Naked Science Forum
Non Life Sciences => Technology => Topic started by: neilep on 05/07/2009 13:44:05
-
Dearest Chipologists,
As a sheepy I of course luff chips !..hmmm..I actually like putting soy sauce on mine !
Take a look at this silicon chip eh !
[ Invalid Attachment ]
A Buxom Wench Of A Silicon Chip earlier Today
My Oh my !!..that's a thing of beauty isn't it ?...If I was a boy chip and that was a girlie chip I'd be getting all jiggidy with that !!..oh yes indeed !
However, it seems my affair could only last for 30 years or thereabouts !
Why's that then ?..why is the lifespan of a silicon chip only 30 years
Ewe see, I just don't know...I want to know though !
Hugs & shmishes
mwah mwah mwah !
neil
I worry About The Lifespan Of Silicon Chips
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
-
Some of the chips in my Commodore Pet are datestamped 79 and it still works.
It's difficult to know how long they will continue to do so.
At least some chips die due to damage from alpha particles produced by decay of radioactive materials (accidentally) present in the matrials they are made from.
Does anyone out there have any chips older than 1979 and still working? Any seriously "retro" calculators perhaps?
-
Computer CPUs die from overheating: chips are cooked.
-
In normal operation I don't think a semiconductor 'wears out'. BUT packages can lose their seal (thermal cycling is often responsible) and the atmosphere can get in and pollute the chips. You have to pay extra to get better quality packaging to cope with this.
-
Lifespan of chips is not easy to predict. There are a number of potential failure mechanisms and any one of these can result in end of life. Flash memory is likely to have an end of useful life when sufficient memory cells have failed that the "wear levelling" mechanisms and error correction systems can no longer cope. The fundamental failure mechanism is leakage of charge from the floating gate due to a build up charge in the isolating oxide. There may be some oxide damage too. Lifespan could be between 3 and 10 years but depends hugely on usage and specific design. In more normal logic devices, gates and/or interconnect can fail and this in turn can depend on inherent design weaknesses in particular parts of the chip. Gates fail because the oxide can breakdown due to a random event, crystal dislocations may change in structure and result in excessive leakage or interconnect can erode away by a process known as electromigration. The interconnect failure is probably the most predictable because it is well understood. All the failure mechanisms have some sort of temperature coefficient depending on the mecanisms empirically defined activation energy. All get worse exponentially with temperature so if one part of a chip gets unduly hot then this will lower its lifetime.
Mechanical failures due to bonds becoming detached (vibration or contamination) can also be problematic. High reliability devices are in sealed ceramic packages (sealed with dry nitrogen inside) where lower cost plastic packages can suffer from moisture ingress. The failure mechanisms here are nearly always to do with there being moisture present when the devices are soldered down; the heat shock causes a rapid outgasing that delaminates the package but does not always cause an immediate failure. Measures are taken to avoid this by dry-packing devices and only breaking the seal shortly before assembly. The de-lamination problem is often spotted as a batch related issue though as the delamination often pulls bonds off. A further problem is that moisture can activate chemicals present in the chip that can form acids which attack other parts of the chip. A common problem used to be leaching phosphorous, forming phosphoric acid and attacking the metal interconnects. Another former problem was "purple plague" which was very descrptive of the multiple alloys that could be formed between gold, aluminium and silicon. THese latter problems are rare nowadays.
Irradiation from cosmic rays is one of the mechanisms that can shorten life too but earthbound chips get much protection from the atmosphere. The collision cross section has to be right to produce damage in the chip so events are rare but very highly significant for devices in satellites or high flying aircraft (or of course those intended for miliatry applications). Alpha particles generally wont get through the package though there used to be a problem from weal alpha emitters contained within the glass filler used in some packages. These caused "soft" failures in memories (loss of data but no permanent damage). Cosmic rays van still cause soft failures by loss of stored data but it is quite a rare event and ground level.
Some chips may only last 10 years but generally the industry is fairly conservative and many could last 100s of years. As I said, there are very many factors.
I hope this helps and some of the buzz words may help anyone interested to resarch more via Google.
-
I was going to talk about electromigration but I forgot what it was called until graham.d mentioned it in his very thorough post. From wikipedia:
Electromigration is the transport of material caused by the gradual movement of the ions in a conductor due to the momentum transfer between conducting electrons and diffusing metal atoms. The effect is important in applications where high direct current densities are used, such as in microelectronics and related structures. As the structure size in electronics such as integrated circuits (ICs) decreases, the practical significance of this effect increases.
So eventually enough material is moved to break the circuit.
-
Err, sorry for all the spelling errors as I'm abroad on hols and on a lap top.
-
graham.d
That's a nice comprehensive list of damaging processes. I must say, I was thinking in terms of 'big stuff' like discrete transistors, which have a lot of mass in their active regions. I had forgotten that, in modern chip technology, there aren't that many atoms associated with each circuit element and they would be much more vulnerable.
ps Just lever off the keys and put them back where you would expect to find them! It may not help you but it will confuse the hell out of the person who lent you the keyboard! Get your own back.
-
I think older chips will last a long time, subject to some environmental/encapsulation concerns listed above.
However, with the very latest microprocessors which use 45nm feature-sizes and smaller, I understand that the constituent atoms can essentially flow or migrate over time (a decade or two), enough to cause the circuit to cease functioning as intended.
They are also increasingly vulnerable to cosmic-ray damage.
-
I would not be too certain of that, techmind. The fine geometry chips have to be designed to be more tolerant of transistor to transistor variation, which is much greater than on larger geometry processes. This makes the designs inherently more tolerant to variation that occurs due to Vt changes induced by faults developing or from the effects of ionising radiation. 45nm processes already have a huge drain-source leakage problem to cope with which also means that changes have to be big to be noticeable. The problem is really the vast increase in the number of transistors per chip that results so that maintaining even a similar reliability is quite a challenge. Another issue not fully solved is that the industry demands more density and greater speed, both of which push up the power dissipation much more than the reduction afforded by the lower operating voltages. This certainly badly affects most of the failure mechanisms although these technologies employ many levels of interconnect, thicker than it is wide, and of alloys of metals that are both lower impedance and more resistant to electromigration.
-
Don't throw away you older microchips.. those sneaky little chip designers stamped them with artistic flair... check it out...
Microchip Art..
http://micro.magnet.fsu.edu/creatures/
-
That's an interesting website JnA. It is quite common to use a Copyright symbol and a Mask-right (like a copyright but with "M" rather than "C") as well as mask details used to identify layers and revisions. Chip names/numbers are also put on as well as other essential aids to process and design. Putting pictures and other symbols on is very easy but could be severely career limiting if it causes problems in process and needs another couple of million dollars to fix via a new mask-set. The Design Rule Checking (DRC) software should detect illegal shapes and prevent problems though some chip furniture sometimes escapes full checking (has a "do not check" marker put on it). It used to be common to do this in the past, but less so now for these reasons.
-
I never really thought about a chip lasting that long!