Secondly the pi line, like the Arduino line, along with everything else of that nature, are 'development platforms.' You don't buy a developer kit in mass intending to resell as an end product...

The pi, Arduino, beagle, etc have an excessive number of Io pins, often a variety of sensors, multiple options for power source and corresponding voltage regulators and other features useful for development. When you take your design to market, you simply draw your own schematic/PCB layout and place an order for a board without the unused sensors, io's, regulators, etc. The schematics provided for the pi and other platforms are quite modular and easy to understand. One can simply remove all the io's, accessories from busses, regulators, clocks, and whatever else isn't required. One can then tweak the PCB layout to fit their form factor, power requirements, component tolerances and whatever else to suit their exact use case, to minimize cost.

Beaglebone black-like system boards are used for both development & production. I almost shipped a system like this personally. With the BBB, you can get slightly modified versions, in bulk, and with enough availability to satisfy purchasing. Even if you couldn't, the SoC on the BBB can be bought from TI. I imagine there are similar things for more modern, powerful SoCs. Many companies like isolating the complicated timing sensitive board design (as needed for NAND and off chip memory) from the rest of their design, and don't necessarily have the skills in house to do that type of layout. Pre-made boards like the BBB provide a lower cost option than a completely custom board.

With the RPis, you can't even buy the SoC. I'm not aware of anyone you can buy slightly modified (ie: unneeded components pulled) boards from in bulk.

That's correct. Autodesk uses a modified Beaglebone Black for its reference "open" * 3D printer design Ember:

https://beagleboard.org/blog/2015-09-29-autodesk-ember/

The Ember retails at $7495, but one can experiment with the modifying firmware with a standard Beaglebone Black.

* "open" as defined by Autodesk.

The number of people who view PCB design as simple is significantly smaller than the number of people who develop for the Pi.

Complaining on HN is not the right solution here; if you genuinely want to fill a production run of thousands of devices with embedded Pi, speak to the foundation directly.

Making your own ARM board is fairly expensive for one-off things, but for volume it's not a huge obstacle. There are any number of vendors capable of producing these for you to almost any specification.

Edit: Looks like Farnell UK doesn't sell the Zero or Zero W, and any store that does is limiting themselves to 1 per customer.

It might be different in other countries, but I have to imagine that anyone in the US could get their hands on a Zero if they wanted one.

1 at $5.00 each, 2-5 at $9.99 each, 6+ at $12.99 each

A zero, sure. But I want like 20, haha. Unfortunately my nearest Microcenter is 600 miles away. If I can't get 20 for $100, they're not really $5 each.

Mandate a slightly higher cost-per-unit (5% or somesuch) and sell it as the Raspberry Pi Zero Commercial. With (critically) guaranteed compatibility with the consumer model. I can't imagine there wouldn't be Chinese factories lining up for the opportunity.

Then use the licensing revenue to help fund Raspberry Pi's mission.

I work in electronic device design. On a good project designing a complex device, 2/3 of the work will be iterations of design (software, electronic, mechanical, industrial). When you've finished designing it you've then got another 1/3 of getting it manufactured at a decent cost with a decent yield. And that's assuming that you have an existing relationship in China with a high end manufacturer. Weeks and weeks of trips out, production runs, implementation of QA.

Given that, how would you go about ensuring that what's coming out of a factory in China is actually up to scratch?

And even if you did, the processor on the Pi uses an unusual chip-on-chip technology that most board makers don't have the capability to build.

The difference here is that this is an already proven design? Have a competitive bid process for the right to market under the trademark, eliminate factories you flat out don't trust, then require ability to randomly sample and test on Pi's side.

If pass rate drops below agreed parameters, revoke the trademark and pick another factory. Then let the factory self-QA or face the consequences.

I can't imagine there are no factories in China that can make a profit on a chip that's being made profitably in the UK.

The only argument I can see against it would be that a commercial market doesn't exist in any volume. But I can't imagine there wouldn't be decent profit vs {insert lower priced Pi clone here}.

... The other argument would be if Pi has an exclusivity agreement with Sony UK (who I believe is the majority manufacturer) that precludes this kind of deal in exchange for pricing?

> "unusual chip-on-chip technology"

You mean the package-on-package mounting of the SDRAM on top of the BCM2835? Defer to your experience, but if Raspberry Pi is doing this then how expensive/rare is the investment?

I'm honestly trying to reconcile the following statements: "They're hard to manufacture" + "We make a profit on each one sold" + "We can't afford to make them at commercial scale"

Imagine producing 50,000pc/mo and having to do in-depth quality inspection on an sample size of, say 500pcs. Who is going to do the inspection? That's a lot of Pis.

How do you distribute? Do you ship a box of 50,000 to your warehouse and then mail them out 1 at a time to customers? Do you form retail partnerships?

How do you front the money for a 50,000pc order? Say they cost $3 per board, that's $150,000 up front in capital tied up for, say, 1 month of lead time, 2 weeks of transit time, and 1 month until you get paid by the customer - that's $425,000 at any time tied up in inventory (150k*2.5months). Do you negotiate payment terms with the supplier? It's difficult to negotiate terms without a history. Who does the negotiation?

What do you do when your box of 50k has an defect?

None of these issues are insurmountable, but there is a lot of necessary overhead that comes with mass production of anything. Deciding to MP Pis is an organizational decision that requires significant changes to company structure and resource allocation.

One of their aims was to produce things in the UK. They didn't meet this initially, but with scale managed to bring production at least largely within the UK.

Why?

What I'm getting at is, while QA might be hard I'm sure doing hard things might pay off some way instead of no way at all.

- Iterating through a project and seeing new functionality implemented can be rewarding

- The chase of business negotiations and inter-personal problem-solving can be rewarding

"Why" just strikes me as an off-topic, unspecific, and weird question to ask.

I apologize for striking you as off-topic.

Have you considered that the foundation's goals don't necessarily mirror yours?

Only true for Pi products. If you're willing to spend $7 you can get esp8266 nodemcu hardware delivered from amazon prime in arbitrary quantity.

Theres a problem where no matter how many times its stated, "you can buy ONE if you pay six times MSRP or live within driving range of the one retail store on the continent selling them or if you import from the UK or eastern europe" vs I need six for an IoT experiment involving micropython and I2C environmental sensors. I can't buy six $5 pi or even six $10 pi but I can have six ESP8266 boards by thursday afternoon for $35 total. Which is less than single pi zero plus "development kit of parts" costs, ironically.

A $5 computer that's impossible to actually purchase is much less useful in practice than a $7 computer delivered in two days by amazon prime in any quantity I'm willing to pay for.

Its interesting that above $40 or so there is an infinite supply of rasp pi 3 and beaglebone blacks. Everything is widely available except for cheap pi.

I'm currently working on an agriculture sensing iot platform and going to try to use rpiZero W with the current spectrum sensor from nanolambda. http://www.nanolambda.net . luckily they have the rPi api,sdk already.

https://www.board-db.org/

It's quite possible to get more, if you need to (though probably not multiple ones at the same time). Just use the shops the the RPi foundations recommends and choose to be informed via email when this shop again has some in stock. Then as soon as possible buy one.

Practically the Pi Zero-W would fulfil all our system requirements, it'd probably be better because it's smaller. However, we can't risk it because the supply chain isn't there yet.

I'm not expecting mass quantities on release, but the problem is that you can only buy one which makes development difficult. And if it's anything like the original zero, we'd have no idea when it will be available in quantity. Often it's useful to have more than one board e.g. one bare dev board, one in an enclosure/prototype. Having a month or more lead time for a replacement isn't practical.

because what your saying isn't actually true: https://www.nec-display-solutions.com/p/hq/en/news/dp/Produc... but they aren't the only people

Secondly, have you placed an order with them directly? Have you talked to a reseller?

the point still stands, have they talked to the rpi foundation about it?

You're far better off going with one of the Allwinner-based boards like the C.H.I.P if you intend to do this at some point in the future.

Unless I'm mistaken, the Beaglebone Black uses chips that are available for purchase from TI (and from which I think you can get datasheets, even as an individual), boots the mainline kernel, and provides CAD files for the board: http://elinux.org/Beagleboard:BeagleBoneBlack#Hardware_Files

Their founder and key personnel was/are employed by Broadcom. IIRC, essentially Broadcom thought of the Pi as a fun side project and were completely blown away by the demand.

The other problem is that people have written stuff, and loads of that, specifically for the Pi CPU/GPU - which makes a move away from BCM next to impossible:

- anything involving accelerated video isn't easily portable (e.g. omxplayer)

- anything that relies on a given special function of the CPU mapped to a specific GPIO pin might break with another CPU (e.g. extension boards)

- anything that relies on other rpi-specific hardware features (CSI, DSI) will be hard to port, but then again you don't really have a choice with non-usb/i2c camera or display modules...

> https://getchip.com/pages/chippro

who comes in question as manufacturer of the SoC (hint: Broadcom is not among them). Also according to

> https://github.com/NextThingCo/CHIP_Pro-Hardware/blob/master...

it uses a Mali400 GPU (not a VideoCore IV as the RPi).

Because of the cost? Or do they refuse to deal with small companies?

Rather: originally in support of the mission. :-)

http://spectrum.ieee.org/geek-life/hands-on/build-a-custompr...