Hardware Reliability - So how frequent are IC failures?

[nitrojacob]

Hardware Reliability

You might have noticed a disclaimer in the end of digital IC datasheets(eg. 74ls series), that they are not guaranteed enough for medical uses.

I have no experience with a digital IC failing (unless mishandled). Nor have i worked in any digital manufacturing companies.

So how frequent are digital component failures. Is it more for VLSI than MSI/LSI? How are the medical equipments designed/manufactured?

 

[keith1200rs]

Re: Hardware Reliability

I wouldn't worry about disclaimers. They are just trying to avoid being sued but they are pointless.

Any medical equipment needs to be approved for medical use. How difficult that is depends on what it does, if it is patient connected and what would happen if it failed. If someone dies if a single component fails then it wouldn't be depending on a single device for it's operation.

I guess the same would apply to aerospace parts. While I have designed a few I have never got involved in the testing/certification but some systems I have been involved in have had three parallel systems and a "voting system" so if one fails it is assumed that the two that agree are correct.

If you are designing medical devices you need to read either the FDA regulations in the USA or EN60601 in Europe, I think it is. there may be others - I have mainly only been concerned with electronic products.

Keith.

 

[nitrojacob]

Re: Hardware Reliability

Thank you keith.
That means the life critical equipments are designed with redundant modules.
No matter what the component or module reliability is, we will go for such designs to reduce the chance of damage to life. But what about the actual digital component failure? How often are they? How often does a microcontroller/processor fail, if used entirely within the safe limits specified in the Datasheet?

 

[keith1200rs]

Hardware Reliability

You really need to get someone involved in the project who is experienced with medical devices. Even so, you need to start reading the FDA regulations ASAP otherwise you might have to scrap any work you have done and start again.

There are different rules depending on the class of device. Most of my work has been class II. As I recall, it is not a case of how often things fail but what will happen when it does and designing the system to avoid the adverse consequences. In the case of software it is more complicated to ensure a bug doesn't exist that could cause injury or death. If you don't follow the regulations now you could end up having to scrap all the software you have written and rewrite it.

It is a complex area and I am no expert on it but I have been involves in a couple of medical products and getting them into medical trials.

Keith

 

[nitrojacob]

Thank you keith. But I'm not involved with any medical project at present.

Due to the recent happenings in the automotive sector, every embedded engineer has got some interest towards reliable designs. Everywhere people speak of hardware reliability first. But from my experience, it was not so obvious to design for the hardware component failure. Now i got the idea.

I would like to get some idea about the numerical value of digital component failure.

 

[keith1200rs]

Ah, your first post confused me in to thinking you were designing a medical product.

One possible source of information is manufacturers themselves. Quite a few publish failure data from accelerated life tests.

Keith

 

[123jack]

Look into the concept of MTBF - (mean time before failure)
All products will have this specified somewhere even if you have to ask
the manufacturer directly. Also identify the various MTBF purposes.

When building equipment there are professional statistical approaches to
assessing systems MTBF figures based on component rates. In large products
insurance companies may require such specifications. In practice they may mean very little depending on circumstances.

For fun you might want to find out the MTBF rate of your hard drive (sometimes
quoted in bits for data transfer purposes) then workout how much data you need
to transfer before the manufacturer says you "could" create an error statistically.

When designing equipment you should also investigate the concept of
"failsafe design" - briefly the concept requires your design to do what is says - fail safely under all circumstances. Again there are professional methodologies available.

If something specifically states "not suitable for medical products" this does not
mean "because it might fail" - it could mean a whole range of factors such as RF
issues or manufacturing variations in tolerances etc.

Companies do not put such statements on their products on the off chance they
might get sued. To do so would in itself create liability in some cases.
They do so to inform you clearly to think harder about what you are doing.

Hope that helps.

jack

 

[xaccto]

I simply have a related question to whats discussed above out of interest,

I have seen manufacturers list rad hardened range of products for say aerospace applications, but for medical applications I wonder where what range of products manufactures choose, because thus far I have always seen the disclaimer "not for medical applications" on data sheets.

 

[nitrojacob]

Thanks Jack.
Xaacto, from the discussion and a few readings, what i think, is that, if you don't suspect a high probability of radiation induced errors or any other kind of prominent cause of faults, you can design with consumer grade electronics.

No matter what grade of component you use, you should have provisions at board level/system level to withstand random faults to any of the subsystems/modules/chips.

Correct me if I am wrong.

Another question is, at what level is this usually done? Whether you duplicate/triplicate
the entire main board /
modules like RAM/SMPS /
Chips like RAM chips & SMPS controllers ICs /
is this decision up to the system designer?

 

[123jack]

Normally you wouldn't want to duplicate circuitry (unless failsafe just isn't enough -
of course - for instance - launch control on a space shuttle could have some fail safe circuitry but you'd want the oxygen cleaning system to have duplication of some parts
while in space - because up there - no one can hear you scream....)

It's always up to the designer - but then mandated by house rules and walkthrough
procedures.

It depends entirely on what you're building and who with.

You could build a heart monitor from tin cans and bubble-gum - but who would buy it? There is a marketing input as well as a regulatory stance also.