Welcome, Crowd Supply backers!

Thank you for your interest in supporting this project. Here is where you can find out a bit more in-depth about what's going on.

Open Development Process

This project has been extremely unusual in that it has been a Libre Hardware and Software project right from the beginning. Here is a comparative list of other similar laptop projects:

  • The Ben Nanonote was developed publicly, with a wiki, mailing list and IRC channels used by the developers. More details can be found at Paul Boddie's page here
  • The Novena crowd-funded campaign released PCB designs after they were finalised (but prior to launch), and uses a processor that is fully and actively supported by the Libre community. However it is extremely hard (if not impossible) to find open source tools for the Novena's on-board FPGA.
  • The Pi Top promised to release educational CAD files after it was completed.
  • The OpenPandora released software for early developers prior to launch but did not (and still does not) release PCB or CAD designs
  • The Olimex DIY Laptop released the PCB files under a GPL license (only once completed) but the casework, whilst available to buy (via a single-source China supplier), is proprietary, copyrighted and may not be modified, reproduced or repaired. Additionally, the software used in the Olimex DIY laptop is based on a copyright-infringing SDK and requires a proprietary bootloader.

So by evaluating all these projects and many others (OpenMoko, GTA04, Neo900) we can see that this one is unique in that not only is it intended ultimately to reach mass-volume markets (as opposed to low-volume niche markets) but it truly respects Software Freedom, invites others to collaborate, comment, review, learn and teach in both the software and hardware fields, but more than that, unlike for example Google Project Ara's MIPI Unipro it is based around a patent-free, royalty-free open standard that itself was designed right from the start following a fully public, fully documented, open invitation, open collaborative and open consultation process.

From a commercial perspective, at least one key strategic part of the project needs to be kept proprietary (the Computer Card's PCB CAD) whilst the financial investment so far is recouped. Absolutely everything else has been available as GPL-licensed source code (including all CAD files) right from the start, and, once the financial investment in this project is paid off, the Computer Card's PCB CAD files will also be released as libre hardware-licensed files as well. Here is what's available right now:

  • EOMA68-A20 Schematics and PDFs contains the complete set of datasheets for the EOMA68-A20 Computer Card
  • A20 early boot code contains the very early bootloader code for the A20 processor.
  • A20 u-boot source contains the complete u-boot sources needed for the EOMA68-A20 Card
  • A20 linux kernel 3.4 source contains the complete linux kernel sources needed for the EOMA68-A20 Card
  • Micro-Desktop Schematics contains the schematics for the micro-desktop and other files
  • Laptop Schematics and PDFs contains the schematics for the laptop, datasheets and other files
  • Laptop 3D CAD Files contains the CAD Files for the laptop casework (youmagine instructions link below).
  • Parabola ARM GNU/Linux a guide as to how Parabola was installed on the EOMA68-A20 Card. This was quite awkward and required downloading a live boot parabola x86 CD which was then run under qemu which then ran qemu-arm within qemu... but this was successful. Link is provided for reference, so that Parabola GNU/Linux sources and packages may be reviewed if desired.
  • Debian ARM GNU/Linux - this link is provided for reference so that the sources and packages may be reviewed if desired.
  • Laptop STM32F072 Firmware is available under the GNU GPLv3+ and requires libopencm3 (an LGPLv3+ licensed embedded controller firmware library).

There are a number of independent and openly editable news pages for each of the projects:

  • The Laptop: it can be seen that the first prototype laptop casework was completed as 3D-printed last year. The first prototype PCB set was also assembled and tested with the battery, keyboard, LCD and 4.3in CTP/LCD last year as well.
  • The Micro-Desktop has likewise been successfully prototyped.
  • The EOMA68-A20 Computer Card also has its own page

Videos of all of these can be found on youtube and are tagged EOMA68. As with all Libre Projects there is also a Mailing List and an IRC channel: #arm-netbooks on freenode.

Risks and Challenges Section

This has been an extremely comprehensive and long-running project where an extraordinary amount of attention to detail and learning has been undertaken over the past five years, all of which is documented publicly and may be reviewed online at the arm-netbooks mailing list archives and the Rhombus Tech web site.

Whilst it is literally impossible to comprehensively list here the full set of risks and challenges already faced and dealt with, a small sample can be given here, including a strategy to deal with each of them.

Timezone communication lag and language barriers

One of the very first problems immediately encountered when dealing with China was the 48-hour round-trip on sending and receiving email. Several weeks could be spent just establishing that a supplier did not even have the right part: multiplied by 100 suppliers that needed to be contacted. It has taken several years to find good factory owners who speak english, but often the suppliers do not. Therefore the solution is very simple: move to Taiwan, visit Hong Kong and take trips across the land-bridge to Guangdong.

Component sourcing and supply

For westerners it is bewilderingly and freakishly difficult to source components in S.E. Asia and to find datasheets. Bunnie Huang documents this with his "Gongkai" posts Not only that but components can go end-of-life without sufficient notice, or may have pricing that's well above acceptable levels.

Significant time was therefore spent sourcing the components, searching online for datasheets, ensuring that they are mass produced. Where unusual components are single-supplier only, the relationship with the supplier has been kept up-to-date on a very regular basis to ensure that they continue to stock them.

Also as hinted at, above, there is actually a significant amount of fraud that takes place involving components. Capacitors and Hard Drives can be found to be lower spec than stated on the outside: 3.5in HDDs it turns out are perfect for hiding smaller 2.5in HDDs inside, and 100uF electrolytic capacitor cases are perfect for hiding 10uF. This isn't a joke! it genuinely happens! So, being actually on-site at a warehouse or supplier, and using a factory owner whose suppliers are long-term reputable is critical. Hence why it is necessary to travel to S.E. Asia.

EOMA68 Standard Development and finalising

The EOMA68 standard is strategically critical to the entire project, where this crowdfunding campaign is just the first phase.

Many standards are designed with "optional-itis" inherently built-in. Options on standards have proven time and time again to be the nail in their coffin even before they are launched. An extremely comprehensive review was done of alternative standards (both modern and historic) and documented here

Key to EOMA68's longevity and long-term future is the absolute golden sales/marketing-esque rule "Just Plug It In, It Will Work" combined with an automatic hardware-level "speed negotiation" strategy. NOTHING is allowed to be "optional": all the interfaces selected have been around for several decades - USB, SD/MMC, RGB/TTL, I2C, RS232 and so on.

Over five years since the standard was first created, four painful decisions had to be made to revise the standard and to write off the investment of time and effort in CPU Cards that had been created conforming to older revisions. Each time the revisions were made to ensure that more of the mass-produced low-cost SoCs (http://www.webopedia.com/TERM/S/SoC.html) would conform easily to the EOMA68 standard, whilst also allowing room to expand to ever-faster SoCs over the next decade and beyond.

In short: EOMA68 has been extremely carefully designed, tested and analysed, and is now set not in stone but in diamond.

PCB CAD development and prototyping

Unlike software development, PCB errors cannot easily be corrected, often requiring a complete re-run if even a single mistake is made. It turns out that many PCB production companies will permit a few tracks to be cut and wires soldered in place for a production run but this should not be relied on as it can cost.

The campaign was therefore not even launched until the PCB prototypes were confirmed as operational, with only minor errors requiring only a low-cost second prototype run to confirm that they had been corrected.

Whilst the laptop and micro-desktop PCBs are extremely simple and straightforward (2-layer), the EOMA68-A20 CPU Card is a 6 layer extremely densely-populated and small PCB which costs $USD 1700 to make 5 samples each time. There exists a proven older design which is slightly more costly, but it is considered prudent to update it to use faster, cheaper and easier to obtain RAM ICs. Given that there is a fall-back design, taking the decision to update to better and cheaper RAM is easy and risk-free.

Closed design vs Open design

One of the most annoying aspects of many crowd funded projects is the "promise" to be "open" with schematics, design details and many more absolutely critical details... AFTER the product is finalised and shipping has begun. In such projects, isolated from the users - many of whom will have significant expertise - mistakes are made that could actually jeapordise the success of the entire project, or, much worse than that, the users end up being seriously unhappy with what is actually delivered, because secretly a design change was made without consulting or informing them.

In complete contrast to this extremely risky strategy, the entire EOMA68 project has, like the Ben Nanonote, OpenPandora and many other open projects, been a public open project from day one, when it began almost five years ago. For example, the schematics have been made available for the majority of PCB designs, and design reviews undertaken publicly on the mailing list, and the CAD files for the 15.6in Libre Laptop dock have been publicly available under the GPLv3+ license right from the beginning.

Learning from both the mistakes and the good parts of the OpenMoko, OpenPandora, Ben Nanonote, Pi Top and many others, everyone has and continues to be invited to comment and participate and, crucially, to point out potential mistakes before they could become serious problems (technical or logistical), on the project's mailing list

This strategy of being completely open has several additional benefits for end-users, not least is that they are guaranteed to be able to effect repairs down the line (just like the early very first Personal Computers, which always came with full hardware schematics): in this case the entire product including casework can be replicated and even potentially improved on, down the line.

Laptop hinges

Sourcing laptop hinges has proven to be extremely difficult. A decision therefore had to be taken to construct one in a way that is easy to repair and replicate with off-the-shelf parts, using washers, M3 and M4 nuts and bolts. However, after a year's work this is still not finalised: steel washers have too much friction, so a test will be carried out using brass washers instead (which are self-lubricating).

One of the reasons why Taiwan was picked as the first S.E. Asia country to go to is because it is much more likely to be able to find a supplier there, and to test on-site the parts that their factories make. So whilst it would be nice to be successful in designing an end-user maintainable laptop friction hinge, made from off-the-shelf parts, if that turns out to not be possible then a single unit will be sourced and used instead.

Keyboards and trackpads

It is surprisingly difficult to find keyboards suppliers. Many of them have near-cartel-level relationships with the large laptop manufacturers. So again, here, actually going over to visit them, get the datasheets so that the keyboard does not have to be disassembled and reverse-engineered, is all crucial.

Trackpads are also ridiculously difficult to source: a google search "trackpad supplier" is deluged with hundreds of thousands of enterprising alibaba suppliers for Apple and Dell laptop products. This is one reason why a 4.3in CTP + LCD combo was chosen instead of the more traditional trackpad, not least so that buttons (and scrollbars) can actually be displayed on-screen, and the buttons changed (1, 2 or 3) to suit the end-user's personal preferences.

CE and other Certifications

The cost of FCC, UL and CE Certifications is anywhere between $USD 3,000 and $USD 15,000 depending on who does the Certification, and what is actually required. Batteries are an entirely separate matter, costing an additional $USD 10,000 and up, plus 20 samples to be destruct-tested, in order to get safety Certificates when installed in the product. They do however always, always come with a Certificate from the Manufacturer that says that they are safe to ship.

These Certifications simply cannot be paid for unless the project goes well beyond the minimum threshold set of $USD 125,000. Stretch goals therefore have to be set which include Certification for both the products themselves and the battery for the laptop when it is installed in the product. It's just how it has to be.

"Being overtaken" by "The Big Boys"

The risk is: this project will become irrelvant because a large manufacturer will go, "that's a good idea, let's replicate it and accidentally on purpose destroy this small one in the process".

The strategy to deal with this has to remain confidential for now, but it is worth mentioning that some of the large companies have actually been contacted (who must remain confidential), and they have all said that they cannot make enough profit from mass-volume modular computing. In essence, by creating an Open Standard the risk for them is that their competitors will undercut them. Look at the Motorola Atrix Laptop Dock: its first revision you could buy for $USD 70 and plug in a USB-HDMI Dongle PC with some cables. For only $USD 120 you could assemble a really good laptop. The second version of the Motorola Atrix Laptop Dock was therefore DRM-locked.

From this we can learn that there is a simple financial DISINCENTIVE for mass-volume manufacturers to pursue collaboration. This bizarre situation which leaves the entire market wide open for persistent and motivated eco-conscious dual libre hardware-software entrepreneurs to exploit is covered in more detail in the ecocomputing whitepaper: http://rhombus-tech.net/whitepapers/ecocomputing_07sep2015/

Summary of risks and challenges section

Incredibly this genuinely is just a small sample of the risks and challenges faced already, and gives a near-overwhelming glimpse into what's involved in product development that you just don't normally see. Key thing is: got any concerns or expert advice, please get in touch.

Pledge Levels and costs

As this is a transparent campaign we're doing something that's really not been done before: publishing the parts costs so that people wishing to help verify things or perhaps at a later date do a group buy for components know what to expect.

15in Laptop

  • Battery: $8 (MOQ 250)
  • 4.3in CTP+LCD: $8
  • 15.6in LCD: $30
  • Keyboard: $10
  • Main board PCB1: $25 (MOQ 250)
  • EC board PCB2: $12 (MOQ 250)
  • Power board PCB3: $15 (MOQ 250)
  • 3D-printed casework: $35 (network-3d-printed in china)
  • 1.5mm plywood: $10
  • LVDS Cable: $10 (MOQ 250)
  • Cables: $8
  • Charger: $5

Total parts: $USD 166. hmmm, that's a lot lower than $500, isn't it? how does $166 turn into $500?? turns out that that's quite easy...

Then we need the following fees and costs:

  • Testing: $20
  • Package hand-assembly of kits: $50
  • 12% for marketing and crowd-funding
  • 3% for credit-card company
  • $2.75 packaging and handling fee
  • US Freight costs (estimated $50)

So, we're now looking at around $120 in extra costs.... times 3% times 12%.... that comes to $404. suddenly $500 doesn't look like it's very much, does it! We also have to do make a second run of prototype PCBs out of the $100 spare per unit, cover expenses whilst in Taiwan including plane flights...

EOMA68-A20 Computer Card

This one's a lower unit cost but these are all MOQ 250. Some of the parts are actually MOQ 30,000 however I've negotiated a strategy of ordering 250, then 2,500, then finally 30,000 within a year.

  • A20 CPU and PMIC: $7
  • 8gb NAND: $7
  • 2x 1gb DDR3 RAM ICs: $8
  • Discrete ICs and parts: $5
  • PCMCIA housing: $1.50
  • Bare PCB: $1.50
  • Micro-SD: $0.20
  • Micro-USB: $0.10
  • Micro-HDMI: $0.50

Total parts: $USD 30.70

Then we need the following fees and costs:

  • Testing and assembly: $5
  • 12% for marketing and crowd-funding
  • 3% for credit-card company
  • US Freight costs (estimated $5, the Computer Cards only weigh 40g!)

So now we're up to $50... so we should really be doing pricing at around... $65 just to be safe.

Micro Desktop

The Micro-desktop is really very simple, it's mostly connectors. There are still over 100 components on it however because of the VGA conversion which is done by hand using a resistor-bridge

  • Bare PCB: $2
  • VGA connector: $0.30
  • 2x USB connectors: $0.50
  • Power Jack: $0.30
  • PCMCIA socket: $2
  • VGA circuits: $1
  • Power Management ICs: $3
  • Micro-SD: $0.20
  • Other bits: $1

Total parts: $10.30

Then we need that beautiful casework, which is around $12 for the plywood to be laser-cut. Then we need:

  • Testing and assembly: $5
  • 12% for marketing and crowd-funding
  • 3% for credit-card company
  • US Freight costs (estimated $10)

So now we're looking at $39, so really we should have the Micro-Desktop at around $55 to be safe.

USB-OTG cable kit

This is still to be estimated. These are costs on e.g. Amazon:

  • Micro USB-OTG "Powered" Hub: $10
  • 5V 2A PSU: $5
  • Micro USB-OTG Cable: $5
  • Micro HDMI Cable: $5

So it would be reasonable to assume a comparable cost, but for safety we should be adding a margin... yeah this one really needs to be properly assessed, so it can be added later during the campaign.

Questions

Here's some of the great questions we're receiving, along with full answers.

Can I Put in WIFI? Is it MiniPCIe or USB?

great questions: we're going with USB, these are $7 SoCs, they don't have PCIe, that would require a $20 SoC and a completely different standard to be able to have the PCIe pins brought out.

the laptop enclosure has INTERNAL space for two USB ports, around 90mm x 60mm x 8mm high. so you can put in a USB-WIFI, USB-3G/4G, USB storage etc. without them getting knocked off and damaged. there's a compartment lid in the near left hand corner.

USB WIFI adapters you can get from Think Penguin, there's even a modern atheros-based 802.11n one where the full source, thanks to work by thinkpenguin, is available including the on-board firmware of the chipset. which is awesome.

More info from Think Penguin.

Are you planning to do an 11in laptop?

yes we will look at doing an 11in laptop later: that was actually what i wanted to do, initially, because the power consumption would be lower with a smaller screen - it draws 50% more power than the processor does! however, chris from ThinkPenguin, the sponsor of the project, convinced me that doing a 15in would actually be better for people to be able to do office-style work on, not least because a full-sized keyboard with a numberpad is way more convenient. remember also, this laptop doesn't suffer from the mass compounding problem: it only weighs 1.1kg, and that's with the LCD coming in at a whopping 400 grams!

normally a laptop this size would have to be 2kg or above because it has a 15 to 30W intel processor, which needs a fan and a heatsink. those are made of metal, that means extra weight. you also have a hard drive, those are heavy too. now you have to beef up the plastic to cope, and pretty soon it's all gone quite mad and you're up to 2kg very very quickly. by contrast there's no fans because the Computer Card uses only 3.5 watts. there's no HDD so that weight's gone. the battery can be a 10Ah light-weight E-Bike battery (yes, really!) weighing only 200g.

now, to do a 11in laptop i now see, from having done this experience, would be a new challenge on its own. it would need a complete new design of the casework (which took a year!). space would be challenged... and so would i, which is what i love. but, bottom line is: it would be an entirely new project, and we want to see this one succeed first.

The A20 has MALI, a non-free GPU. Have you sorted out MALI somehow?

we haven't - that's going to be down to either ARM to make their minds up to provide the sources (and you can help there by contacting ARM by telephone and taking up the receptionist's time by asking to speak to someone about why they haven't provided the source: the more money that they lose by having to pick up the phone the better), or for someone to put the funds up for the three years necessary to keep Luc Verhaegen full-time employed to carry out the necessary reverse-engineering.

it may surprise you to learn (as it certainly surprised me) that we can still apply for RYF Certification on the Allwinner A20, as many other people have tried to do in the past. the basic criteria that the FSF are looking for is, "how easy would it be for the average end-user with no significant computing experience or knowledge to ACCIDENTALLY install proprietary software without their knowledge i.e. with no warnings from the OS about the potential dangers of the action that they are about to take?" in other words, Debian GNU/Linux may never receive FSF Endorsement because when you run synaptics, the package manager GUI presents you with incredibly easy options including "non-free repository" and does not tell you or warn you "hey you could end up being spied on or your privacy violated simply by arbitrarily clicking this checkbox".

so in our favour are two things: firstly, we have no non-free co-processors or peripherals. the STM32F072 embedded controller has full GPLv3+ licensed firmware and has been run as a truly libre project right from the start. you can check out the repository, you can see the commit logs and you can see the discussions on the libopencm3 mailing lists. all other projects presented to the FSF using Allwinner A20 SoCs have failed at this point because they have included embedded WIFI ICs on-board which require proprietary firmware to operate. in direct contrast to this we are using USB for WIFI so you can get products from ThinkPenguin that do not need proprietary firmware blobs. you can buy even modern Atheros 802.11n WIFI dongles that even the complete source code for the firmware in the underlying Atheros chipset, thanks to ThinkPenguin, which is amazing!

secondly, it turns out that the MALI GPU is entirely memory-mapped. if you try to do "lsusb" or "cat /proc/cpuinfo" or any other kind of exploration from userspace, if you haven't compiled up mali.ko you LITERALLY cannot even SEE the MALI GPU from userspace. guess what we will not be doing? :) we will not be adding CONFIG_MALI=y to the linux kernel build process. thus, it becomes literally impossible for the average end-user to accidentally end up installing the proprietary non-free MALI 3D GPU code. this is why we believe that the special exemption being applied for will be approved.. but ultimately that has to be a decision made by the FSF.

but, given that we are targetting office, email and internet usage for these computers, the lack of 3D is genuinely not a problem. you don't need 3D graphics to edit a spreadsheet, in other words. that's not to say that technically-aware people cannot compile their own source code and include mali.ko: both we and the FSF recognise that people have that absolute right to do that. we just don't endorse it, that's all.

What are the differences between the Libre Tea and the Practically Perfect?

ok it's down to the FSF's rules when applying for RYF Certification, which when explained to me made perfect sense.

basically think of it from an average end-user's perspective, you're looking around, you want to buy something that won't spy on you, you hear about this thing called "mumble, mumble, mumble", you go to the web site, you click "buy" and when it arrives it has Debian GNU/Linux or worse Ubuntu where your screen is sent off to Amazon when you do a local search and they do OCR on your private desktop and start sending you product adverts based on CONFIDENTIAL FILES.

turns out that there was a drop-down menu that you were supposed to select and you didn't know that, and ended up paying for a proprietary OS.

now contrast that with this scenario:

you hear about this thing called "Libre Tea Computer Card", it's a funny name, it's unique, and crucially it's a COMPLETELY different name from the product that does NOT have RYF Certification, you click "buy" and you get what you were expecting to buy.

in other words, the names are there to make sure that the average end-user doesn't get the wrong OS by mistake.

But why sell a version without the RYF Certification?

Very interesting question. This project is the beginning of a mass-volume OS-agnostic standard, EOMA68, where we recognise that it's only going to make a huge environmental difference if it becomes mass-volume rather than niche. We can always give people the option to choose to easily transition to RYF-Certified products if they're easily compatible at the hardware level. Right now, as things stand, there is no choice. Try buying a laptop today that truly respects your freedom - feel free to explore that space... when you have, come back and we'll be able to help.

Is JTAG available?

On the EOMA68-A20 CPU Card it's available via multiplexing on the 6 SD/MMC pins - you can look them up and get a breakout board.

Is there Linux mainline kernel support for the A20?

Yes - more info here but not all the hardware is supported. Linux 3.4 is what supports all the hardware, reliably.

I'm missing Ethernet and SATA on EOMA68, and what about power consumption?

ok, this is a really good question. to answer it, there's quite a lot of history which might be expected: please bear with me while i go through everything: it's related to what low-power SoCs are available now, and what are likely to be available in the future, as well as what would be reasonable to support in a low-power system. remember that EOMA68 has to be both lowest common denominator as well as have a working life spanning at least a decade.

so the original version of the EOMA68 standard did in fact have both ethernet and SATA, and only one USB port. however over the next two to three years we learned that the number of processors that could be accessed that were entirely libre, non-GPL-violating, affordable and accessible without NDAs was exactly two out of almost a hundred processors world-wide. if however we removed SATA that number went up to around five, and if we removed ethernet it went up to around ten. so we had the option of having a near-irrelevant but powerful standard, or a relevant but less powerful one.

the rather painful decision was taken in two stages to make EOMA68 more relevant, replacing SATA with a second USB port and then replacing Ethernet with first USB 3.0 and then USB 3.1. we figured that as USB 3.1 is what... 10gbit/sec it would provide a much more powerful upgrade path in the future than sticking with Gigabit Ethernet and SATA, when even affordable Micro-SD cards available today are exceeding the kinds of transfer rates of the fastest available SSDs from only four years ago.

you also have to bear in mind that these low-power processors are only $2.50 to $7 - there's even a 64-bit quad-core processor for $4! they're not even anything remotely approaching the cost of Intel's tablet-style processors which come in at around $21 and above, which probably goes a long way to explaining why Intel has shut down its smartphone and tablet division. so the irony is, if a $2.50 processor hasn't got SATA or Ethernet, it has to be added with a slew of components: $1.50 for USB-to-SATA, $1.50 for USB-Ethernet and another $1.50 for a USB Hub IC because you used up the only 2 available USB ports of the processor, otherwise. add that up: you've got $4.50 worth of peripheral ICs to add to a $2.50 processor and you're now cost-incompetitive versus a $7 processor which has SATA and Ethernet built-in! not only that but you've exceeded reasonable power-budget expectations as well.

... is this starting to make sense? :) it's about what i can reasonably and easily get hold of, what can be kept within a reasonable power budget, and many other factors.

so what we've done with the Micro-Desktop is, we kept it to the bare minimum. there's a small internal header if you want to do arduino-style experiments. the two USB ports from EOMA68 are put out directly via USB 2.0 sockets. the video output is converted to VGA. the SD/MMC is connected to a Micro-SD socket. all of that keeps the cost down. there's a 15 watt power management chip on-board which is more than enough to cover powering the Computer Card and the two USB 2.0 sockets. it's possible to plug in say a wireless Logitech keyboard-mouse combo into one port and a USB-Ethernet dongle or WIFI dongle into the other, or for anything beyond that you can plug in a USB hub.

for the Laptop it's a little different (or there is more to take into account). we considered putting in USB-Ethernet and USB-to-SATA but the amount of power used would have had to be increased by just under 10 watts to do it, including the power draw of the HDD or SSD. we're now up to a 25 watt budget. that is way beyond what a simple single-cell low-cost power management IC can handle, and it's also way beyond what a single cell battery can handle: the batteries are 4.2v and it would actually start to be of some concern to be drawing 6.25 amps continuously from a single-cell lithium battery.

so to add in USB-Ethernet and USB-to-SATA we would have to take out the $10 Power PCB and replace it with a $25 dual-cell or triple-cell Power PCB, add in fans (which could be a mechanical point of failure), redesign the case to house two (or more) batteries, add in $10 worth of extra ICs, and it quickly becomes a cascade-effect that in effect is a completely different project with at least double the financial budget required of the simpler design that we've come up with, today.

that's not to say that we're not going to do it! we just feel that it would be better for us to do this one, first. in the meantime, there are three available USB ports on the laptop: one of them is external, but the other two are internal in an easily-accessible compartment. you can put in a USB storage and WIFI dongle INSIDE the laptop, where it will be safe from being knocked off. also it's possible to put in two of those ridiculously-fast Micro-SD cards i mentioned, which go up to something like 256GB each at the time of writing, and can do 80gbytes/sec transfer speeds. pretty amazing, but also it means that having a SATA drive is not so critical.

in short: we have to learn to walk before we can run, but i believe you can see that we have thought this through very carefully, and provided some alternatives.

You're using raw NAND, isn't that supposed to be totally unreliable?

Original question asked and answered on CNXSoft It's actually very common to use raw NAND, then to have good error correction and wear levelling, which the ubifs filesystem (the successor to jffs2) does very well. After learning that MLC can actually corrupt itself just by reading (not writing!) we will be considering putting in SLC (single level cell) NAND instead.

But, the A20 is considered to be "unbrickable". Basically if it all does go belly-up (which is unlikely given that the internal boot rom just keeps marching through NAND blocks looking for an expected boot signature), you can always boot at absolutely any time from a specially prepared MicroSD card.

What other OSes could run on the EOMA68-A20 Computer Card?

The list is huge! Fedora, CentOS, FreeBSD, ArchLinux and Parabola, Debian-based derivatives, Android, and even the L4KA micro-kernel is up and running, so l4linux and l4android both work. The A20 really has been around long enough to have this level of comprehensive OS support.

Why create a standard instead of using an existing one or a common devboard?

This extremely good question came up on Phoronix but basically... there aren't any, so we had to make one! The EOMA introduction does a quick evaluation of all the available standards that have been found so far: none are suitable. And if you put a devboard into your pocket, it's going to get destroyed very very quickly. This is why we picked legacy PCMCIA because you get a robust metal case.

Why don't you use (insert my preferred SoC here) for EOMA68?

The evaluation process for making a hardware project is a bit of a shock to anyone who is used to Software Devlopment or any kind of general Project Management. Outlined here is an example where someone suggested to use a qualcomm processor because its 3D GPU supports freedreno (a libre 3D graphics engine).

Now, don't get me wrong: I'd love to use (insert your preferred SoC here) but I've been at this for five years, I have some ethical standards that are not up for sale, I've been around computers continuously for almost forty years now, and I have the ability to assess things in a bit more comprehensive depth so that I don't fall into many of the pitfalls that many projects I've seen come and go.

That doesn't mean that I can do everything myself, so if you see a good SoC please do contact me and tell me about it, I'd like to do an evaluation.

What can we do to help (other than pledge)?

Thank you for asking! Join the mailing list, tell your friends, co-workers, post on social media, go on to various free software mailing lists (debian, fedora, arch linux) and let them know that this project is available, emphasise that it's upgradeable and that the cards can be used as low-power co-located servers (there's companies that do this already), that sort of thing. GNU/Linux just reached 2.5% desktop share so it's not that unusual any more.

Also: although these devices will go out as fully-working computers, there will be quite a lot of software improvements that can and will need to be made, not just to support EOMA68 but also to improve the mainline linux kernel support for the A20 for example, which currently doesn't support LCD screens (only HDMI - see question above: it's why the Computer Cards are going out with a Linux 3.4 kernel).

So thank you for asking!

There was a lot of fuss about Allwinner security in the news, what gives?

A really good and informative article is on Arstechnica. So, basically, all the GPL-violating products which were supplied to Chinese factories with binary-only images that went out a few years ago were totally vulnerable. No surprise there.

However we are not going to be distributing binary-only GPL-violating images, because that would be kinda dumb - and completely against the ethical standards that we have committed to. We're only distributing license-respecting software, so that you and many others have the absolute and fundamental right to review it, help each other to fix it and improve it, and generally make it easier for everybody to use and trust. That's how Software Libre works: mutually beneficial collaboration. And yes, we are prodding Allwinner. With a sharp stick.

Will the PFY (and PIY) Laptop come with build instructions?

I'll be using the first few assembled units to make very clear videos, photographs and precise build instructions. The example that i will be following is that of Chris Palmer's "mendel90" build instructions (with the addition of videos), as Chris is one of the world's top most competent 3D printing engineers.

I'll also be available on the mailing list and IRC channel for questions online, and i have one person in Europe who is willing to do hands-on workshops showing people how to assemble them.

Plus, the mailing list will be available for people to help each other out, as well. The idea is that you will be part of a community that doesn't just repair your own electronics, but knows how to build your own electronics. This project goes well beyond "Right to Repair", it's the "Right to Build"!

I'd like to add a 3rd monitor, how can it be done?

USB-VGA and other adapters should work perfectly, the UD-160A was tested on ARM as far back as 2009 so there should still be mainline linux kernel support for all DisplayLink devices (without requiring non-free firmware). However we do not have one of these devices to actually test so if someone can confirm that would be very helpful.

I'd like to use bitcoin, money order or wire transfer please?

Yes this can be arranged, please contact support@crowdsupply.com for details. Obviously these things have additional costs, bitcoin for example is volatile, so you should make sure you are comfortable with these realities, but yes: you can use these methods for pledging.

Wark! The pledge level of the assembled laptop is as much as a Macbook!!

Take a deep breath: we get a lot of people freaking out about this one, so it's worth doing an in-depth update, but here's a summary until then: Macbooks use back-doored processors, are designed by a company that's known for taking libre-licensed and open-licensed software to make profit without giving anything back to the community on which those profits were made (I've been around long enough to know the people who pressurised them to comply with the GPL, over fifteen years ago), and they're at the "efficient mass-volume" manufacturing phase.

By complete contrast this project is run along ethical lines, it's just getting started, is run as a Libre project that respects your privacy even at a hardware level, and the pledge levels were set up to help us get at least to the MOQ level of 250... and it's modular and upgradeable for a fraction of the cost of a macbook (literally a fraction: under 5%). Oh and it's a libre hardware project as well.

In short: there's quite literally not a single level - no common ground whatsoever - between a Macbook and an EOMA68 Libre Laptop Housing, so it's quite hard to understand why people are making the comparison to a company whose CEO has only just woken up to the consequences of what they've done: "We did it to ourselves"

What's web browser performance like?

The Allwinner A20 is a tablet-designed processor: it's low power, it's not going to win any prizes for high performance. You should adjust expectations and configure (or choose) web browsers accordingly. A full update is being written which will go over this in a bit more detail, demonstrating a wide range of both simple and complex web pages in a range of web browsers.

The basics are: use ultra-light-weight web browsers such as netsurf or dillon, or if the latest web browsers have to be used, install u-block to delete adverts, and consider adding "User Agent Switcher" or equivalent so that you can deliberately enable identifying your web browser as a "mobile" browser. Mobile web sites are designed to be less cluttered and to provide much faster response times. The general assumption with the top web sites unfortunately is that if you are running "Desktop" browsers, you must be running a machine with a 10ghz 20-core Intel Processor with 128mb or greater of DDR6 RAM (i.e. a good order of magnitude faster or greater than any machine on the planet that's reasonably affordable, and doesn't actually exist yet).

Our expectations and tolerance of the insane overload of our devices basically means that we need to ratch those expectations down a notch (or five) with the interesting side-effect that by deploying those same tricks on "Desktop" browsers we end up with an extraordinarily quick and productive workflow.

Always remember, though: this is just the first Computer Card. We have to start somewhere. There will be more. By backing us you're helping us to reach better performance in incremental stages... and importantly, do so without requiring VC funding and the associated loss of control of the project that would go with VC funding.

Updates List

This is where we're maintaining a list of updates that are either done or TODO:

  • First update questions and fbturbo g2d driver
  • Standalone Computer Cards OTG Adapter Cable set including HDMI, evaluation
  • Update explaining MOQs
  • OTG Adapter Cable set including HDMI includes WIFI demo
  • T-shirts, mugs etc. ask questions what do you want to see on them
  • 3D-printing demo
  • TODO evaluate USB2-HUB4C from plugable
  • TODO T-shirts, mugs etc. put onto crowd-funding
  • TODO Add an encryption RNGon crowd-funding
  • TODO: add link to phoronix discussions about SoCs. two key links SystemCrasher and robclark
  • http://tllts.org/rsspage.php - keep an eye on podcast DONE, podcast added
  • follow-up with bruce byfield DONE.
  • put break-out board onto crowd-funding DONE
  • write update about 410c with arrow. TOO MUCH WORK, QUALCOMM BLOB AUDIT NEEDED
  • try SUGAR OS. works okay, some bugs, see other TODO.
  • try gcompris suite. works okay, update done.
  • try trinity armhf on devuan - works great.
  • done: keep in touch with albert about french keyboard and stm32f072, update.
  • TODO follow-up with makezine and wired
  • TODO add update for phonebloks and profit stuff, explain about project ara and how it's not really working out
  • TODO special lowrisc update, getting lots of questions about it http://lists.phcomp.co.uk/pipermail/arm-netbook/2016-July/011183.html
  • TODO do FreeBSD
  • TODO look up GuixSD and NixOS (info from leo: runs on x86 only)
  • TODO try slackware / freeslack
  • TODO try gentoo
  • TODO name suggestion for devuan: Humble Purple, Devot moine (devomuan)
  • TODO name suggestion for fedora24: hodenplank
  • TODO try linux-libre kernel https://jxself.org/linux-libre/
  • TODO try freedombox OS.
  • TODO test morph.is
  • TODO try chromeos
  • TODO report sugaros bugs to developers
  • TODO investigate Cortex A7 hardware crypto engine
  • TODO try Gnash http://lists.phcomp.co.uk/pipermail/arm-netbook/2016-July/011275.html

Processor Evaluation Table

Processor Ingenic jz4775 Rockchip RK3188 Allwinner A64 Qualcomm Snapdragon 410c Allwinner A20 Freescale iMX6 Any Intel Atom SoC
Power consumption 2 watts Est 4-5 watts ~danger Est 5 watts (clock-limited) ~danger Est. 4 watts Max of 3.5 watts ~success 6 watts ~danger TOO MUCH ~danger
NDA NO ~success YES ~danger YES ~danger YES ~danger YES (most info is public) No ~success YES ~danger
Datasheets YES Sort-of Yes Mostly YES ~success Yes Yes
PMIC Available YES Yes Yes Yes YES ~success Yes Yes
Reference Schematics YES Yes (thank you Tom Cubie) Yes Yes YES ~success Yes NO
Reference PCB YES Yes (thank you Tom Cubie) Yes NO (PDFs only) YES ~success Yes NO
Tech support Community-based Community-based Community-based NO Community-based ~success Yes NO
GPL-compliant bootloader YES MAYBE (have to check) NO (libdram) ~danger YES YES ~success Yes Under NDA only
GPL-compliant u-boot YES Yes YES (sort-of) YES YES ~success Yes NO (contains spying hardware) ~danger
GPL-compliant kernel YES Yes YES YES YES ~success Yes Yes (but with spying software) ~danger
GPL-compliant 3D GPU YES NO (MALI) ~danger NO (MALI) ~danger YES (Freedreno) NO (MALI) ~danger Yes (Etnaviv) ~success YES
GPL-compliant Video Engine YES NO NO (reverse-engineered partly) NO ~danger NO (part rev-engineered) ~danger NO (proprietary) YES
Example OS (full source) Yes (but none "Libre" (too old) Yes Yes YES YES (huge!) ~success Yes YES
Estimated time 2-3 months 2-3 months 2-3 months 4-6 months 2-3 months 2-3 months 8-10 months ~danger
Estimated cost $10,000+ $10,000+ $10,000+ $25,000+ $10,000 appx $10,000 appx $100,000+ ~danger