Raspberry Pi

http://www.theregister.co.uk/2012/02/29/raspberrypi_mania/

"Farnell spokesman Ken Leitch told El Reg: "It's been a phenomenal day. This is an incredible little computer, we knew anticipation was huge and it sold out very very quickly, within the hour. We had a massive amount of interest across different territories - half a million interactions with our site in 15 minutes at its peak."

That's a damn popular bit of kit then! Wonder if it'll really get kids into coding? I wish I had gotten into computing at school but the computer they had at school didn't even have a screen just a line printer, couldn't get my head round it at all.

I did my first programming on an ASR33 Teletype hooked, by serial thankfully, to an HP 2000E timeshared system, back in '72. I thought I was damn lucky, as most folks then did programming using punch cards on batch process systems and had to actually wait a day for their results.

I wonder how many of these will end up in the hands of kids?

My younger daughter is interested in this kind of thing, I should pick up something like this for her.

Looks like I won't get mine until the end of April. But given I don't have an actual use for it yet, it's not a huge deal.

What really irked me, was the reaction of numerous people in the comments and forum sections of their website throughout the whole process. There's the ones that bitched about the design, the ones that bitched about the production in China, about the lack of pre-orders and about them entering into agreements with RS and Farnell.

None of them seem to grasp the complexity of designing, building and selling something like that, especially when none of the people involved with it work full time on the project. 'A bunch of fucking arseholes' doesn't even begin to describe what I think about people like that.

Glad you got that off your chest Andy

Week of the 14 May they are quoting for mine. You must have been earlier in the queue than me Andy

I managed to officially express an interest at Farnell as soon as their website became useable again, and I put an order in about 5 minutes after I received their email inviting me to actually purchase one.

I'm quite looking forward to having a fiddle around with it. It's certainly more capable than previous embedded boards I've tinkered with in the past. About 6 months ago I had considered a Beagleboard to stick in the car. But the fact the Pi was so much smaller and cheaper, while still retaining much of the useful features, I thought it would probably be a better bet.

Sorry for sounding off in my previous post, the Raspberry Pi seems to have really attracted a lot of very self righteous wankers. It certainly goes to prove the "Greater Internet Fuckwad Theory".

/me looks at his "first 400" voucher and smiles

Do you guys have any specific projects you intend to use them for? I want one, but don't have a particular use right now. I already have a 24/7 server/NAS and a mythtv box I can use if I want something like a mini webserver.

I have to say I'm quite impressed with the Quake III demo in terms of performance.

I'm using it as "just another really interesting target" for our distro: http://www.merproject.org/

For RasPi, I'll be the mentor for using the Mer build systems.

Looking at their forum yesterday it also seems to have attracted people that want to run Windows on it, why?

Well with a touch screen wired up to it, it might making an interesting Windows 8 device.

Though I thought Microsoft were restricting Windows on ARM to OEMs, so that might be a bit challenging.

Given that Windows on ARM is supposed to be getting Office, it would also make for an interesting low cost desktop machine for the office. But again, only if you could actually get WOA on it.

Maybe their plan is to ensure that the RasPi adheres to the Microsoft ARM "secureboot" lockdown?

In the front of my is an in car pc to give the functionality for playing music my empeg gave me, but also coupled to a dvb-t tuner for DAB radio and OBD and maybe some bespoke PIC circuits for car stats/information display.

I can't see myself getting one, but then I can't see any of it's alleged target market (kids) going for one either.

I've got a couple of things I'd like to try.

Firstly I'd like to take a crack at whole house audio. I'd look at using something like Shairport and a set of amplified speakers in each room, but also look at running something like Softsqueeze as well to cover all bases. I can use either the built in ethernet or a WiFi USB dongle to make installation straightforward, instead of running line level audio or speaker cabling through the house. I could even look at using the built-in GPIO interfacing to automatically power on and off the amplifier as and when it's being used. The cost of the R-Pi makes it the cheapest part of the whole exercise.

Other than that, I suppose I'll look at building my long awaited Empeg MK3. My Parrot Asteroid has been a serious disappointment. Great hardware and form-factor, let down by really crappy player software.

Kids themselves aren't going to buy them. I think the plan is that their schools will be. Even if R-Pi's primary goal fails, the nerds of the world will be buying them a dozen at a time!

I can imagine the BeagleBoard lot must be spitting feathers.

Absolutely, which I guess is not a bad thing but not what they intended

I managed to order one through farnell shortly after they started accepting orders.

My estimated shipping mate is week commencing the 12th March, the only shipping date I've received was in the initial order confirmation, I understand quite a few people have received other emails with revised shipping dates.

http://www.theregister.co.uk/2012/03/08/raspberry_pi_delayed_networking/

Yeah, that's a bit amazing. No first article inspection at the factory? No test plan that checked even one unit manually before they ran 10,000 down the line?

I'd be very surprised if the factory weren't rather nervous about the Raspberry Pi guys (who obviously have never done any manufacturing before), who probably assured them they didn't need any testing and just to ship the boards out...

Given that it's dual-sided SMT on that board, the rework is going to be manual and they'll likely damage some boards whilst they're doing it. Not cheap either (plus shipping all the boards back and forth another time), which is even more of a concern when you're not making anything on the product in the first place.

Ouch.

Interesting take.

It sounds to me (from the limited information in that article and the necessarily one-sided info from the RasPi forum) like the *factory* cocked up. They were given a spec for a component and used the wrong one.

"where we’d specified jacks with integrated magnetics in the BOM and schematics, the factory soldered in non-magnetic jacks"

I'd say the RasPi guys (who may not have the experience to know that when you pay someone to do a job, and quite possibly outsource QA to them too... will still screw up) are the ones who should be nervous about the factory - and indeed maybe what you find amazing is actually that they trusted them to do the complete job.

I'm interested in why you say the RasPi guys told the factory "they didn't need any testing and just to ship the boards out..." and not the factory telling the RasPi guys?

In my experience (and granted, it's a million miles away from this environment) suppliers rarely say "hey, there's a good chance we'll screw up, I hope you have a plan to test our products match your spec before you trust us enough to ship"... quite the contrary.

I think what Hugo means by "have never done any manufacturing before" is that perhaps they don't realise when a price is too good to be true. There are contract manufacturers who work the way you expect, and who have the forces of contract law ranged against them if they don't. But there's also an underworld, wherein you can get things manufactured much more cheaply if you're prepared to stand over them the whole time keeping them in line. In the underworld they will replace specified components with cheaper ones they had lying around if they think it will still get past board testing -- prototype Trekstor Vibezes came back with a zero-ohm resistor instead of the very-low-value one for current sense. "Experienced" production managers know to watch out for this, in the same way that experienced project managers know who to bribe in tinpot dictatorships.

The unfortunate thing is that the cost difference is so wide, that over time the underworld always wins.

Peter

It's very possible that the factory swapped a part out without the RasPi guys being aware, but that doesn't actually happen that often - usually they will ask to swap it and send a datasheet, but if they buy an incorrect part the customer isn't going to pay for it, so this isn't so common.

First article inspection is pretty important because you're very likely to get 10,000 units identical to the first one - that's how factory lines are designed to run. Possibly their testing was booting from an SD card and checking a screen picture came up, which would have caught the more tricky issues like BGA soldering problems, but wouldn't have helped a network issue.

Factories tend to push hard for testing because if there's a problem with the goods they ship, it comes back to them and it's a PITA. They would like to know, with some certainty, that they are shipping working product.

By the sound of it the biggest issue is their naivety. We do contract production of clothing for mostly UK retailers, it's sometimes shocking how little people that may have been in the clothing business for decades actually know about the manufacturing side of things. Even when you know the business it's not always a given that you're going to get what you specced, we had an incident a few years ago where a supplier in China made a whole batch of jeans with rivets that were not guaranteed nickle free. This was not picked up by anyone, including our paid for inspection service in China, till the merchandise hit the stores and someone had a reaction to the rivet. A very costly mistake.

Did the Empeg team do the deal with Rio purely for the money, or the access to manufacturing and marketing capability? I guess the same sort of thing happened at slimp3, manufacturing your own kit is never easy.

I find this whole problem interesting so I did some digging.

Summary:


More detail on QA: http://www.raspberrypi.org/archives/781#comment-14934



This link mentions a remote-testing rig which, as you suggest, doesn't seem to verify the network (in fact it doesn't sound like it even boots the device... would need to dig more)

Also: http://www.raspberrypi.org/archives/781#comment-15451



So given the level of transparency of this team I'm giving them the benefit of the doubt. It does sound like a factory cockup and it's hard to see anything the RasPi team could have done differently short of sending someone to China to oversee the factory.

And there in lies the solution, they should have either run a limited quantity to ensure that what would come off the production line will work or had somebody out there to test one. If they'd done this, they would have spotted that there was a problem with 10,000 of them before they committed. If they subsequently received the production batch and they didn't work, then they could go back to the factory and point fingers.

I'm with Hugo and Peter, this is a situation that has been caused by Raspberry Pi not checking that their production build will function correctly, the way they've done this can either be put down to inexperience or lunacy.

This is part of the reason that we have our our pick and place setup, we can assure quality, our PCB's are bare board tested for connectivity.

This reminds me of us recently purchasing some 7 segment drivers for a product that we build for another company, we ordered them from china, they arrived with the correct part number on the devices. We put one on a production board and low and behold it didn't work, the numbers were coming out all wrong. Turns out the chinese manufacturer of this device had copied the (long out of production) original part but had labelled the standard inputs part with the BCD part variants.

Baring in mind that this was a large through hole device, we would have been in a whole heap of doo-doo if we'd just blindly trusted what the parts were and had them soldered into production devices.

Bottom line is, never trust any electronic component that is sourced from china, chances are it's not what you think it is.

It would have taken ONE sample unit to verify this problem. With a handful of samples they could have also torn at least one down completely to verify against their BOM. Before running 10k.

Really that's not even a mistake a first timer should/would make. They were probably just busy and excited and didn't bother with due diligence.

Not sure what bit of "We received one initial small test batch of boards... they tested fine" and "The first production batch had a different jack in place" you missed?

They did exactly what you say you would have done.

The factory appear to have swapped components between the initial test batch and the production run.

Maybe I should assume you were just a little too busy and excited to fully check the links (which I admit seem to only point to the general vicinity of the relevant comments) before you posted your analysis?

I have nothing to do with RasPi... but if you're going to use my post to call them inexperienced lunatics then I think I should at least verify the logic you use.


They did - so actually what you would have done *didn't* solve the problem. So given your first solution was the same as theirs does this following quote apply to you? If not, why not?



Your second suggestion:

is apparently they are now doing with a Farnell/RS rep in china. So again, they did exactly what you'd have done - how does that justify the labels you applied?



Yes, indeed - and I'm not seeing much evidence that they did.

You mentioned Hugo and Peter's posts - actually that's not how I read Hugo's posts - he very much assumed that the factory would be professional - Peter on the other hand almost certainly nailed it.

As far as I know, they did an initial run of boards in the UK, i've not personally seen any posts where they indicated that they did a limited production run with the supplier they were using the make the large batch, although I'm more than happy to be proved wrong on this. If they did do a mini run and the part was swapped out after, then I'll remove the lunacy tag, they can keep inexperience because they should have had somebody there to ensure the quality of the product.

And even if they are doing it now with farnell, it's a bit like locking the stable door after the horse has bolted.

Hugo placed the blame with Raspberry Pi, I concurred. No matter what the factory did or didn't do, the onus lies soley with raspberry pi to ensure that their end product is built to the BOM and their exact specification.

Like I said, inexperience or lunacy. Take your pick, it's one or the other.

I got focused on the part where they said they ran tests remotely and via "imagination"

Oh dear...

Seems that they've now been told by their distributers that they won't sell them until it's been though EMC testing.

What's bizarre, is that R-PI are claiming that they've been and done it, but they don't know at this point whether it's going to delay anything, my experience of EMC testing is that you know pretty darned quick if something is wrong and you then spend most of your time fitting chokes and makeshift cans to get it inside the curve. My guess is it failed and that they couldn't fix it on site.

Chances are they even use the same test centre as us seeing as they're based in Cambridge.

dB tech FTW! (is that who you use?)

CE is self-cert, so if the testing has all been done, and it passes, then there's no delay. You just print out some CE stickers and be done with it.

To ship immediately to the USA you'd need to go to an FCC certified lab and get FCC testing done there (it takes longer if you do it at a non-cert lab and then send the paperwork off) but again that's just a day.

My guess? They ran compliance testing without all the wires connected, which is an easy way to make something pass but doesn't actually satisfy the requirements (FCC in particular insist that every port is connected - the empeg test setup even had a dot-matrix printer connected to the PC that was connected to the empeg's USB port).

They do appear to be learning how to manufacture electronics very publicly... embarassing.

Yeah, we use dB technology. They're super helpful there - even if the outdoor testing in the middle of winter can be described as "less than fun"!

Yeah, we test our equipment with a everything that could possibly be connected, erm, connected. Even though in a real world situation that would be very unlikely - the more nice pieces of floating cable you add, the more potential antennas you create.

As you point out, it takes very little time to test and certify something. If they've done the tests and are holding back, then it sounds like me that it failed.

Depending on how badly it failed and how easy it is to attenuate the problem frequencies will give them a scale from "oopsie" to "houston, we have a problem".

If they've totally fucked up, then they've just made 10,000 boards with BGA's on that their suppliers won't sell.

Seriously, in the grand scale of things it'd have cost them like a £1000 to spend a day at the test centre with a pre-production prototype and do all the necessary tests. I'd certainly be doing that before pressing the big shiny "go" button on 10,000 boards with BGA's on them.

Can I reinstate my lunacy tag?

I guess if they need to do a v2 with added components they could still bring the v1's back and sell them direct, or via a small distributor.

In fact I know there are hundreds of sponsored Pis reserved for selected developers - they should really get those back ASAP and send them out so the key developers can start messing around!

The problem is that they'd still be selling the product direct to end users, if the Pi was a component part of a another system then the compliance issue would be the problem of the integrator, but that's not the case here.

Raspberry Pi would be liable for all legal issues that subsequently arise because of EMC problems. Trust me, you wouldn't want to be liable.

Their statement on the issue seems pretty clear to me:

But they can't do that, the scale of the product being shipped means that the potential for a complaint is high. They always knew they were shipping 10,000 in the first batch, so to say "we were going to do it later when it's cased" is just either stupid or nieve and stupid.

If they're shipping to end customers without a case, then that's the configuration that should have been tested. It's even likely that they wouldn't have to do a full set of tests on the boxed version later if the unboxed version passed ok.

I just don't get their statement, "we intended to defer". Great, you also intended to ship product in the real world. For a bunch of engineers, they sure seem dumb.

As Hugo says, they're learning the hard way on how to manufacture electronics - or at least making stupid mistakes and doing it the hard way.

I don't know anything about compliance testing, but do eval boards need to be compliance tested? What about all the various Arduinos of the world, are they all tested and certified? I kind of doubt it. What's the difference between them and Raspberry Pi?

It depends how blurry your spectacles are.

I believe that Arduino and such like get away without certification because they're sold as development systems, which is true, without you putting your code on it and connecting up your own electronics (or other boards) it doesn't do anything, it's up to you to provide the magic.

Raspberry pi on the other hand have a full system which I think they'd have a hard time convincing anybody that it's a development board, especially considering that they've gone out of the way to show that it runs a consumer media player distro and hooks up to your tv - and without restricting sale to developers (how do you even do such a thing) and making a huge batch, they really should have realised that this was going to be an issue.

Of course, they could still just say it's a development board and ship them, but I think they put themselves in a very grey area legally, especially if the thing doesnt even pass the radiated emissions tests.

I dunno. the Pi looks very much like a bare development board to me -- very much like any of the Arduinos. I don't know if the latter have ever passed EMI tests, maybe they have?

It looks pretty bare to me too. I can't imagine anyone expecting a real "consumer" device buying one, not yet and certainly not in it's current form.

They've obviously been naive with this product launch, but looking at the product and it's intended market they're probably not the most consumer product savvy people in the world.

I'm sure that this experience will have taught them loads about outsourcing, testing, and QC. If the next batch has similar issues then they probably need to give up.

Like I said, it depends how blurry your spectacles are. You wouldn't class a PC motherboard as a development board would you? I don't see that as being any different to the pi.

Could average joe technophobe do anything with an arduino in 10 minutes? No.

Could same average joe technophobe have his raspberry pi connected up to his tv, running xmbc and playing media in 10 minutes? Yes.

I personally think they'd have a very hard time trying to convince the relevant authoritative bodies that it was only a development board in the case of a problem with emissions.

The sales numbers that are floating around 100k+ and the interest in the media playing says to me that a lot of these units will sit behind a tv and have certainly not been bought by people interested or capable of software development.

(sorry for any errors in this, I'm currently sitting in schipol airport waiting for my flight and using my phone)

No, but there's no one in my circle of friends/acquaintances that's actually bought one. I've got a friend who's been waiting 8 weeks for me to pop over and stick some extra RAM in his PC, despite my reassurances on how easy it is.

Of course I'd expect a MB to be properly tested and certified, but right now I'd only buy the Pi in a case of some kind with a pre-installed OS and some readily available apps. So to me it's substantially more geeky than a MB.

I don't think the same Joe Technophobe would buy the bare board to put behind his TV either. I think he would wait until it came in a case with a PSU, which was what they apparently were waiting for before they did the compliance testing. I don't think he would have more problems plugging in a shield with an LCD and some button, plug in the USB cable and load up a pre-made sketch than he would have with a Pi, where he would have to connect up a PSU, get XBMC onto it somehow and connect it up to the TV...

They apparently had orders for 100,000 of the things.

I guarantee that had they not made a big thing about xmbc and video playback on it that they wouldn't have sold anywhere near that. I frequent some other forums which are non tech related and I was even surprised to see talk of the pi. Nobody mentioning development, just people excited about running xbmc.

As for getting the os on it, xmbc will have a single click tool for creating the sd card for it. Put sd card in, connect hdmi & power, job done.

According to this article:

Currently, rival devices such as the Beagleboard can be delivered to UK customers without undergoing the CE testing.

I have no idea what a beagleboard is and agree that they should have had it tested and certified regardless, but it wasn't a black and white case as far as the Pi team were concerned.

The Beagleboard I fished out of the returns bin at the Farnell trade counter in Leeds has got a CE mark, but no FCC mark.

Here's the latest quote from the guy who designed the Pi.



You can read his whole post here:

http://www.element14.com/community/peopl...-and-compliance