Systemic IC failures

[dexters_lab]

So in my adventures fixing up various vintage quantel products i am coming across a particular IC that always seems to fail. From what i understand the failure also happens in other equipment so it's not like quantel were using this out of spec. The particular ic in question is just known to 'go bad'

In this case the IC is a ST Micro STV1601 and it's partner the STV1602, they are a SDI (serial digital video) serialiser and de-serialiser pair and were manufactured in the mid 90s to sometime in the early 00s. The 1601 takes 10 bit parallel data and converts it into serial digital video and the 1602 does it the other way around. Quantel used them for SDI video input/output on their broadcast equipment.

I understand they were originally manufactured by Sony as the SBX160x and these were even more notorious for going bad but at some date the Sony parts seem to disappear and be replaced by ST Micro branded parts and the device name changed to STV160x. It also looks like manufacturing switched from Japan to France.

Some people have said it's heat related, but i have just been through some of my spare new-old-stock quantel boards that use this IC and all of the STV1601 devices seem to be dead even though they don't appear to have been used, so it could even be an age thing. Interestingly the 1601 part seems to be much worse than the 1602 and quantel in later versions of their products switched from using the 1601 to a entirely different Cypress part but retained the 1602 indicating the 1601 is particularly troublesome.

So out of curiosity i am just asking what are some of the causes in systemic failures in a particular ic like this, why one would be worse than the other and how it can continue to be manufactured with what is clearly some kind of flaw? heat, age, poor design, poor manufacturing? 

The STV160x devices are a ceramic PGA device as pictured below.

[max_torque]

Guess you'd have to work out what actually fails in the IC first?

[amyk]

What happens when they fail? Does it completely fail to give any output, are bits stuck high/low, etc.? Looking at the datasheet and die it seems to be a relatively simple and large-featured IC so you may be able to reverse-engineer it easily. It might just be dirt but I see a black spot slightly above and to the right of that die.

[philpem]

I'd go with the usual culprits initially - electrostatic discharge is a possibility.

With that said - if NOS parts which have been stored correctly are failing, I'd be looking at failures in bondwires, failures inside the IC (that broken-open Sony part looks more like an encapsulated hybrid), and so on.
Humidity might be another possibility if the ceramic package isn't quite sealed.

Some Tektronix hybrids had issues which could be resolved to an extent - I can't remember how though!

[Wimberleytech]

A look at the datasheet indicates a bipolar process, so ESD is less likely. 

Something physical like bond wires...maybe.

Extremely difficult to know without doing FA.

[dexters_lab]

What happens when they fail? Does it completely fail to give any output, are bits stuck high/low, etc.? Looking at the datasheet and die it seems to be a relatively simple and large-featured IC so you may be able to reverse-engineer it easily. It might just be dirt but I see a black spot slightly above and to the right of that die.

they seem to slowly die... you'll start getting random sparkly pixels on the video that gets worse, then glitchy video, loss of sync, then eventually nothing

A look at the datasheet indicates a bipolar process, so ESD is less likely. 

Something physical like bond wires...maybe.

Extremely difficult to know without doing FA.

yea hard to tell and tbh i am not expecting to find out exactly what the cause is, i am just interested to know that if you take ESD and mechanical issues out of the equation how an IC can die ('scuse the pun!)

[Alex Eisenhut]

Purple plague? Bad passivation?

https://en.wikipedia.org/wiki/Gold-aluminium_intermetallic

[CatalinaWOW]

Might just be a bad design for one or two key paths in the metalization.  Current density right at or just above the limit.  NOS has a significant failure rate.  Same thing might have happened to the OEMs using these chips.  So the failures you see in the field are the survivors, the ones that were good enough to make it through test, but still not quite good enough to go forever.   Low yield is never desired, but it happens.

[james_s]

I recently worked on an Asteroids Deluxe board that had 3 or 4 bad 74LS157's, nothing else. In the process I noticed there were 2 more that somebody had replaced previously. The ones that were present were all the same brand and were spread across different sections of the board. Suspecting a manufacturing flaw, I replaced all of the '157's on the board. I don't know what the actual cause was but I would guess some form of contamination or moisture ingress.

[dexters_lab]

i can understand how you can have bad batches that fail, but this device seems to be sketchy throughout it's entire production life

[james_s]

Perhaps it was just flawed by design? Some aspect of it may have been pushing the limits of the available process? It would be interesting to do a detailed failure analysis on a significant sample of defective parts but I certainly lack the resources to do so.

Is there anything semi-modern that can be used to replace these?

[NiHaoMike]

What about a PCB to adapt a modern chip?

[dexters_lab]

There's nothing available that's pin compatible but there are other SDI ICs that do the same job

On this system the STV1601s are mounted on small daughter boards which also contain the SDI cable driver IC and some interface logic to mate with quantel's internal video, although we don't have the schematic for the actual SDI daughter board (it's basically the STV1601 reference design + an altera EPLD anyway) we do have the schematic for the whole video output board that these SDI output modules plug onto. It looks fairly simple so we're having a play around with some FPGAs which could allow us to create the SDI output. As the modules are plug-in it could allow a replacement board to be made.

Doing the same on the video input board might also be an option as they are very similar.

For the curious i have attached one of the video output boards, this one dates from 1994 with the SDI module highlighted in green, there are two STV1601s as one does the 'key' channel

[bd139]

Interesting. Back in the late 1990s I saw a lot of PPC 603e failures in an embedded system which I had the joy of being the stats monkey on. They had the same type of package. They mostly failed during environmental testing (vibration) and in the field after a few weeks. Test cases would start to intermittently fail. Not sure what was done to resolve it as I quit from boredom. Correlation?

[John Heath]

Fast processing comes to mind with video. Fast usually means running hot. Does it run hot? Maybe it needs a heat sink to be a happy camper.

[mikerj]

Fast processing comes to mind with video. Fast usually means running hot. Does it run hot? Maybe it needs a heat sink to be a happy camper.

That wouldn't explain why new old stock parts are dead.  My guess would be corrosion failure, either caused by contaminants on the die at manufacture or moisture ingress from insufficient package hermeticity.

It's an interesting look back in packaging though, it's a huge device considering the size of the die.

[T3sl4co1l]

Corrosion (ingress of atmosphere, including gaseous diffusion) and intermetallic formation (also diffusion mediated) would seem likely.  Diffusion does go faster at elevated temperatures; is there correlation between failure rates and high operating or storage temperature?

Everything else can seem to be ruled out, as stored parts failed.  ESD wouldn't fail whole lots, electromigration, overheating and other operating factors are N/A, etc.

What about building a riser board to replace the PGA itself?  If compatible parts (but not footprints) are available, that would seem almost trivial.  Populate a small board with press-fit pins, and there you go.

Tim

[james_s]

Interesting. Back in the late 1990s I saw a lot of PPC 603e failures in an embedded system which I had the joy of being the stats monkey on. They had the same type of package. They mostly failed during environmental testing (vibration) and in the field after a few weeks. Test cases would start to intermittently fail. Not sure what was done to resolve it as I quit from boredom. Correlation?

Perhaps they were as bored as you were and quit for the same reason 

[dexters_lab]

they probably were fast devices for the time, they run from a 27mhz clock which is PLL'ed 10x to 270mhz to get the 270mbps for SD-SDI

they are supposed to dissipate 2 watts, but i dont think that's wholly accurate as some of them gett 'effin hot, i put heatsinks on all mine that are still working

[dexters_lab]

is there correlation between failure rates and high operating or storage temperature?

i believe so yes at least for the operating temperature

[james_s]

Corrosion (ingress of atmosphere, including gaseous diffusion) and intermetallic formation (also diffusion mediated) would seem likely.  Diffusion does go faster at elevated temperatures; is there correlation between failure rates and high operating or storage temperature?

Everything else can seem to be ruled out, as stored parts failed.  ESD wouldn't fail whole lots, electromigration, overheating and other operating factors are N/A, etc.

What about building a riser board to replace the PGA itself?  If compatible parts (but not footprints) are available, that would seem almost trivial.  Populate a small board with press-fit pins, and there you go.

Tim

This does seem likely. I do not have a background in IC fabrication but I recall reading somewhere about the process of prototyping one of the later Apple II series that used ASICs to replace some of the discrete logic on the motherboard. As I recall some of these prototype ASICs had no lids for whatever reason and the engineer mentioned that they stopped working after a few weeks due to exposure to the environment. Prior to that I hadn't realized that IC dies would fail if exposed to the air but it makes sense. There was a similar problem with many Bosch ECUs in the late 80s-early 90s that used ceramic hybrids coated with a clear goo. They failed in large numbers after about 20 years, people suspect due to moisture diffusing through the protective coating and contaminating the IC dies.

[GeorgeOfTheJungle]

Yes these Bosch hybrids with goo fail very often

[james_s]

That looks quite a bit higher density, the ones I dealt with were LH-Jetronic, always thought about reverse engineering one of the hybrids but never got around to it.

[John Heath]

That is an interesting hybrid. Half IC half PC board. There was one of those in my car inside the distributor cap , honda accord. It failed and I had to change it. Their little hybrid module was 150 bucks and the distributor cap assembly  including hybrid module was also 150 bucks?? Did not have to
think too hard on which one to buy , ha.  Cool eagle by the way.

[floobydust]

I had field failures with TI CD4000 series parts, after a lot of hassle it was traced back to TI's manufacturing changes in the DIP package. Bonding wires were failing. Their new plastic DIP package allows more humidity and flux/wash ingress.

TI blamed it on PCB "aggressive" flux ingress corroding the bonding wires. Had to use TEM to resolve it.
SEM Lab examples