My first tech job was testing software at Hewlett Packard, circa '96-97. One of the projects I worked on was HP's first (I think) digital camera. The hardware testing was apparently running behind due to challenges the hardware team was having with abrasive qualities of some of the paints they'd chosen, and one day a bunch of us were asked to go help expedite some hardware tests.
Every button, latch, and moveable part on that camera was rated to a certain number of presses and uses. A team of folks (which for one day that winter included myself) sat in a room pressing buttons and drawing notches on a sheet of paper until the product failed or qualified. It was kind of eye opening, actually. For weeks afterward I was nervous about whether the next button press on my tv remote was the last :)
I'm curious, 16 years later.. How much of this testing is remains so manual? At HP I think the number of products and rate of change in early development made building fixtures to do this sort of thing pretty costly compared to low wage human testers.
Regular testing that you'd expect to do many times, and that's part of your product development process, is almost certainly automated.
But one-off tests are still squarely in the territory of Ye Poor Intern. I remember interning at an auto parts plant years ago where I spent two whole days stuck in a windowless lab where I laser-measured the precise diameter of a batch of motor shafts because there was suspicion of a defect from the supplier.
It's the sort of mind-numbing repetition and boredom that threatens to liquefy your brain, and to this day gives me a greater appreciation for how good I have it writing code for a living :)
(the fact that I was using frickin' lasers to do my job was novel for only the first half hour)
Certain parts of automated testing are still manual too. In the aforementioned plant I was responsible for going into the oven and making measurements every few hours and measuring hundreds of things as part of the heat-cold cycle testing, even though the actual temperature shifting and vibrational stress was fully computer-controlled.
I have similar experience in an an auto-parts manufacturing setting, where they had tried (and failed) to develop an automated system for their defect testing. However, since the area was "visual defects" and not functional defects (this certain component was in a very eyeball-attracting part of the vehicle cockpit), it was discovered that humans are far more capable than this kind of affair than computer vision at this time.
The Light Bulb Conspiracy is a great documentary about planned obsolescence. It starts with a printer programmed to fail (at the end of the movie Marcus downloads a homemade Russian driver which mysteriously repairs it). It deals with many more subjects and products, including titular lightbulbs (one of them is working for over 100 years already), ipads with irreplaceable batteries lasting 3 years, macs and other PC waste dumped in Africa as "used computers", degrowth, Bernard London, Phoebus, wooing consumers with short-lived products instead of forcing them (reminds me of STEAM).
Unfortunately, only fragments of it are on youtube (except for Polish version, in full).
A fascinating comment, and now I want to watch the documentary! Could you clarify:
"wooing consumers with short-lived products instead of forcing them (reminds me of STEAM)"
There might be a typo or I might just be not reading it right, but I can't seem to parse this, probably due to confusion about who/what "them" refers to and what action is being "forced". I want to understand it though because the comparison to Steam could be an interesting one! (Assuming that is Valve's Steam - another slight confusion, I'm not used to seeing it written in all caps so perhaps you're referring to something else.)
The documentary states that planned obsolescence was invented in the US, in wake of the stock market crash of 1920's. Some economists wanted to find a solution for the depression. Initially one of the economists came up with the idea that planned obsolescence should be controlled by the government. Check out Bernard's London paper on the right of this page:
http://en.wikipedia.org/wiki/Planned_obsolescence
The idea was that products would later be gathered and destroyed by a commission or something. A similar approach was used by Phoebus Cartel, which fined factories which made light bulbs lasting over 1000 years. According to the documentary, it still exists after multiple name changes.
I don't remember the exact details, but the documentary then proceeds to show many commercials and famous designers who promoted new goods with short life span. Instead of the government making people give up their stuff and destroy it once time is up, the focus was on "low quality" and advertising.
It's dubbed, but you can hear many interviewed people talk in english. I'm a native polish speaker. Treat it as a teaser :-). Points of interest:
- nylon stockings at 27:10
- at 16:10, an interview with a woman who met Bernard London. A few minutes later the section with commercials starts. "Planned obsolescence. A desire by the consumer to own something a little newer, a little better, a little sooner than is necessary..."
The documentary doesn't mention STEAM, but I think it's similar. STEAM is DRM in sheep's clothing. It comes bundled with so many extras (I can live without) such as friends lists, IM, automatic patching. These days, people are asking developers to publish on STEAM. They are seduced.
As for games in general and planned obsolescence - after watching this documentary I think I know why so many games these days are linear, story-based and short.
It was automated in the early '80s at a manufacturer of switches intended for aerospace. I'm sure it was long before that. You would need to be Thor to press the roller switch used in landing gear for example. They even were under glass vacuum chambers so the testing could be simulated at various altitudes.
After 4 days 6 hours it did 36,720 iterations.
After 10 days it failed on 37,112 iterations.
It takes 10 seconds per iteration. My calculations say it would have failed after 4 days and 7 hours. I wonder what happened to the other 5 and a half days.
Very cool experiment. It's this type of testing that makes me feel reassured in mechanical engineering. It's about reassuring that a car air bag will work after years of inactivity, or a car seat belt buckle will not break under pressure in a collision. The only way to know is to test, test, test.
I'm the guy that built it: During the tests, I had to guess the current count based on the estimate that one full iteration was about 10 seconds. This was not a good estimate and it made me overshot the actual number. When the test was finally over I was able to see on the SD-card the final 'real' total of 37,112. I'm making a new video explaining some of this.
Honest question: do you think it matters that the bricks are attached with a radial motion rather than linear? From my personal usage, I tend to press all areas down at the same time rather than one side fully contacting before the other. I would imagine the wear is different between these two cases.
I like your plan of testing LEGO bricks[1] from different eras, but I would recommend you use the same rotary mechanism for all the tests. If you change the mechanism to a linear one, you won't be able to compare with confidence the first (rotary) test with any follow-on linear actuator tests.
OTOH, if you repeat the (rotary) 70's era test with the linear actuator, that would be interesting estimate of whether a rotary actuation is better or worse for the brick longevity. Of course, you will still have a sample size of one, so your confidence interval will be meaningless. I can see this turning into a lifelong obsession... ;-)
Once I have a new test machine that his much faster I will indeed retest all the eras (with multiple tests per era). Of course this time I plan to build a real sound proofing system around it. This last machine was really annoying.
Do you think you could perhaps make it faster by doing computer vision and making the mechanism go up-down only and not waiting for the sensor, and then checking the photo while it moves?
The sensor was not the slowest element, it was the servos combined with the design. For the next build I plan to use a big step-motor rotating in continuous mode and transfer the rotation motion into a linear one like steam powered machines do.
Congratulations, sir. I just about lost my shit laughing when I saw the arduino in the first picture. Good work, it's really cool seeing people doing interesting projects like this. The maximum number of lego presses is something I might never have thought of (I'm a k'nex person!), but I'm really happy to now know.
I am so impressed by the mechanism used to hold and release the bottom piece -- it's an elegant, simple solution that fully demonstrates why I lack the talent to be a practical engineer.
This might motivate you to keep trying: http://www.youtube.com/watch?v=BI23U7U2aUY
What Ira Glass says is, that you get into something because you really like it and you have good taste. But for years, or decades, you're work will not be good enough to satisfy your good taste. Eventually though, this taste is what allows you to produce high quality work. But, you have to keep trying.
But in all honesty, I've experienced this in other areas of my life -- writing, programming, cooking, guitar playing -- and can attest to what he's saying.
(I'm the guy that built it): Totally true, I build a lot of stuff and the goal of this build was to make it work fast. In doing so I realized that I underestimated the strength needed to assemble and disassemble LEGO bricks. I actually planed to build a much better one using linear motion instead of rotation motion.
And, just so it's completely clear, I'm also by no means claiming you aren't talented (although I think the greater compliment is to observe that you are a hard worker). I'm just trying to make sure this guy knows that what you've done is easily within his reach, if he wants it and works at it.
Agreed. I used to run a test lab for a company that made electrical cables and interconnects. We'd create fixtures all the time for specialized mate / demate tests. Most of the fixtures were created by technicians who had built up experience over time and got good at rigging them up quickly. You do need to have a tinkerer / puzzle solver type of mentality, though.
I remember at Burningman in maybe '03 or '04, there was this roller coaster with flamethrowers called "The Roaster Roaster" - a friend of mine (who's girlfriend ran straight into the queue to ride it) was asking one of the operators about it:
"So, who designed and built this?"
"Dave did!"
"So who's Dave? Is he an engineer or a roller coaster designer in real life? How did he know how to build it safely?"
"It's _Dave_ man! Dave can build _anything!_"
Hahaha, well we were a chip company so the potential disaster scenarios weren't as pronounced. (though of of the NPI guys crushed his finger in one of the testers :()
Alternatively, don't use the bricks for 20-30 years and they will fail within a dozen assemblies. Plastic loses its elasticity so the older bricks just crack when are forced to connect.
--
(edit) "Plastic" = "plastic used in Lego bricks that were included in Lego Space sets sold in the early 80s in Japan."
Yeah, that edit was needed. My daughter plays with the lego I had as a kid (going on 20-something years now) and it certainly appears to have just as much stick as before. The small ones still defeat my fingernails.
"Plastic" is not one material, but many. Some degrade spontaneously over a timescale of a few decades; others don't. Is your statement a result of testing Legos, or a guess based on experience with other things, perhaps made from a different plastic? Chuck McManis above said his experience was that his 40+-year-old Legos work fine.
Given that LEGO bricks are made of ABS plastic it very well could have had any number of formulations where some years it was weaker than others. Should be possible to test that.
Interesting, The Wiki says up until 1963, didn't know they changed the formula at all. Be interesting to see how well ones from before and after the change held up comparably.
I have a bunch of the very early ones, the differences with early ABS are:
- changed shape slightly, some warping
- the colours faded
- they became quite transparent
- they don't bind at all or very very loose
The early ABS ones are about as good as new other than being dirty and a couple of scratches. No discolouration, they bind just fine, no warping as far as I can measure.
Cool, they must have had some great QC with the ABS to do that well for so very very long. I'd love to see the acetate ones in a picture if they went transparent, that's a really neat "failure" mode.
PS: One of those strange and seemingly random usage differences between British and American English is the latter's use of a plural, "Legos" (or perhaps more accurate to say, use of the word "Lego" to denote "a Lego piece" with resultant plural), which is never used in the UK and always sounds odd to my ear (British Lego being like sheep ...)
It's LEGO-fan usage more than British/American difference. A fan always says "I play with LEGO, this is a LEGO brick, hand me some of that LEGO" whereas a casual user might say "Scoop up some of those LEGOs; I play with LEGOs."
Yeah, I've never heard anybody in Australia say "LEGOs". It's always been an American thing to me.
Also, depending on the particular American accent involved, I often find the pronunciation quite unusual. Sounds more like "lay-goes" than "leh-goes", similar to the way some Americans pronounce the word "leg".
In the last year I've probably put together 20-30 sets of various size (50 pieces to 1000 pieces) and some lego don't hold at all out of the box either. Lego is super easy to deal with for broken or missing pieces from sets which is the only reason it doesn't bother me too much.
That was such a fun read. I have wondered about this but never tried it.
Strangely perhaps some Legos from my childhood still snap well (and they are like 40+ yrs old). Kinda makes you want to figure out how the formula changed over time.
I am sure I have not come close to 37,000 times. I have observed that Legos we got for our kids vs ones that we had from when we were kids, are different. I'm tempted to test both 'vintages'. I've also got some from the Lego Dacta catalog when we were building robots out of Legos using the MIT 6.270 build guides. They are slightly different too.
One of my favourite sales gimmicks is the wooden butt at Ikea - the testing machine, displayed alongside the Poang chairs, endlessly "sitting" and "standing".
It would be interesting to see the cycles-to-failure distribution for these parts. I wonder if the textbook "bathtub" curve would appear even for such a simple piece.
Every button, latch, and moveable part on that camera was rated to a certain number of presses and uses. A team of folks (which for one day that winter included myself) sat in a room pressing buttons and drawing notches on a sheet of paper until the product failed or qualified. It was kind of eye opening, actually. For weeks afterward I was nervous about whether the next button press on my tv remote was the last :)
I'm curious, 16 years later.. How much of this testing is remains so manual? At HP I think the number of products and rate of change in early development made building fixtures to do this sort of thing pretty costly compared to low wage human testers.