Discrete vs. IC reliability

...BUT you fail to understand...smaller is better because it is faster and i can use less silicon, meaning i can make more profit. Do not confuse the issue with mundane facts.

Yup! Remember at Mother Faichild about all the fuss over ceramic(for Cer-DIPs) causing major failures..materials came from the wrong mine.

There was one case (though I don't remember all the details) where Intel couldn't get a radioactive gas that they used to test the porosity of plastic packages so substituted a short-life alpha emitter for a longer life beta. They injected failures into the packages that haunted them forever.

It depends on the circuit and the application.

Firstly, he would draw on his own experience and that of others to figure out which is the most reliable method of obtaining the required function.

This generally rules out anything new.

Still interested?

RL

On Jan 7, 2016, legg wrote (in article):

The basic answer is to use reliable parts (get MTBF data from manufacturer), derate them (keep below one half of all the data sheet max limits, and above twice for min limits), and keep them cool (failure rate doubles for each 10 C rise in temperature).

Joe Gwinn

All good principles.

In addition, you might look at some details of the

projects. AMSAT builds amateur radio satellites based on ordinary (COTS) components (since available at reasonable prices). Some might function for a few days or a few decades in space, but you do not have way to know before, which will be the final lifetime.

But in general, keep any temperature (differences) low and use double or triple redundancy (if no weight constraints) to keep the total _system_ reliability at a reasonable level, without using extraordinary high cost specialized components.