Author Topic: Is "integrated circuit burn-in" a thing?  (Read 6185 times)

0 Members and 1 Guest are viewing this topic.

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #25 on: April 24, 2019, 10:38:45 pm »
Is there anything in the industry about IC burn-in? Perhaps specific to MEMS devices in particular?
Burn-in is a very expensive process so only used on parts which can support the cost.
I didn't mean burn-in at the foundry. I meant after assembly. Maybe "conditioning" is a better term... basically some period of time with power applied during which "errors" aren't to be considered fatal, while the component conditions or ages or settles or any number of other similar terms. In discrete electronics this wasn't uncommon, you'd allow freshly assembled units to run with power applied for some number of hours before bothering to calibrate them for the first time (again, to be clear, I'm not talking about warm-up after each power application, I mean the initial application of power). This was because many types of components tended to "drift" at first, then settle into a long term stable condition after being powered up for a while. The amount of change quickly went asymptotic after some number of minutes/hours and then could be relied upon to be stable thereafter even after repeated power cycles.

I've never heard of that for IC's, but these parts sure act like it and it would explain what I'm seeing.
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #26 on: April 24, 2019, 10:40:00 pm »
Even so, burn-in is normally used to detect early failures, not to somehow magically make faulty parts start working. I'm not aware of any physical mechanism which would cause a damaged part to start working just because it's been switched on for a while.
I haven't heard of one either, hence this thread.
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #27 on: April 24, 2019, 10:41:40 pm »
Interesting thought on the moisture sensitivity. I wonder if being powered on for a while would bake excess moisture out (for that matter, I wonder if the package is hermetic). If the problem is solely moisture interfering with a good SMT reflow, that is certainly not going to improve with power-on time. I'll confirm with the assembly shop and report back. Thanks!
I received their response: Sealed packages from DigiKey in this case, a reputable supplier. And they didn't notice anything unusual when opening the packages nor loading the reels into the pick-and-place machines either.
 

Offline m98

  • Frequent Contributor
  • **
  • Posts: 634
  • Country: de
Re: Is "integrated circuit burn-in" a thing?
« Reply #28 on: April 25, 2019, 08:06:50 am »
I was just about to design an ST accelerometer into a product. Guess I will use an alternative part instead.
There is a recovery period for some parts, some hygrometers and gas sensors for example only perform to their specs a week after soldering.
 

Online SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15796
  • Country: fr
Re: Is "integrated circuit burn-in" a thing?
« Reply #29 on: April 25, 2019, 02:07:13 pm »
Burn-in of active parts for high reliability devices is certainly a thing. It's meant to help getting rid of early failures.

In your case, I'd venture there is probably something very wrong that is way beyond the scope of burn-in. Failures in a matter of a couple hours in normal conditions with several devices doesn't look right. So either the batch of ICs you got is all faulty, or your design is? I never saw such a problem with MEMS accelerometers. Admittedly I only ever used AD and TE accelerometers, but ST ones can be found in many mobile devices. You should probably contact ST for support.



 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #30 on: April 25, 2019, 03:09:53 pm »
In your case, I'd venture there is probably something very wrong that is way beyond the scope of burn-in. Failures in a matter of a couple hours in normal conditions with several devices doesn't look right. So either the batch of ICs you got is all faulty, or your design is?
The evidence is mounting that this is an ST problem. I'm up to ~72 continuous hours on some BMI160's with zero issues (just scoped them again before typing this). That's the exact same PCB, exact same BOM, assembled at the same time as the LSM6DS3 boards, same same same EXCEPT for which part was stuffed in that footprint. As noted earlier in this thread, those two devices happen to share a compatible footprint so we can build boards that are identical in every way except for which device is stuffed into location U7. The LSM demonstrates this weird "burn in" behavior, while the BMI appears not to.

Quote
You should probably contact ST for support.
:-DD :-DD :-DD  |O |O |O See my comments about ST support earlier in this thread....  :rant: :horse:
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #31 on: April 25, 2019, 03:18:34 pm »
I've just remembered one "difference" between the two devices: Their "typical" supply voltage.

The LSM has supply specs of 1.71V (min), 1.8V (typ), 3.6V (max). The BMI has supply specs of 1.71V (min), 3.0V (typ), and 3.6V (max). We run both parts on a 2V5 rail that was already in this board's previous design. I personally contacted both vendors and asked them if their 1V8/3V0 was actually preferred, or if 2V5 (which is obviously well within both device's specs) had any disadvantages compared to their "typical" supply voltage value. Both companies replied and explicitly said that 2V5 was perfectly fine and that voltage had no disadvantage compared to the "typical" values they listed in their spec sheets.

I don't think that could cause this LSM problem, but I did notice it and got the manufacturer's opinion before committing. That said, 2V5 isn't their "typical"  value for either part and is 50% higher than the LSM's "typical" while still comfortably within its acceptable range.

Just tossing that out there in case it rings a bell for anyone. Again, I can't imagine how a part would stop locking up after it "grew accustomed" to a different supply voltage....  :o
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #32 on: April 25, 2019, 03:26:36 pm »
I was just about to design an ST accelerometer into a product. Guess I will use an alternative part instead.
The Bosch-Sensortec BMI160 (compatible footprint, though not firmware compatible) is looking good electrically. However, it appears less tight on its axes-to-package physical alignment. I'm seeing misalignments of 2+ degrees on some BMI160's. Your downstream firmware can compensate for this by doing the appropriate 3D coordinate rotations, but it's something to keep in mind.

Some earlier MEMS devices included guaranteed specs for the alignment of their die axes with those of the package. I've noticed that particular spec has disappeared from most spec sheets now. My guess is it reduces production cost for them by not having to individually characterize every device, and they probably assume most installations are doing some sort of in-place calibration anyway so why duplicate that. The downside is that with these earlier devices their alignment was often close enough for many non-critical applications as long as your pick-and-place equipment did a good job, but now it appears you cannot rely on that unless +/-2 degrees of misalignment doesn't matter. Maybe it doesn't for game controllers or something, but in our case it DEFINITELY matters. This had me originally favoring the LSM but given that part's apparent reliability issues the BMI is looking better every day!
 

Offline SilverSolder

  • Super Contributor
  • ***
  • Posts: 6126
  • Country: 00
Re: Is "integrated circuit burn-in" a thing?
« Reply #33 on: April 25, 2019, 05:33:16 pm »

Could it be that you were just unlucky, and if you bought some more parts they would behave better?

 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2851
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #34 on: April 25, 2019, 07:41:59 pm »
The evidence is mounting that this is an ST problem. I'm up to ~72 continuous hours on some BMI160's with zero issues (just scoped them again before typing this).
Do ALL ST devices go through this issue?  If so, I'd suggest getting a few new units and hand-soldering new parts on your boards and repeat the test.
It may be that the stresses of reflow soldering in their process may be exceeding what the chips can handle.  If you can hand-solder the ST parts in and they are fine, it has to be something to do with the soldering process (or maybe even the P&P machine is shocking the part when planting it on the PCB.)

Jon
 

Offline AndyC_772

  • Super Contributor
  • ***
  • Posts: 4315
  • Country: gb
  • Professional design engineer
    • Cawte Engineering | Reliable Electronics
Re: Is "integrated circuit burn-in" a thing?
« Reply #35 on: April 25, 2019, 07:57:21 pm »
Hand soldering imposes far greater thermal stresses on components than a correct reflow profile.

I've wrecked several ST accelerometers this way. IIRC they were the first parts I ever had to learn to reflow using a preheater and hot air station, and once I'd figured out how to minimise the peak temperature they ever experienced, they worked fine.

Offline Berni

  • Super Contributor
  • ***
  • Posts: 5050
  • Country: si
Re: Is "integrated circuit burn-in" a thing?
« Reply #36 on: April 25, 2019, 08:32:30 pm »
That is a very weird problem there.

I wouldn't really expect any failure in the MEMS structure to actually kill the SPI communication. I seen weird behavior of some MEMS IMUs due to a floating pin, funny enough the problem on that one only seamed to appear when it was at below freezing temperatures. I suppose some noise or bad decoupling could cause the internal CPU to rarely misstep and crash, or perhaps the timings,levels or rise times on some of its pins are capable of putting its internal logic into a weird meta stable state that eventually gets it to crash.

As for explaining the "burn in" period that can possibly be caused by soldering flux that was not properly cleaned from the board. It very slowly absorbs moisture from the air, causing its conductivity to drift and making for seemingly time related behavior. Tho these resistances tend to be so high that it takes a floating pin as mentioned above to detect it. It usually takes a very heavy layer of flux and lots of moisture for it to become conductive enough to mess up an actively driven digital signal (But i seen it happen).

We also found that ultrasonic cleaners have the ability to kill MEMS devices and barometers
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7678
  • Country: ca
Re: Is "integrated circuit burn-in" a thing?
« Reply #37 on: April 25, 2019, 08:59:59 pm »
What kind of flux, pcb wash and dry are you doing? That can cause problems with moisture ingress and your "burn-in" is just waiting for things to dry out.
The reflow profile has to be decent, not sloppy with too much heat or too long. Excessive PCB flex during handling can also cause shifts.

Best to be careful and use scientific method (change only one thing at a time) to troubleshoot your problem.
Encountering hassles now is much better than when in production...
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #38 on: April 25, 2019, 10:22:45 pm »
Could it be that you were just unlucky, and if you bought some more parts they would behave better?
Maybe... but how many more before they're "trustworthy"? If I have 50 bad parts now, and 50 more all work great, that's a 50% reliability rate. I wouldn't design with a part that is only 50% reliable.
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2851
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #39 on: April 25, 2019, 10:24:19 pm »
Hand soldering imposes far greater thermal stresses on components than a correct reflow profile.

I've wrecked several ST accelerometers this way. IIRC they were the first parts I ever had to learn to reflow using a preheater and hot air station, and once I'd figured out how to minimise the peak temperature they ever experienced, they worked fine.
Maybe, maybe not!  Yes, what you say is possible.  But, if you solder one pin at a time, for just a second or so, the BODY of the part never gets heated more than 10 C above ambient.  When reflowing, the whole component is heated to 200C+ above ambient.  Maybe this part is really sensitive to that internal temperature.

OK, you say you had to use a hot air station, that concentrates the heat on the LEADS, not the component body.  That's what I'm talking about.  My guess is the OP's contract assembler used ordinary reflow, heating EVERYTHING to 250 C for a minute or more.

Jon
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #40 on: April 25, 2019, 10:26:19 pm »
Best to be careful and use scientific method (change only one thing at a time) to troubleshoot your problem. Encountering hassles now is much better than when in production...
Amen to that! Which is why I'm digging into it now, rather than later.

The bottom line for me, so far, is that the ST part demonstrates this burn-in behavior while the Bosch-Sensortec part does not. Literally everything else is the same, they were installed on the same group of panelized-then-separated PCB's, all other components are identical and from the same reels, stuffed and reflowed at the same time, etc. There just aren't that many variables to play with, especially since "the other part" doesn't demonstrate the behavior.

It's aggravating.
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #41 on: April 25, 2019, 10:29:00 pm »
My guess is the OP's contract assembler used ordinary reflow, heating EVERYTHING to 250 C for a minute or more.
That's exactly correct. However, they are very good at what they do. In literally tens of thousands of units they've done for us over the past 3-4 years I've never had a reflow problem, and never seen a problem like this one. That's why I'm so focused on the part, because it's such an outlier and only ONE part is exhibiting weirdness while its second source does not.
 

Offline amyk

  • Super Contributor
  • ***
  • Posts: 8526
Re: Is "integrated circuit burn-in" a thing?
« Reply #42 on: April 26, 2019, 12:04:04 am »
The "typ" values in datasheets are usually chosen to demonstrate the best performance in one of several areas (such as power consumption), but as long as the parameter is within limits it shouldn't matter.

I recommend also the "buy several and solder them yourself" test, if you could also get a sample from what your assembler was using then that's even better.

This part has breakout boards available and is popular enough to be "Arduino-ised", if there was such a weird problem with it we would've heard many many times from others already.
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #43 on: April 26, 2019, 12:48:44 am »
This part has breakout boards available and is popular enough to be "Arduino-ised", if there was such a weird problem with it we would've heard many many times from others already.
That's what I felt as well. Which is why one of my theories (noted earlier in this thread) is that the part doesn't "sit still" very well. My long term tests have been unusual in the sense that most people don't just power up a MEMS motion sensor and then let it sit motionless for dozens of hours. Also, so far (fingers crossed) the prototypes we've send into the field haven't reported this error... and they don't sit motionless either. I'm thinking of building some sort of motion table and letting a long term test run on that, to test this theory, but not sure I have the time to spare anytime soon.

EDIT: All of the protos in the field came from the same, single build as the ones I'm using on the bench. So again, no variables there. Just a different use case, mine can sit motionless for hours at a time while those are in constant motion when powered up. Hence my theory....
« Last Edit: April 26, 2019, 12:50:24 am by IDEngineer »
 

Offline AndyC_772

  • Super Contributor
  • ***
  • Posts: 4315
  • Country: gb
  • Professional design engineer
    • Cawte Engineering | Reliable Electronics
Re: Is "integrated circuit burn-in" a thing?
« Reply #44 on: April 26, 2019, 05:56:22 am »
Maybe, maybe not!  Yes, what you say is possible.  But, if you solder one pin at a time, for just a second or so, the BODY of the part never gets heated more than 10 C above ambient.  When reflowing, the whole component is heated to 200C+ above ambient.  Maybe this part is really sensitive to that internal temperature

Every surface mount component you can buy is designed, tested and rated for reflow soldering.

These parts are tiny (~4x4mm) QFN packages; you're not going to quickly solder the pins one at a time by hand without heat conducting into the bulk of the package.

Offline SilverSolder

  • Super Contributor
  • ***
  • Posts: 6126
  • Country: 00
Re: Is "integrated circuit burn-in" a thing?
« Reply #45 on: April 26, 2019, 04:12:38 pm »


If the part is available as an Arduino expansion board, maybe it is something to try? (just to see if a different process makes a difference)
 

Offline amyk

  • Super Contributor
  • ***
  • Posts: 8526
Re: Is "integrated circuit burn-in" a thing?
« Reply #46 on: April 28, 2019, 02:04:49 am »
Which is why one of my theories (noted earlier in this thread) is that the part doesn't "sit still" very well. My long term tests have been unusual in the sense that most people don't just power up a MEMS motion sensor and then let it sit motionless for dozens of hours. Also, so far (fingers crossed) the prototypes we've send into the field haven't reported this error... and they don't sit motionless either. I'm thinking of building some sort of motion table and letting a long term test run on that, to test this theory, but not sure I have the time to spare anytime soon.
"doesn't sit still" is an interesting theory, but perhaps more interestingly, what orientation are they mounted in when they're sitting still?
 

Offline IDEngineerTopic starter

  • Super Contributor
  • ***
  • Posts: 1950
  • Country: us
Re: Is "integrated circuit burn-in" a thing?
« Reply #47 on: April 28, 2019, 02:24:31 am »
"doesn't sit still" is an interesting theory, but perhaps more interestingly, what orientation are they mounted in when they're sitting still?
Various. They're mounted on a 2x2in PCB. If it's one I'm actively working on I have a little PCB holding jig that allows me to rotate and freeze the board in convenient orientations. Otherwise the boards are plugged into test harnesses, left on the bench, and the PCB's just sit in whatever orientation the harness wiring twist happens to settle into. There's no "normal" orientation, I'd bet no two of them have been in the same orientation when they've locked up. The wiring harness twist and weight is more forceful than the weight of the PCB's so the boards can't force themselves into some sort of repeatable orientation.

EDIT: But once the wiring harness "settles" each PCB is basically motionless in that (random) orientation because I'm not touching it and nothing moves the bench.
« Last Edit: April 28, 2019, 02:27:30 am by IDEngineer »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf