Physical age vs electrical age of component and counterfeit detection

This is strictly a question to provoke a discussion and not necessarily generate a solution. (and who knows, if you come up with the solution, you patent it, and retire to a beach of your choosing)

With the ever increasing number of counterfeit components entering into supply chains, is it possible to design/build in a circuit that would "age" or "degrade" based on the time a component is actually powered on? If some quantity could be measured with a relatively simple definitive test, one could then determine the electrical age of the component, thus knowing I am not getting an unused component, but one that may have been stripped off another circuit board and resold.

Obviously components marked with a date code gives us some idea as to the physical age of the part in question.

Your title is misleading... those devices are not counterfeit, they are pulls.

And there is no "clock" that measures on-time on a device.

thorneko said:

is it possible to design/build in a circuit that would "age" or "degrade" based on the time a component is actually powered on?

Click to expand...

Perhaps the mistake on this question is the wrong assumption that the running time is the only factor to derate a device. Actually, the long term exposition to higher temperatures affect much more the component integrity.

You can embed hot-carrier-stressed devices and sense a drift
vs a lesser- or un-stressed device. But this has a lot of
variability with lot processing and with temperature history
(baking out to some extent and accruing "hot" carriers more
at lower temps). If you had NVM you could even put an
"odometer" in there and store cycle counts. Clocks, or
power-up cycles, could be set after some "hall pass" period
that accommodates test (and potentially burnin?) so that
a part on the loading dock is OK, but a part that's been
used to any significant fraction of its advertised life will
have a "mystery bit" set somewhere only the ATE knows
where to look.

But all of this pales in comparison to the threat of "walking
wounded" EOS'd or ESD'd devices which a crude board strip
might leave you with, irrespective of time in service or any
such measures. A time=0 reject from a "porous" third party
test floor can also find its way back into circulation. What
would such methods have to offer, for that?

Also, time in service at room temp "ages" most parts applied
per limits, not at all. Dielectric wearout and interconnect
electromigration are strongly temperature accelerated and
a part designed for 10 years at 85C should be good for
hundreds of years at 25C (depending on the actual Ea
and exponents for the materials and processing at hand).

And by implementing such measures, you had best not
compromise the usefulness of the part for someone who
really wants to make a durable product. A "test flag" is
one thing, but a flag that causes misbehavior is another.

andre_teprom said:

Perhaps the mistake on this question is the wrong assumption that the running time is the only factor to derate a device. Actually, the long term exposition to higher temperatures affect much more the component integrity.

Click to expand...

Yes, I'm not disagreeing with this point. I do test engineering for a living, testing all sorts of components across temperature.

Fault of every engineer everywhere. Going straight to the details while ignoring the big picture.

Step back and look at the big picture for a moment.

I buy a packaged part from a vendor.
Now this part is either "counterfeit" or a genuine part. A counterfeit part could fall into many categories.
Off the top of my head... 1) different silicon inside, but same functionality 2) part with a new label (say a commercial part upgraded to industrial), 3) a pull (part has been used before).

How much electrical on time would you expect a new packaged part to have? I'd say less than an hour. x amount of time to test at wafer sort. y amount of time to test the final packaged part.

Especially in case 3 above, the component in question if it has been used before, will most likely have been run for at lease 1 hour electrically. This could also apply in case 2 if it fell into the category of 3.

Hence my original question, what is the best method to determine the electrical age of the part. Anything over an hour (even taking into consideration usage at temperature, etc...) would indicate the part has been most likely used before, and is therefore unable to be qualified as a genuine "brand-new" component.

Clear as mud, right?

Why are you using a vendor who isn't trusted enough to give you true, genuine parts?

The idea of embed an 'odometer' is worthy of consideration, because sometimes in the lack of sufficient stock on trusted suppliers we have to resort to second line suppliers. Anyway, considering that most modern components are manufactured in SMD packages, it would be hard to reuse them without leaving some evidence of that.

SLK001 said:

Why are you using a vendor who isn't trusted enough to give you true, genuine parts?

Click to expand...

Thinking like an engineer again. It is not a perfect world, parts as I describe do enter into the supply chain, "trusted" vendor or not.

thorneko said:

Thinking like an engineer again. It is not a perfect world, parts as I describe do enter into the supply chain, "trusted" vendor or not.

Click to expand...

I ran across an article about this subject of the supply chain and fake/repackaged parts (written by some supply chain consortium). The gist of the article stated that you should always use authorized distributors for components (not just trusted, who says they can be trusted?) as the authorized distributors either send you parts from their inventory (inventory was directly supplied by the manufacturing plant) or they order those parts from the manufacturer and then they send them to you.

The point they were making was it is important for them to keep any chance of fake/repackaged parts from entering the distributor supply chain as that would compromise any "trusted" status they have. There's little chance of ever seeing grey market parts coming from the likes of Avnet, Arrow, etc.

I just came across this, so it is still considered a problem.
https://www.govtech.com/policy/Cybersecurity-Legislation-Seeks-to-Ban-E-Waste-to-China.html

So far the odometer idea seems the best of the ideas presented, but ideally a pass/fail flag seems like the best solution for chips with low pin counts.
Wafer sort results in 1 electrical power-up. Packaged part testing results in 2 electrical power-ups. Burn-in requirement makes 3. Counter increases on each power down? Ideally then the customer would read a register (odometer) at power-up and at the most based on component history see x in his odometer. This is a workable idea. I also like the idea of a measurable quantity (such as resistance of a material) that undergoes degradation that correlates to the number of hours of actual electrical operation.

The link you supplied makes no mention about how the parts got into the supply chain in the first place. Unless the place has stringent requirements on buying from authorized distributors only, then due to extreme schedule pressure I've seen buyers purchase parts from basically the 2nd tier of distributors (which don't necessarily have any tractability of their parts back to the factory), or the grey market (no guarantee of being anything but bad/fake/reused parts).

Parts that have been rejected, have been stored incorrectly, etc that are then disposed of end up in the supply chain of these non top tier distributors regularly from what I've read on the subject previously.

Your "odometer" idea would only work to identify pulls. Counterfeits would also have your "odometer" that would register in the range that you would like, so this idea won't do what you want. Bad parts will also have your "odometer", so what good is it?

Also, you are overlooking the fact that an odometer will take up silicon and thus will cost more than a manufacturer may be willing to shoulder. I maintain that if you want good parts, purchase from authorized dealers or manufacturers.

All the discussion brought me something to mind: It would be costly for uC manufacturers to include some read-only register on which could be recorded either the S/N or Lot number ? At least this way someone could to know the birthday of the part...

For example, in the case of the counterfeit memory chips sold to L-3 Communications, the supplier in China selected and sent L-3 Communications’ U.S.-based distributor a sample of 18 parts to test. Once those parts were tested and validated as authentic, the China-based supplier sold the company more than ten thousand of the chips. L-3’s process at the time allowed the company to accept those chips without additional testing from an independent laboratory.

THe Broker Association for Electronic Parts in the USA https://www.erai.com/ is well aware of these practices. THe brokers are poor at Quality inspection expertise to verify if qualified samples match the follow-on shipments. THese would have been new parts, most likely new parts as used are easily identified by solder plating marks on leads.

Identifying unknown parts from valid OEM requires Visual, XRAY and ATE testing on every part to be 100% certain.

There is no method to determine proof of aging. ( like carbon dating)
Any company found to ship fake parts is often disqualified immediately even if Certs of authenticity were fake provided to them.

Having been personally burned twice by counterfeit devices, I can agree that the best solution is to purchase from reputable distributors exclusively.
Sometimes it is a hassle, since on many instances the legit components have long delivery lead times. So your company must have a good planning/scheduling department to plan ahead with orders.

And it is not only ICs or expensive semiconductors. One of the times it was a fake tantalum capacitor which had the correct capacitance but the incorrect voltage rating (it was marked with a higher voltage).

Unfortunately, these issues usually arise because the parts are obsolete, (in some cases poor planning and long lead time, but not the military) hence surplus market must be used supported only by Brokers, not franchised distributors, hence, the situation is unavoidable, and more scrutiny and penalties are needed. Buyers beware.