OT: Electronics word

indeed it does, but you can make those flip flops out of nand gates. Actually, you can make pretty much any logic function out of nand gates.

Sometimes called 'random logic'. This makes it sound like the inclusion of a controller imposes order... HAH!

It doesn't matter how it's made - it matter's what it does. A combinatorial circuit does not have storage. It's the definition.

Bob

The code, if in EEPROM/flash ROM, can be erased by ionizing radiation. UV windows were arranged in old EPROM microprocessors to allow this to be done deliberately, of course, and those packages would eventually be erased in sunlight as well as in UV erasers.

Some microprocessors can reprogram their own memory stores, but that usually would only cause code loss if the code was faulty to begin with... no real loss there!

Thanks for all the responses. Yes, combinatorial versus sequential logic is the distinction I was looking for. But my knowledge of this field is a little rusty, and I now remember that it's possible to build a sequential logic circuit that's a good way short of being a microprocessor. So I can see that the distinction is a grey area.

If it helps, the chip I had in mind is a DTMF tone dialler.

On a related point, does anyone know if it's possible for a simple microprocessor, in which the code is stored on-chip, to unintentionally lose its code under any normal circumstances?

Best wishes,

Chris

Or discrete transistors, or vacuum tubes, or relays...

Yes.

especially the tyoes with eprom storage technology - 8748 etc. (Not all of them had windows for erasing)

10 to 15 years was quoted in one of the early data books.

This discusses initial and long-term failures briefly:

In my limited technician's understanding the amount of charge injected into the floating gate can vary, also it may not be a perfect insulator. Hopefully that's caught at test. When the 2716 was current I was a 'test engineer' for a semiconductor test equipment maker and we played with fast programming techniques for it, mainly testing for correct output after each programming pulse, thus spending the minimum time on each bit. IIRC the number of pulses or bit-blasting time to just barely program all bits on a device followed a bell curve. In addition to increasing throughput the method gave SPC feedback to the wafer fab people.

See "Fast or Intelligent" programming.

jsw

A "circuit."

;-) Rich

I've even heard the term "discrete logic," illogical as that may seem to one who has actually worked with discretes. ;-)

Cheers! Rich

That's OK - when you introduce clock edges and FFs and stuff, it's "Sequential" logic. :-)

Cheers! Rich

Thanks all. Does this mean that a circuit without a microprocessor stands a better chance of lasting a long time, or are other factors more important?

Chris

Not really.

The real limiting factor seems to be capacitors.

The electrolyte tends to dry out over time.

Does funny things to RC time constants. :)

Connectors are another weak link unless gold plated. Following Mil Specs helps reliability considerably. Industrial standards are good too, but avoid Commercial if you expect long life.

Conformal coatings can improve reliability but they make repairs difficult:

If the job is worth the trouble you could assemble several boards and run them for a week while temperature cycling them. You need to attach a circuit that captures faults, which can be intermittent. A simple version looks for a difference between the outputs of two or more boards with an XOR gate.

I built automated environmental chambers to test large batches of boards but a toaster oven and your refrigerator would work. 0C to 70C is the usual minimum. High reliability automotive and military circuits are tested between -55C and +125C.

jsw

Thanks, Jim. Nice explanation. I actually had failures within IC chips in mind, but it sounds like these probably aren't the biggest issue in the reliability of a circuit board?

Chris

The lifetime on erasable programmable EEProms - was based on the number of word erase or block erase commands. Ion migration was the killer. Something like the 2832 IIRC. I had a customer that wore out his set. We used it to store calibration values. The values were good for a time between two weeks and a month. This customer calibrated every shift. He lasted many years before all heck happened - I was the go to guy who was the last in a line of tech support and Applications. I was able to determine what to change by talking to the tech, who I worked with before.

Mart> Christ>>

Testing weeds out the weaker parts, which is why I wrote to use Military or Industrial grades rather than Commercial. Extra testing and lower yield add to their price just like proof tested transport chain.

What is your application? The possible failure mechanisms depend on the stresses the board and devices have to tolerate. Outer space and submarines present different problems.

What are the consequences of a malfunction and how easily can you swap boards? Can you add a lower performance "Limp Home" mode like car engine controls?

High reliability can be expensive if you build ten units and test some to the limits of destruction just to get one good one.

NASA is the expert, followed closely by the military and the auto industry. I was only a "Test Engineer" when the company need one quickly, I learned by OJT.

Static discharge testing:

jsw

Other factors in circuit boards affect reliability, as others have mentioned. Traces can crack if boards are stressed, etc.

Reliability within IC's depends to some extent on temperature and feature size. If they run hot they won't last as long. Complex chips like microprocessors, DSP's, memories etc having smaller features in the silicon are more vulnerable to damage by cosmic rays, so devices that employ them are more likely to fail than older technology devices.

A guy I know did a lot of failure analysis work for manned spacecraft at Boeing Rocketdyne. After looking at a lot of stuff thru electron microscopes, etc, he said it's somewhat amazing that any consumer electronic stuff works for more than 5 years.

How many people do you know that have used the same cell phone for 5 years?