The tech tree

As regular readers will know, I’m an engineer by trade, and I have some high tech toys of my own, and therefore some high tech contacts.

I can tell you about a company that builds and sells high tech equipment here in the UK, I can name the company, people within the company, the entire product line, and I have been pretty much everywhere through their HQ facility.

None of that matters, only one thing matters, they are having serious ongoing issues with bought in parts and components failing QA on the point of assembly.

To the point where 4 to 6 week build schedules are running 7 to 8 weeks, and anyone in any kind of production will tell you that that is all kinds of bad for all kinds of reasons, not the least of which is throttling your production rate by 25%.

The thing is, anyone in any kind of high tech production will know exactly what I am talking about, and it is a problem that is getting worse with time.

Let’s be brutally honest here, it even affects giants like Samsung, just look at the Note 7 exploding battery debacle.

Go to a full olympic sized swimming pool, take a shot glass, dip it in the pool to fill it, and put it to one side, that’s about the ratio of li-on cells used in mobile devices vs used in cars, except in a car it won’t be a pocket or a hand that is burnt.

There is a high tech item that I myself am buying that is running two weeks behind schedule because of these same QA issues, thankfully in my case it is all made from things that *can* be fully scanned and tested by QA before it ships, so it just means delays and at least I am assured that when I finally get my hands on it, it has been fully tested first.

This is one of the tacit hidden secrets in the motor industry, only those marques with fully equipped test cells that test *everything* on the vehicle on a rolling road prior to it leaving the production line have any sort of remaining name for quality and reliability, everything else is a fucking crap shoot.

One of the real hidden secrets in modern manufacturing is when you start making everything with a CPU and sensors controlling it, instead of using that processing power to highlight problems and put them back into the remedial line for fixing and rectification, increasingly that computing power is being used to compensate for and mask those fucking problems.

Of course different production facilities face different issues, in the early days of silicon wafer fabs it wasn’t unknown for sites to go bust trying to nail down elusive problems that beset yield.

But even at the low low low end of someone going to a local hardware store to buy plywood to make a dolls house or bird house, some of that material you buy is going to be crap.

While many will be commenting that a rise in the minimum wage from 9 buck to 10 bucks will send businesses out of business, almost nobody comments about the production and yield issues faced by anyone buying in goods or materials.

Back in the day when British Standards and ISO paper trail certification was brought in, some people had an inkling and tried to bring in some system to rectify it, but, it doesn’t.

if a laser assembler buys in 10 galvo scanner heads and only 8 of them meet specs when assembled in a machine, it doesn’t help to be able to paper chase all the way back down the assembly and component line to identify what went wrong, because the fact is that in real life 99% of the time there aren’t the processes in place to be able to rectify it, cheaper to say awww shucks and ship out two new ones and say sorry, and then reclaim back from your supplier, and so on down the food chain.

The problem with this model is the labour at each stage and the production capacity at each stage is lost.

And that is lierally like burning money, it never comes back.

Just look at AvE’s teardown of a juicero, all that shit to do a job that could have been done just a well with hydraulic pressure or a pair of pinch rollers, and at 5% of the build cost.

And of course if it is all proprietary shit, it’s not like you can go to another supplier for a standard off the shelf part, don’t like denso bearings, go buy a fag bearing…

So basically in 2017 we live in a world where one of three things happens.

1. You buy a product that is *thoroughly* tested for QA compliance and it only ever gets signed off and shipped when it is right and to spec.
2. You buy a product where a *sample* are tested, and the rest are taken on faith… if the sample passes, they all pass.
3. You buy a product where *some* compliance and tolerance tests are made, but nothing is actually fully tested, that stopped at the production prototype stage.

Going back to the Samsung Note 7 debacle, scrapping and recalling the *entire* production run and model range is the most expensive possible and least desireable possible option for a manufacturer.

Even if 30% of them were shit, it would be easier and cheaper and better to recall that 30%, happens all the time with motor cars and washing machines and so on, so the ONLY POSSIBLE reason you do not do it is you *can not* do it.

The problem is that lots of industry insiders tell me that this is an ever increasing problem, and nobody apparently knows how to fix it, or at least, they do not know how to fix it and stay in business, because the fix costs money, lost of money.

Enter the iphone 10, over a thousand bucks, for *meh* features that most of the android world has had for 2 years, but the bendy iphone did the damage, so apple did the only thing they could do, which was tighten up on QA throughout the production process, and now they have a US$1,000+ phone.

And the reactions from many “safe” sales customers the world over is “fuck that”, I’ll wait until the price drops 20%, or I’ll buy something else, or I’ll keep what I have.

Yes, I want a quality item, I just do not want to pay that much for it.

Now, this is a REAL big fucking problem for mass manufacturing of consumer items, there is literally no comparison at all between a shirt factory that has to make a shirt every 30 seconds to stay afloat, and a tailor, who makes shirts to order.

I used to make and build bespoke PC’s and server for people years ago, business wasn’t great, I was never going to retire, but it was steady, because there are a steady minority of people for whom the latest dell isn’t what they want, they know this for a fact, because they tried that shit 3 years ago, and have regretted it ever since.

Apart from one PC that got used as a football by UPS on the way to the client, I never had a single warranty claim of any kind, ever, yet, I’d buy components much the same place as everyone else.

But I soak tested every single machine for 48 hours before shipping it, and I took care with the component selection and matching and assembly… I *did* have a couple that failed at the soak test stage.. the customer never saw them.

I’ve been building PC’s since 300 mhz slot 1 pentium days, I have *never* been able to replicate industry figures for hard disk failure rates, but then I’ve never treated them as cavalierly as everyone else does, including mike dell et al.

I still *plan* for HD failure so it doesn’t kill me as / if / when it happens to me, but the fact is I doubt I have had 4 ever, I have 5 spinning 24/7 in a nas box so it’s not like I never spin the things up either.

Which again brings us back to robotics in assembly, the great white secret hope in assembly plants world wide, not to cut wages bills, or eliminate bolshy staff, but to try to address the burgeoning problem with QA.

You can trivially program a robot to never use more than 1G of acceleration / shock when assembling something.

The places with all the reject components at QA stage are in class 1 from the 3 point list above, and they all have staff who are skilled and conscientious, yes, real sorry your shit is going to be a week late buddy, but you will get exactly the quality you ordered and paid for, and not one iota less, and not masked by cpu intensive software calibration to make it seem as good.

And none of them are into volume production.