How do we test reliability and quality so as to minimize the risk of board failure in the field?

Question

We are developing a smart cable for a customer. The market potential is hundreds of thousands of units. The vendor who designs and supplies the boards (with firmware) that will be built into the cables is making prototypes now. We can easily test these for correct functionality, but as far as long-term reliability and quality, I'm not sure how best to reduce the risk of systemic or high incidence field failure, which in those quantities would be an absolute nightmare scenario for a small company like ours. How do we test prototypes and mass production first article samples to minimize such reliability and quality risk as much as possible?

Answer 1

There are several different ways to approach this problem. Typically one does testing where the device is operated under stressful conditions to reduce its lifetime. This can include elevated temperature, temperature cycling, vibration, humidity, etc. Sometimes the test protocol runs to failure. The failure may be repaired and the test resumed until the next failure, etc. Often many samples are run at the same time.

For more information see: http://en.wikipedia.org/wiki/Highly_accelerated_life_test

There are many companies which specialize in this type of testing service. I suggest that you contact one of them.

Answer 2

The first step is common sense. Does it look robustly designed? Are there obvious mechanical stress points? Is there proper strain relief wherever something flexes? Are all the datasheet limits carefully adhered to in all possible corners of normal operation with some reasonable margin? Does the design handle the obvious expected abuse? This is both mechanical like someone yanking on it or stepping on any part of it, and electrical like ESD shocks. Get someone who has done this before and has experience with what fails. This might actually be two someones, one for mechanical and the other for electrical.

Take a few 10s of these things and abuse them. Do some deliberate stress tests with mechanical abuse, temperature and humidity cycling, ESD zapping, etc. Some of these will be beyond spec. The point is you want a bunch to fail so that you can see if there is a common trend to how they fail. Make that part more robust, rinse, and repeat.

You also have to test for the things that it didn't occur to you to test for. Give some to the least technically skilled people you know. You want people that don't know what they're not supposed to do with a cable. Let a few four year olds play with them, and don't try to tell them what not to do or limit what they do. Assume the four year olds are more imaginative than you are. You can decide later that jumping rope with the cable or playing tug of war with the dog thru a muddy puddle aren't things you are going to protect against, but you might uncover some interesting failure mechanisms anyway. And maybe a dog chewing on it isn't all that out of line compared to it lying on the floor and getting stepped on regularly. Don't expect people to treat your smart cable any better than a extension cord. If it's long and thin and looks like it could be stepped on it will be.

Answer 3

What they all said, plus:

Be very wary about assuming that the Macromelt encapsulation is going to make up for poor mechanical construction or stresses in wiring or connectors or joints.

If you design the result with the Macromelt as part of the design and include formal knowledge of and allowance for its mechanical characteristics and long term changes under your environment then it can be a legitimate part of your system. If it is just "a magic protective coating that makes everything OK" then you can end up with pretty plastic covered junk.

---

You mention built in boards not made by you. If these are non accessible they must survive but even if accessible you'd like them to survive even if not your responsibility. Is there adequate field protection against EMI, ESD, moisture intrusion, ... How do you know? Who said? What do they know?

In addition to Jason's FMEA](http://en.wikipedia.org/wiki/Failure_mode_and_effects_analysis) and Olin's 4 year olds {maybe not these ones} you can try something easy and potentially informative by finding ways to automatically beat it to death in ways which may not be representative of real world use but may invoke useful failures. eg put one in a tumble dryer for a few hours. Or days. Pass OK? Try it on hot. Find a way to wave it to and fro continually so cable to connectors is stressed. Something akin to a windscreen wiper action but there is liable to be something prebuilt that can be used. A few washing machine cycles sound like fun. Should it survive that? Would it need to? Why not? Do you own a concrete mixer? :-)

Is there a built in backup battery to 'go flat'. Is it soldered in place or in a battery holder. How hard do you need to hit it, stress it, pull on it to make the battery lose contact permanently or instantaneously. What about anything else inside the 'cable'.

Is there a piezo annunciator? If so, can you tap it or otherwise deliver a sharp mechanical shock to it. If you can, how big is the generated voltage transient and what does it do? (Equipment has died when people applied force to a housing and mechanically stressed the piezo which caused overvoltages.

Is the internal temperature higher than ambient or is it well ventilated. If ventilated, how are the water, ants, spider, ... kept out. If sealed, does it use aluminum electrolytic capacitors and what is the design life (noting that a depowered Al ecap lasts less long at a given temperature when powered off than when powered on.).

Will it be frozen or heated to > 60C. Who says not? What do they know? Any consequences?

Can you connect a plug backwards? Who says? What happens if you try really hard. (I have seen DB9 connectors inserted 180 degrees rotated !!!!!!!!).

Is mains powering involved. Can you power it from the wrong mains (high or low)? What happens?

Does it have an external power supply? Can the wrong one be used? What happens? If the right one is used can it fail? If it can't fail and does what happens?

Is it ROHS compliant? Does this matter? Why did you think it didn't?

Does it need EMC certification? Who says?

Can the 4 year old swallow it by mistake :-)? Or, more relevantly, some small detachable part of it.

Can it have low voltage on exposed connector pins? If these are wet or damp can his destroy the connector or device due to electrolysis. (Worst case, electrolysis can destroy a connector in minutes.)

All the above translates to: Have you kept Murphy's protection payments up to date ? :-).

Is there a path from touchable metal to interior which makes it "static electricity" prone. This is covered above by formal ESD tests but being aware of gross means of making ESD less (or more) liable to cause damage can help.

Will it survive being TASERed? There should be no need, but will it? (A favoured way of putting security camera systems out of action in some countries, I'm told).

Answer 4

The simple answer is design for failure, then prevent it.

Look up FMEA, MTBF and reliability calculations. There is commercial software that helps you with this (Relex is one I can recall) but it can be done in a spreadsheet quite easily - though it's a tedious task.

There are a large number of mil-spec (e.g. MIL-HDBK-217) and avionics documents that discuss the issue of reliability. Basicaly you take into account each component, it's use, it's stress, its ratings, location and process used to solder it, then by looking up in tables (in the mil-specs) and manufacturer data come up with a figure for it's MTBF. Often this involves finding from the manufacturer how it is made.

Consider every physical property, e.g. voltage, current, temperature (in and out of service), vibration, power, duty cycle, mechanical shock, etc.

Once you know the MTBF for each component, you can calculate when a unit will fail (on average). You factor that into the lifespan and lifetime support for the product.

For example, generally you try to reduce stress on the components, over-rating them to compensate. E.g. For a capacitor that has 5V across it for it's entire life, you might choose a 50V part rather than a 10V part.

FMEA considers what will happen when each component fails. You usually only consider a single point of failure. For each component (again in a spreadsheet) decide what will happen to the performance of that product if:
It goes out of tolerance
It shorts
It open circuits

For multi-pin devices, consider each pin separately.

Use some common sense to decide what the short might be to and document that decision.

In high reliability (e.g. Nuclear power stations) they have to take more care to consider (or mitigate against) seemingly impossible shorts.

Once you have done the above it will give you information to feed back into the design to make changes which will improve it to reduce the risk of failure, then you do the whole thing again.....

Answer 5

drop test

In addition to the many good suggestions here, you might consider a drop test.

One standard drop test involves "a drop test onto each face, edge and corner, a total of 26 times onto a surface of plywood covered concrete." (For more details, see military standard MIL-STD-810G, METHOD 516.6, Procedure IV - Transit Drop). Some independent test companies seem to prefer dropping directly onto bare concrete.

I hear rumors that 3 meter drop tests are popular for consumer cell phones. a b c d

Some independent test companies seem to prefer "testing to destruction" -- gradually increasing the drop height until something fails. Exactly what fails and how is valuable information -- often that information enables you to tweak the design to make it significantly more rugged at little or no additional manufacturing cost.