[pjc50]

It's (speculated to be) flexing PCBs under BGA chips again.

This is a serious problem when making devices thinner: they become more flexible, but the joints are not flexible, so after a while you get a crack all the way across a joint and it either becomes intermittent or capacitively coupled.

A decade ago I had a white plastic-bodied Macbook which developed a similar fault in the graphics. But it's not just Apple, the famous "red ring of death" was a similar problem induced by thermal cycling rather than physical bending.

[rwmj]

It's reminiscent of the iBook G4 problems with cracking RoHS solder (caused by heat, not bending). Apple denied up them until the National Consumer Agency of Denmark investigated and produced this excellent report:

https://web.archive.org/web/20071021235338/http://www.forbru...

It happened to my computer. Apple delayed and denied and charged me, but wouldn't fix the problem. Eventually I gave up, but I've never bought an Apple device since that failure.

[abakker]

Thermal cycling issue in macbook pro 17, 2011. After 3 years, I had to put the logic board into the oven for 10 minutes, ever month. After a year of that apple finally issued a service bulletin and replaced that board.

[chenster]

I used a hair dryer on high and that worked really well until Apple issues official replacement.

[Mtinie]

Presuming that manufacturers aren't about to stop trying to make thinner and thinner devices...

Is this class of issues correctable (to a degree) with a change in the design of those components -- for example, proactively splitting the board at the point that the fissure is likely to appear and coupling it in a flexible way -- or is this a materials science concern where we need to find new methods to build the components' substrates so they are innately flexible?

[pjc50]

You can't split the board, it's the solder joints on the surface that are the problem. So far as I know there is no flexible soldering technology, and creating one would be pretty extreme materials science.

(To see the problem, place a coin on a credit card and bend the card - note that the coin no longer touches across its whole surface)

The normal solution is to make the PCB stiffer (thicker, or invent something better than FR4), or to make the overall casing stiffer. Either by changing materials or changing the aspect ratio. Fundamentally a long flat thin object is going to be bendy or brittle. The older iPhones that were smaller with glass front and back were an extremely good design from this point of view.

[kalleboo]

> Fundamentally a long flat thin object is going to be bendy or brittle

I think GPs idea was to make the PCB no longer long, but have two smaller PCBs with flat flex between them. Then when the phone bends, the PCBs wont bend but let the flat flex take the stress instead.

> (To see the problem, place a coin on a credit card and bend the card - note that the coin no longer touches across its whole surface)

Now cut the card in half and connect it with half a cm of tape and bend it - the two halves will be flat and you'll get a V shape instead. Put the coin on one side and no problem.

I guess it depends on how much open space is in the phone or if the whole PCB is completely flush with the bent casing.

[pjc50]

Doable, and might even be a good idea for long thin PCBs, but it messes with both your routing and placement and therefore consumes PCB area. Also upsets your controlled-impedance traces. And adds an extra assembly step. To estimate doing this, take the PCB image and Paint and try to clear a 3mm gap in it. How many components do you have to move? Do you have to make the whole thing 3mm longer? Aren't most of them decoupling capacitors that must be kept right next to their corresponding IC?

(Look at the smallest .04 x .02 resistors on there!)

More cynically, the PCB only needs to be made as bend-resistant as the display: you don't care about your BGAs if you've cracked the screen.

The article does mention a fix the repairers have been applying, involving the "sticker shield" - not shown in the hero image, it's the metal casing that's been removed and you can see the edges of. That's a few mils off the surface of the PCB and evidently stiffens it enough (on the I-beam principle of operation) to avert the problem.

[kalleboo]

And after all this discussion I'd bet Apple already fixed this with the 6s by simply stiffening the actual case...

[pilsetnieks]

They did, they even said in the keynote that they're using a new aluminum alloy[1], and it's mentioned on the website as well[2]

[1] http://www.theverge.com/2015/9/12/9315471/iphone-6s-new-alum...

[2] https://www.apple.com/iphone-6s/design/

[yincrash]

In the article, it's mentioned that in the 6S, the chip location has moved to where it's not near the most flexed area.

[Retric]

Alternatively, let the case flex slightly and have the board float in a gap attaching to the center as apposed to the edges. You still need to connect the buttons to the main board, but that's not a major issue. basically: [---|--|---] with [ ] as edge, -- as board and | as attachment points.

This is not going to be as flat as possible, but it let's you play with some flex vs thin and completely stiff.

[Waterluvian]

But you'd have to make the hundreds of broken connections between the boards you split. So isn't that a really delicate ribbon cable?

[talsit]

Not really, you do the split according to functionality, not just a straight split. CPU/memory on one side, WiFi/BT/Baseband on the other.

[quantumhobbit]

So something like Google's project Aria that they just cancelled with separate modules for each function. Only not user serviceable.

[snowwrestler]

Simply making the cross section of the iPhone rectangular again would go a long way toward stiffening it up. The curved edges are beautiful to see and great to hold, but they're quite a bit weaker as beams.

[stcredzero]

So far as I know there is no flexible soldering technology, and creating one would be pretty extreme materials science.

There are already polymer glue solder alternatives. They are nowhere near as good as solder, however.

The normal solution is to make the PCB stiffer... Fundamentally a long flat thin object is going to be bendy or brittle

What about changing the layout, such that you can introduce many, many voids in the circuit board? If you can divide up the board into many separate "compartments" then each individual segment can be proportionally stiffer and less flexy. (Long, thin things are "bendy" because the material can act as a lever against itself, so many short stubby things can be very stiff locally.)

[Retr0spectrum]

It would probably be very hard/expensive to manufacture, but you could have something like an LGA socket, with the springy contacts soldered to both the IC and the PCB. Thay way, the board can flex all it wants without overly stressing the solder joints.

[zaroth]

Do they add any flexibility around the mounting points? So the PCB can continue to lay flat, while the case flexes ~1mm around it?

[captainmuon]

It's not even just consumer electronics. I work on a particle physics experiment, and some sensors we use to detect particles are ball-bonded to the readout electronics. It turns out that thermal cycling between say -20 and 20 °C (operational vs. room temperature) a couple of times the balls break and we loose sensitivity in part of the sensor.

It seems people hate these ball bonds and want to move away from them. One alternative is wirebonds, which have their own problems. They are pretty fragile and time consuming to set. I've also learned today that they start to vibrate and break if nearby wires carry a signal at their resonance frequency, which is pretty crazy.

The golden alternative would be to build everything (sensor + electronics) from one monolithic wafer. Instead of making a silicon sensor, some silicon ASICs, and joining them with a PCB, you'd directly put the electronics in the sensor and make it one big CMOS circuit. However it will take a couple of years until we can do that.

[celticninja]

except with the RROD Microsoft admitted the issue and provided a replacement or repair service. Apple are just ignoring the issue and leaving the customer suck up the cost and hassle.

[marksweston]

Are they?

My 20 month old (i.e. out of warranty and with no Applecare policy) iphone 6+ started showing the symptoms a week and a half ago. The "Genius" at the Apple Store immediately recognised the symptoms and processed a free replacement (refurbished) phone without any prompting or negotiation on my part.

So while it's true that they are staying very quiet about it in public, it does seem that knowledge of the problem and a free replacement policy has been communicated within the company.

[microtonal]

Note: if you are in the EU, settle for nothing less than a new phone (rather than refurbished).

There was recently a lawsuit in the Netherlands where someone sued Apple after receiving a refurbished phone after replacement under warranty [1]. She won the case:

https://www.iphoned.nl/nieuws/rechtszaak-apple-garantie/

The judge based his verdict on a verdict of the European Court of Justice [2], so it's likely that the outcome would be the same in other EU countries.

[1] 20 months would still fall under warranty in the EU.

[2] http://curia.europa.eu/juris/liste.jsf?language=en&num=C-404...

[colejohnson66]

What's tong with refurbished? With regular devices, only a few are tested each batch. With refurbished, every single piece of the device should be tested.

[cududa]

Of the test batch of new phones, the standards tested for are much higher.

On refurb, many (but absolutely not all point of contact for chips to PCB) are tested, but require lower tolerances

[specialist]

I'm content with refurbs, under warranty. I consider it an extended burn-in period. (Maybe I'm wrong?)

[yodon]

Your experience is an example of how capricious Apple's handling of the issue is. I walked in with a 14 month old phone in the exact same situation as yours, including the immediate identification of the problem by the Genius Bar rep, and I was told buy a new one. Period. No help, no willingness to help. Just buy a new one.

[sottitron]

So when I've gone in for service at an Apple Store, they always take my Apple ID down. I always wondered if those who buy more get better service. For example, the person who has a purchase history of four or five iPhones, a few iPods, and pair of iPads, and a few Apple laptops might have a better chance of getting an out of warranty fix than someone who has only ever purchased one or two iPhones. It always seemed like this would make sense because Apple wouldn't want to alienate its diehards. Just a theory...

[strictnein]

I've wondered that too. I've gotten much better service in terms of hardware replacement at Apple stores than my friends, and I spend a lot more with Apple than they do

[cududa]

Could just be that you "speak tech"/ know how to talk to them better than non-technical friends

[celticninja]

Did you read the article? Considering the number of phones they talk about it is certainly not widespread or even Apple policy to replace these phones if out of warranty.

Perhaps a timely reminder than the plural of anecdote is not data.

[ebbv]

Microsoft denied the RROD for a long time and then made people jump through a lot of hoops in order to get a replacement or repair. Also RROD was far more widespread (in terms of % of devices affected) than this issue appears to be.

Most people I know who got RROD ended up just throwing their 360 out.

[pjc50]

.. admitted the issue eventually. There were a lot of people home-reflowing their xboxes, with varying degrees of success.

[pjmlp]

It is quite sad given the premium price of Apple devices, but people should have already learned that premium doesn't equal quality.

[celticninja]

Yeah I have started buying Chinese phones, started with a OnePlus One and I have just ordered a Xiaomi Redmi Note and most people consider them to be inferior because they are Chinese, oblivious the fact that their Apple device and many Androids are manufactured in China too.

Also never had an issue with the Oneplus, it has taken a battering and I have successfully changed the screen 3 times, survived a full submersion while turned on and numerous drops. It is currently missing the on\off button due to my carelessness but it is still going strong.

[PacketPaul]

Damn you are hard on phones :-). Yea I agree. The days of buying $849 phones are over. The $300 OnePlus or Blu is just as good.

[lloeki]

> It's (speculated to be) flexing PCBs under BGA chips again.

Also, operational heat (from SoC, battery, whatevs...) can repetitively flex boards ever so slightly as it heats up and cools down, and on the long term will cause BGA chips to lose contact over time. That's what famously happened on the Xbox 360, but it more silently happened on numerous other devices too.

BGA is a necessity nowadays due to pin density, but it's a tech that's incredibly easy to ruin at the slightest design or manufacturing mistake.

[jacquesm]

ROHS/lead free solder is a huge part of the problem too.