From the open source RISC-V processor and Open Compute Project server designs to prosthetics and other items made with 3D printers, open source hardware is here to stay.
Producing free and open source software requires few resources except a computer and an internet connection. By contrast, open hardware requires not only components, but also the means to buy in bulk. For that reason, the spread of open hardware is decades behind that of FOSS. In fact, not long ago open hardware was considered impractical. It is only in the last decade or so that alternatives in social trends and hardware have started making open hardware practical — specifically, its proven success in non-profits, the rise of crowdfunding, and the availability of inexpensive microcontrollers and single-board computers, and 3D printing.
Building on FOSS
Open hardware could not exist without the prior success of FOSS. It has been twenty years since the Dot.com era, when FOSS was an untried idea. Since then, other groups based on the ideals and practices of FOSS, have grown into successful semi-independent communities of their own, such as Open Access and OpenStack. FOSS ideals no longer have to be proved, so open hardware does not need to be defended, either.
If anything, open hardware has gone on to have its own successes. Like FOSS before it, open hardware has an affinity with academia, where the exchange of ideas is a norm analogous to copyleft licenses. When academics venture into manufacturing, they are likely to organize under the same principles.
FOSS-based ideals are especially common in non-profits. Probably one of the biggest successes for open hardware is in the field of aesthetics. A traditionally constructed artificial hand costs upwards of $30,000. That price is beyond the reach of many families in a developing nation like India, where the average family income is about $21,000. By contrast, a custom-made artificial hand is sold by an open-hardware company like Open Bionics for $400. Although the cost of an open hardware hand is still high by the standards of developing nations, it is at least within reach, especially with charity. It also means, of course, that seventy-five open hardware hands can be made for the price of one proprietary one. Building on FOSS, open hardware has gone on to prove its own practicality.
Alternative Funding
Like FOSS in its earliest days, open hardware was initially of limited interest to established hardware manufacturers. That is not surprising; as Erich von Hippel has pointed out, existing companies are rarely a source of major innovation. However, this lack of interest also meant that would-be manufacturers of open hardware were enthusiasts, with neither the contacts nor the experience to raise private financing or venture capital. At best, possible open hardware manufacturers were small firms, focused on engineering rather than marketing and production.
Nor were many of the first open hardware ventures major innovations. Often, they were minor enhancements or niche markets — not at all the sort of products likely to interest entrepreneurs.
Crowdfunding has changed all that by providing an alternative source of financing. True, compared to venture capital, the amounts raised by crowdfunding are relatively small. So far, the highest amount pledged was for the Ubuntu Edge phone, and that campaign failed to reach its goal of thirty-two million. Amounts of over a million dollars are still relatively rare for open hardware, and most are for much less. However, that has not stopped the rise of small open hardware firms like Purism or Keyboardio, that have served as proofs of concept, showing that businesses can thrive on open hardware.
For these small companies, crowdfunding has become a way of life. Many return to crowdfunding with each new product. One advantage of this habit is the campaign to raise money also becomes a preliminary marketing campaign. By the time a product is funded, the new product has already attracted consumers and created a buzz on social media.
Admittedly, open hardware has yet to have produced the next Apple or Google. So long as it is financed by crowdfunding, it is unlikely to. But in place of a multi-national corporation, it continues to produce small businesses that are viable in their own ways.
Do-It-Yourself Manufacturing
A third boost for open hardware comes from trends that allow do-it-yourself manufacturing. Instead of developing hardware from scratch, open hardware manufacturers can rely on low-cost, off-the shelf components.
One of these trends is the rise of micro-controllers and single-board computers like the Raspberry Pi or Arduinos. Generally priced at under $60, Rasp Pis and Arduinos are available in a variety of models that can be quickly modified for a specific purpose. Both reduce the time spent on prototyping and production, reducing costs in a way that was not previously possible.
Another is the development of reasonable affordable 3D printing. In some cases, open hardware no longer needs to be manufactured, stored or shipped. Instead, all that is needed is to make the schematics available for customers to print their own. When repairs are needed, users can make replacement parts as needed.
Currently, parts made by 3D printer are mostly plastic, which can impose certain limitations. For instance, prosthetic artificial hands made of plastic are of little use in places like rural India, where open fires can melt them. Moreover, while a small 3D printer for plastic can cost under $4,000, depending on its size, a 3D printer that can use a variety of metals with the use of a furnace can cost $360,000 to set up. Still, 3D printers can be used in enough cases that they can contribute to the spread of open hardware.
Waitng for the Big Time
Open hardware is starting to attract more notice. One especially promising sign is that traditional companies like Google are getting involved with the development of the open hardware RISC-V chip. Part of this interest may be due to the increased likelihood of tariffs driving up the cost of proprietary chips.
What holds back open hardware is not only name recognition, but the fact that those paying attention are often new entrepreneurs, with few contacts among those who manufacture for vendors. Many manufacturers are uninterested in the small production runs of open hardware vendors, and in Asia, quality control can be an ongoing problem. Those interested in open hardware should read the Keyboardio blog for a summary of the setbacks that beset open hardware production. Usually, they start with an explanation of what open hardware is, and why anyone should care.
Still, open hardware is succeeding, case by case. However, it is taking a long time getting anywhere. Its present status could be compared to that of free software about 1998: it has proved itself to a knowing few, but still remains unknown except in scattered circles.
Bruce Byfield has been involved in FOSS since 1999. He has published more than 2000 articles, and is the writer of “Designing with LibreOffice,” which is available as a free download here.
Overall, your article is encouraging, thanks! The only negative I see is the involvement of Google in hardware development…
I don’t know where to begin.
Purism is NOT an open hardware company. Not even remotely close!
The Raspberry Pi isn’t open either. It has a locked bootloader with proprietary firmware that requires a special built kernel to operate and the wireless/bluetooth can’t operate with fully free firmware/drivers.
A lot of single board computers (most but not all raspberry pi clones and others) have non-free components or require manufacturer provided kernels, meaning you can’t update when you want, and you are tied to the manufacturer’s support or lack thereof for new OS images.
Just because a system can run gnu-linux doesn’t mean it is open hardware.
Open hardware means supporting fully libre software/firmware and providing schematics such that anyone can replicate the device.
There are a handful of truly open hardware companies out there, but giving credit to the wrong companies is harmful to the ones actually working to help free hardware.
There’s also a tendency for companies to claim ‘open hardware’ regarding their boards (sometimes even providing schematics) when they contain components which are most decidedly NOT open and WON’T run FOSS. The situation is pretty sad, actually. It’s similar to the situation in software, where companies will try to claim their non-FOSS license is actually FOSS. Openwashing is rampant.
Trying to learn from manufacturers what parts of a single board computer are open or closed is like pulling teeth. No one wants to list this information or talk about it.
Purism is working towards open hardware. Its freedom roadmap has not been updated for a while, but as of a year ago, the company described its laptops as 99% free at the bios level. It is obviously working towards open hardware (https://puri.sm/learn/freedom-roadmap/).
As for Raspberry Pi, you are correct, and I’m surprised that I didn’t specify the fact. However, products produced with the Pi are often free in themselves, aside from the reliance on the Pi. If Debian had a hardware repository, they would be in the contrib section: neither free nor non-free.
Sorry Bruce, but you are quite wrong on this one.
Being able to run libre software is insufficient to be open hardware.
Purism does not make their hardware designs available, nor are they reproducible. Ergo, not open hardware.
Beyond that, their freedom roadmap is a nice bit of marketing, but little more. Essentially Purism laptops are no more libre than any laptop running coreboot (not even libreboot, because it has proprietary blobs in the BIOS). They are still Intel systems with proprietary firmware and are not 100% free of Intel’s secretive management engine. Nor are they likely to ever be anything more free than that. Neither are they “designed chip-by-chip, line-by-line, to respect your rights to privacy, security, and freedom” like they claim. Do some research into the bluetooth chip on Librem 15 laptops…the one they used to mention in the system spec but no longer do (because it doesn’t work with non-free drivers, a fact called out by their customers). How did this chip get in there if it was designed for freedom? Ask yourself if it is more likely that they paid for an existing reference design and just used that (they admit this, elsewhere).
With regard to the Pi, the analogy to Debian’s contrib repo is a good one. All those free projects on the Pi depend on the not-so-free Pi, just as free software in the contrib repo depend on non-free software to function.
The Pi does have schematics available, but contain components that can’t be sourced by an individual. As I understand it, Broadcom engineers were a part of the original design team and so the Pi uses a lot of Broadcom components, regardless of their proprietary nature.
A slightly more-free clone of the Raspberry Pi is the Tritium board from https://libre.computer
They make schematics available and the components are sourceable by an individual. The bootloader is unlocked and is supported by mainline versions of uboot and the linux kernel (no blobs!). It does lack a wifi/bluetooth adapter, but is otherwise a better Pi than the Pi itself is.
They produce other boards, but as far as I am aware, only the Tritium has mainline support. The only negative is the CPU is from AllWinner, who is hostile to open source in general and a habitual GPL violator (and linux foundation member!), but the CPU has been reverse engineered by the SunXi community, https://sunxi.org for inclusion into the mainline linux kernel.
Correction,
When I said CPU above, I should have said SoC. Also, to avoid creating a false impression, it is worth pointing out that while a number of Allwinner Soc’s have been largely reverse engineered, there remain bits of functionality that aren’t there yet (audio over hdmi is one such feature listed as non-working on the sunxi site)
Another correction,
When I said “because it doesn’t work with non-free drivers”, obviously I meant “because it only works with non-free drivers”.
RISC-V , interesting; wasn’t this the processor family of many UNIX systems around 30 years ago?
No, Richard. I think you’re thinking of MIPS, another reduced instruction set computing architecture, which was also made open source earlier this year. RISC-V came out of a recent project out of US Berkeley, and has only been around about five years. Here’s a link to my most recent article on the subject that’ll get you up to date: https://www.datacenterknowledge.com/hardware/companies-pushing-open-source-risc-v-silicon-out-edge
Bruce’s comparison of the Raspberry Pi to ‘contrib’ software in Debian got me looking for open source hardware definitions that distinguish between ‘contrib’ and ‘free’. I couldn’t find any. This is unfortunate.
For those unfamiliar with Debian, Debian divides software into three categories:
free – is FOSS
contrib – FOSS that depends on Non-FOSS
non-free – Non-FOSS
To continue the analogy into hardware:
free – open hardware
contrib – open hardware that depends on closed hardware
non-free – closed hardware
The Open Source Hardware Association has certification programs for hardware to be listed in a registry as ‘open source hardware’, but it makes no distinction between those which are purely open, and open but containing closed components. They cite the difficulty in obtaining purely open designs, but it seems to me a separate designation for purely open hardware would be useful and desired, even if only to show us how little hardware falls into that category.