in the field when the amplifiers are being
installed by inexperienced people. If the amplifier is playing at full
power and the speaker leads are allowed to touch (especially if they
touch right at the speaker terminals on the amplifier), the outputs
'sometimes' fail before the protection circuit and/or fuses can react.
The transistors are protected with one 15 amp fast blow fuse per three
25 amp transistors.

To make the amp less likely to fail, I tested different transistors to
determine which were most durable. Of the two brands tested, one
consistently failed first (note that the testing was done under
conditions that were far beyond that of normal operating conditions).
Both transistors had exactly the same specifications. They were in the
same style package. The heatsinking was identical. I even reversed the
location of the different brands (swapped channels) to make sure there
were no heatsinking, driver or out-of-tolerance problems.

Why would this be a useless form of testing?

As said by others, this has nothing to do with quality of a device but
is rather a problem introduced by the equpiment designer to use Reg
Edwards terms.
What you are doing is in my opinion just a work around: You have
understood the problem within your design, this is the protection unit
not fast enough to catch the short circuit situation. But instead of
curing the problem at its root, you try to find devices which for
basically ununderstood reasons survives a situation which is most likely
far beyond every data sheet specification. The reason that the
transistors from manufacture A survive could for example be a slightly
higher lead resistance limiting the peak current, which allows the fuse
to blow before the transistor is (noticeable) damaged. In my opininion
this has nothing to do with the quality or reliability of a device.
Even if the reason for devices A surviving this condition is really a
higher peak current capability than claimed in the data sheet, I would
not conclude that manufacturer A produces higher quality components than
manufacturer B - in fact, if the protection circuit would ensure that
the transistors are driven within their specs, it could very well be
that the transistors of manufacturer B are more reliable than the others

Just my two cents
Klaus

 

[Ken Finney]

Do a Google on "Electronic Component Reliability" FITS.

Don't forget fits.

There are problems with FITS, which are debated ad naseum, but one thing
that is agreed on is that FITS are only relevant before device wearout
factors start to kick in. Just two days ago, I got a chart that a very
major IC manufacuturer is going to release for one of their hi-rel
avionics-grade parts. They wanted my comments before they released it,
because they are afraid a lot of negative market reaction might take place.
The chart shows that at 120 C junction temperature, 50% of the parts will
wear out within 5 years. At 130 C, the time drops to less than 3 years.
But such is the nature of the beast when you start getting into deep
submicron.

 

[Perry Babin]

As said by others, this has nothing to do with quality of a device but
is rather a problem introduced by the equpiment designer to use Reg
Edwards terms.
What you are doing is in my opinion just a work around: You have
understood the problem within your design, this is the protection unit
not fast enough to catch the short circuit situation. But instead of
curing the problem at its root, you try to find devices which for
basically ununderstood reasons survives a situation which is most likely
far beyond every data sheet specification. The reason that the
transistors from manufacture A survive could for example be a slightly
higher lead resistance limiting the peak current, which allows the fuse
to blow before the transistor is (noticeable) damaged. In my opininion
this has nothing to do with the quality or reliability of a device.
Even if the reason for devices A surviving this condition is really a
higher peak current capability than claimed in the data sheet, I would
not conclude that manufacturer A produces higher quality components than
manufacturer B - in fact, if the protection circuit would ensure that
the transistors are driven within their specs, it could very well be
that the transistors of manufacturer B are more reliable than the others

Just my two cents
Klaus

I understand that you're questioning the quality of the protection
circuit but... I've repaired thousands (easily thousands - been in the
repair business since 1986) of amplifiers with blown output transistors.
If it were a problem of design, that means that there is no engineer in
the car amplifier industry that knows what he's doing. There is no
amplifier manufacturer that builds an amplifier that can not be damaged.
If I can employ good protection circuits AND better transistors, this
will make the amplifiers even better.

As a side note, the same manufacturer passed the short circuit AND the
stress from playing into a load.

What would be a good test to determine the better part?

PB

 

[Tim Shoppa]

The transistors are used in audio amplifiers. The amplifiers are all
burned in at full power (~400 RMS per channel into 1 ohm) until they
thermal (at ~80C). Then they are allowed to come back on (after they
cool down enough for the protection circuit to disengage) under full
load and full power. There is VERY rarely a failure when they are tested
this way. The problem comes in the field when the amplifiers are being
installed by inexperienced people. If the amplifier is playing at full
power and the speaker leads are allowed to touch (especially if they
touch right at the speaker terminals on the amplifier), the outputs
'sometimes' fail before the protection circuit and/or fuses can react.

I suspect that SOA (Safe Operating Area) limits are coming into play
in short-circuit circumstances. Look at what your two different
manufacturers publish for their SOA curves, for starters?

The transistors are protected with one 15 amp fast blow fuse per three
25 amp transistors.

Just fuses, no electronic (sense resistor) protection?

The tranistors are paralleled, is that what you're saying? What value
resistor do you use in the emitter to allow for device-to-device
gain and V_be variations?

It could be that the one brand is better "matched" and less variations
mean that they share the load better, while a different brand is less
well "matched" and they don't share the load so well. The degree of
matching could be due to distributor issues just as much as manufacturer
issues. And sometimes you just get lucky.

Tim.

 

[Jeff]

I'm sure that's the case. The amplifier has ~96 volts rail to rail and
the output impedance is essentially the resistance of a 6" piece of 14g
wire. Under short circuit conditions, the instantaneous current flow is
quite high.



In the amp used for testing, the fuses are the only overcurrent
protection. If the amplifier is simply loaded too heavily (below 1 ohm
per channel, the fuse will protect the amp every time.



The resistors are Dale 0.1 ohm 1% 3 watt components. The transistors are
always from the same batch. I never install unmatched (from different
batches/date codes) transistors if they are going to be run in parallel.
In testing, I tested with single transistors and with all transistors (3
in parallel) in place.



You're probably right. I usually assume that the components from a
particular batch will be matched (very close Vbe). In the past, the
components with the same date code were very close but that may not be
the case now. I'll have to start checking them again to confirm.

Just to add:
If these are japanise transistors the markings on them
include the Beta grouping and are usually well matched
by the markings.
I have not found this to be true with other area made
transistors, (american or european) they will require
"hand selecting" curve trace matching in many applications where more than 1
or 2 are used, audio and high current
are the first I can think of.
Also some manufactures like to design around the specs
of a part to such a precision that using a substitute part
may require a little redesign.
Jeff

 

[N. Thornton]

The transistors are protected with one 15 amp fast blow fuse per three

In the amp used for testing, the fuses are the only overcurrent
protection.

Hi Perry. This makes it clear that the design is defective. If a fuse
is the only protection the s/c current is going to be very high, well
outside the capabilities of the transistors. So your amp will be
vulnerable to destruction by shorting - no surprise.

Solution is put safe area protection in. If you dont, your going to
get a failure rate.

I gather you're trying to cut corners by putting trs in that do
survive this abuse. But that survival at way over spec currents is
anything but guaranteed. One batch change, manufacturer change etc,
and you could end up with piles of failures. Which I gather is exactly
whats happening now.


Regards, NT

 

[Dan Fraser]

Also be aware that there are counterfeit transistors out there.
Specially with high power audio output transistors. We got bitten once
with counterfeit Toshiba transistors. When we opened the duds, we found
maybe a 6A die where a 30A one should have been. We found he fakes did
not have the BVceo of the real thing. Counterfeits are a growing
problem. If the price seems to good to be true, it often is.

No major maker, after the nearly billion dollars it takes to make good
semis is going to risk it all on cheating the specs. However, there are
third world pirates who will take reject silicon and package it in
anything that looks good on the outside and sell it.

This is destroying the non-OEM brokers who were so useful in helping us
though spot shortages but what can you do besides buy from authorized
distributors only.

--
Dan Fraser

From Costa Mesa in sunny California
949-631-7535 Cell 714-420-7535

Check out my electronic schematics site at: http://www.schematicsforfree.com
If you are into cars check out www.roadsters.com

 

[David Harmon]

This is destroying the non-OEM brokers who were so useful in helping us
though spot shortages but what can you do besides buy from authorized
distributors only.

I don't know, but how much would it cost the broker to decapsulate
the part, take a photograph of the die, and look for the original
insignia?