[sdiy] advice please -- aging chips and caps

[Magnus Danielson]

From: ASSI <Stromeko at compuserve.de>
Subject: Re: [sdiy] advice please -- aging chips and caps
Date: Sun, 10 Aug 2003 00:58:42 +0200

Hi Achim,

> On Saturday 09 August 2003 20:33, Scott Gravenhorst wrote:
> > Are recently manufactured ICs still subject to this?  If the old
> > ones only last 15 years (expected life), then what of new ones?
> 
> Each and every one is designed for a certain operating lifetime under 
> certain operating conditions. That doesn't necessarily mean they will 
> fail after that time, just that they are not tested for a longer 
> lifetime. Most of the parts should have a far longer than the 
> data-sheet lifetime, both because the failure statistics mandate it and 
> because there is some margin in the qualification tests. Exceeding the 
> maximum operating conditions however shortens the lifetime drastically, 
> even if it happens just once and for only a short time. About the only 
> things that can damage an IC while it is not being used is ESD and 
> (large) temperature cycles.
> 
> The ESD robustness of original 4000 series IC is not really good, so 
> you should really always follow the proper handling procedures. I don't 
> know if the later improved series have gotten an ESD overhaul,

They have! The 4000B series have what eventually became the standard input
and output protection that we live with today. I've even seen how Xilinx for
instance refer back to an old RCA app-note (ICAN-6572) on the protection
networks (see also ICAN-6525, both written in 1975-04). I haven't found them
on the web, but they sit nicely in my RCA catalogue. Looking at the 4000B
series protection scheme and that of modern CMOS chips, nothing has really
changed except for natural issues of geometries, doping etc.

However, the 4000A family wasn't without protection, it just wasn't as good as
the new one. A totally unprotected CMOS setup should be handled with as much
care as individual JFET/MOSFET transistors.

A word of notice, there are a few chips which deviate from the normal
protection-scheme, so it might be a good idea to check a propper catalogue for
the exact number (4049A vs. 4049UB for instance). The analog switches can't
have the same form of protection as normal digital gates.

> but good 
> ESD structures take up a lot of area and IMHO that doesn't go together 
> with parts that often cost only a few cents more than their package. 
> Most people assume that the IC dies instantly from an ESD pulse, but 
> that is typically not the case (especially if someone is trying to 
> prevent ESD by insufficient means) - degraded performance and 
> subsequent early fail is much more likely. Also a pre-damaged IC will 
> likely not live through another ESD event, the protection structures 
> are normally not designed to withstand repetitive ESD.

There are several types of ESD failures which can occur abruptly or come
creeping in to show their ugly face at some later time.

> Metal migration actually comes in two flavors: stress migration and 
> electromigration, the latter is significantly different for DC and AC 
> currents. They are both kept in check from the manufacturing side by 
> optimized processing, additional layers and anneals. Some of these 
> things have only recently been developed and/or understood sufficiently 
> well to take full advantage of them. For the design side there exists a 
> slew of electromigration design rules. As with all design rules you 
> have to make some assumptions and for instance a 4069 being used as an 
> op-amp probably wasn't on the mind of the designer.

Don't bet on it. Motorola's "McMOS Handbook - Products - Characteristics -
Applications" from 1973-10 spends a whole chapter (Chapter 8 - "Analog Basic
Circuits" by Robin Hodgson) digging into various analogue aspects, including
how to make use of them as linear amplifiers, noting the level of distorsion
for various supply voltages etc. for different overtones. Hillarious!

I've also not heard that the 4069UBs in the Pearl Syncussion is the most likely
thing to go... but then again, I do agree with you! ;O)

"Out of spec" is always "Out of spec" and you have to use your brains to make
sure it is safe there. It is more responsibility comming your way, and you
better take it. Sometimes you can pull it off and make use of things even if
that was not as intended by the original designer, but hey, that's the fun of
it! ;O)

> >  I'd prefer that they outlast *me*.  This is pertinent for me
> > because I have a 4000 series fetish.
> 
> Hmm, what kind of longevity do you expect for you and what's the duty 
> cycle on your 4000 series IC?

It's just his life-support kit, but I'm sure he don't mind if his pacemaker
fails before him. ;O)

Cheers,
Magnus