Life expectancy of eletronic components 1

[SupportMan]

I have to produce a customer support quote for a 20 year period for a product we design.
Does anyone know of any source of information (website)which will enable me gauge a rough idea of electronic component life expectancies without the need for accessing the data sheets for each component within the design?
The assumption that the components will be operated safely under their data rating, within acceptable conditions and are from reputable suppliers can be made.
Future obsolescence issues and presently known MTBF's aside, I need to factor component breakdowns and shelf storage life into the equation.
Anyone want to swap??

 

[MacGyverS2000]

Resistors from 50 years ago are still going strong (unless trashed by other failed components). Same with inductors. Caps not as much, particularly in power supply type applications. ICs have a near infinite shelf life if operated within spec, but they're usually trashed by other components going belly up.

Other than on a component-by-component basis (i.e., reading datasheets for each), you can only make generalizations.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[SupportMan]

Thanks Mac,
I was working under similar assumptions to the ones you have stated and was going to focus my efforts on the caps. However when future proofing for 20 years I think educated generalisations are probably my only starting point.

 

[MacGyverS2000]

When spec'ing caps for 20 year service life, you would do well to heavily over-rate them for the application. Anything with a non-solid electrolyte is going to be your worst enemy. If anything will vibrate (service vehicles, battlefields, contains fast fans, etc.), you need to think about thickness of component leads and/or vibration dampening equipment. If anything is powered by voltage regulators, can they withstand the abuse if the regulator begins to fail.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Suggest that you look at MIL-HDBK-217 as a start:

https://assist.daps.dla.mil/quicksearch/basic_profile.cfm?ident_number=53939

While the lifetimes from 217 are pessimistic, other similar approaches like Bellcore are what the millitary uses for reliability predictions.

TTFN

FAQ731-376

 

[VEBill]

For example:

RelCalc for Windows

http://www.t-cubed.com/

based on MIL-HDBK-217

The above sort of process will give you the predicted MTBF figures for your design. Based on those MTBF prediction you can plan for 20 years support...

...AND THEN SLOWLY GO BANKRUPT (if you've signed a contract using this raw output)...

I'm not blaming the software vendor. It's not their fault. It's the people that fail to adjust the theory to match the real world.

In the real world, actual box level MTBF figures are in the range of hundreds or thousands of hours (assuming the design is good and reliable). Not hundreds of thousands of hours that might be predicted by MTBF theory.

For example: circuit card with 1000+ components. Predicted MTBF is 150,000+ hours. Should essentially NEVER fail in theory (at usage rate). But out of a dozen units, we've already had several failures.

There's also the issues of 'No Fault Found', damage and accidents. If you do have a reliable product, then these issues can swamp out internal failures.

There's also other specialist software to make the top level ILS predictions. These include consideration of 'turn around time' for repair (where "90 days" is often a year).

In summary: long term support is VERY expensive. Vastly more than might be indicated by just the MTBF predictions.

Good luck.

 

[electricuwe]

Be aware that calculating lifetime and MTBF are two different things although both are measured in units of time.

The first refers to the steep increase of the bathtub curve on the right, the other to the constant failure rate represeneted by the bottom of the bathtub.

Achieving 20 years of lifetime is much more demanding than 20 years MTBF, if you components prone to aging effects in your system like electrolytic caps or power semiconductors exposed to cyclic stress.

 

[IRstuff]

From a strict cost prediction perspective, it doesn't matter; the only thing that matters is that you've accounted for the probable cost of maintenance. Whether that cost arises from random failures or wearout failures is secondary to the fact that you will get returns, you will have to diagnose them, and you will have to repair them.

In military systems the "life" of a system is often defined as the point at which the cumulative repair costs have risen to 50% of the cost of a new unit.

Your fist task is to simply determine what the bathtub curve looks like, and whether you can do anything to change when the curve slopes up. The longer you can postpone the rise in failure rate, the fewer returns you will get.

Secondly, you need to determine where in the system the failures will occur and whether you can make it easier to replace or repair.

Thirdly, you need to assess the potential obsolence of parts and what strategy you will take to deal with scarcity of parts later in the product's life cycle.

TTFN

FAQ731-376

 

[biff44]

Well, you want to do a few things:

1) avoid certain types of components that are prone to failure, like electrolytic capacitors, etc.

2) You want all the semiconductors running cool, like 80 deg C junction temperature max! So LOTS of heatsinking.

3) You want to derate everything dramatically. Have a capacitor that has 10 volts across it, buy a capacitor rated for 25 volts at the temperature you expect it to run at.

This will get you part of the way there.

After this, the next level involves buying components that are specially qualified and hermetically sealed, etc.

http://www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting