New Crowdfunding Campaign Offers Modular EOMA68 Computing Devices (crowdsupply.com)

A new crowdfunding campaign by Rhombus Tech "introduces the world's first devices built around the EOMA68 standard," which separates a "modular" CPU board from the rest of the system so that it can be easily used in multiple devices and upgraded more simply. Rhombus Tech is now offering a 15.6-inch laptop, a laser-cut wooden Micro-Desktop housing, and two types of computer cards, both using A20 dual-core ARM Cortex A7 processors. The cards are available with four flavors of the GNU/Linux operating system, and they're hoping to receive RYF certification from the Free Software Foundation.

"No proprietary software," explains their campaign's video. "No backdoors. No spyware. No NDAs." They envision a world where users upgrade their computers by simply popping in a new card -- reducing electronic waste -- or print new laptop casings to repair defects or swap in different colors. (And they also hope to eventually see the cards also working with cameras, phones, tablets, and gaming consoles.) Rhombus Tech CTO Luke Leighton did a Slashdot interview in 2012, and contacted Slashdot this weekend to announce: A live-streamed video from Hope2016 explains what it's about, and there is a huge range of discussions and articles online. The real burning question is: if a single Software Libre Engineer can teach themselves PCB design and bring modular computing to people on the budget available from a single company, why are there not already a huge number of companies doing modular upgradeable hardware?

Comments

Re:Won't happen

[Aighearach]

What you don't realize is that none of that means it won't happen, it just means Joe Average User won't end up buying it.

There is lots of open, modular electronics already. Your boogeyman didn't pop out.

Because.se one size does not fit all

[zemned3000]

The reason that this isn't already a common approach in the industry is that forcing constraints on form factors for SoC devices has some intractable issues. If you have a powerful SoC it demands high power and needs to dissipate heat; so the upper bound of what you can achieve in the packaging and with the connector will be rapidly met u.nless it is massively over specified, and then it will be large and expensive. Also, display technology is not fixed in time, parallel interface signals are already quite out of date as an interface specification , although the actual limit here will probably be down to the PCMCIA connectors impedance discontinuity and consistency after numerous insertions when more modern differential display protocols are adopted. It is a laudable aim, but I doubt this will save the planet from computer waste.

Re:Because.se one size does not fit all

[lkcl]

The reason that this isn't already a common approach in the industry is that forcing constraints on form factors for SoC devices has some intractable issues. If you have a powerful SoC it demands high power and needs to dissipate heat; so the upper bound of what you can achieve in the packaging and with the connector will be rapidly met u.nless it is massively over specified, and then it will be large and expensive.

we're dealing with that by setting a hard limit for the [re-used, pin-incompatible] PCMCIA "Type I" and "Type II" sockets which are 3.3mm and 5.0mm respectively: the hard limit for these two thinner card types is 5.0 watts. so at around 3.0 to 3.5 watts there's still absolutely no need for fans or any kind of special thermal considerations: passive cooling is all that's needed, and the SoC happens to be in contact with the stainless steel case, which happens to be in contact with the aluminium of the keyboard (in the case of the laptop).

just over that (up to 4.0 watts) and it is possible to use exactly the same graphite paper that's been developed for mobile phones. cheap, readily-available.

at around 4.5 watts it would be necessary to seal the package and flood it with thermal gel.

Also, display technology is not fixed in time, parallel interface signals are already quite out of date as an interface specification , although the actual limit here will probably be down to the PCMCIA connectors impedance discontinuity and consistency after numerous insertions when more modern differential display protocols are adopted. .

right. i spent five years analysing this and the impedance of PCMCIA (which, again, just to emphasise, we are *NOT* electrically or electronically compatible with: EOMA68 merely REUSES the PCMCIA connectors, housings, sockets and assemblies) is 100 ohms.

the EOMA68 standard uses RGB/TTL because that allows you to go all the way from 320x240 up to 1366x768 which works out to be around 80mhz. 80mhz over 100 ohm impedance is just about acceptable: you remember those "gold shields" on PCMCIA? those were designed to reduce EMI. the cards we're using for the initial prototypes have the metal case covering the entire connector, both sides.

why use RGB/TTL? this is covered in the eco-computing white paper in detail, section on "interface selection" http://rhombus-tech.net/whitep... basically if you consider the cost of 320x240 LCDs and take a look on http://panelook.com/ they're peanuts cost and they're *all* RGB/TTL. if you added a converter IC it would be a massively-disproportionate percentage addition to the BOM. however if you go up to a 1024x600 which costs $18 approx and you add a $1 SN75LVDS83b LVDS converter IC.... that's not so bad, is it?

Re:Because.se one size does not fit all

[chris2net23]

Your missing the point and value of this project entirely. We're giving the community and users control over the design of there computing devices. Right now we're stuck behind crappy poorly supported proprietary systems full of backdoors and malicious code. Modern Intel and AMD systems, televisions, every Android device, most Raspberry Pi-like boards, tablets, and even routers have backdoors and other malicious features.

You can't replace wifi cards in most devices these days because manufactures like Intel are integrating the chips into the boards. All modern NUCs for instance ship with proprietary intel wifi chips. Companies like Linksys are undermining our freedom and calling it "open source" trapping us into a system that ensure we won't retain control over critical aspects of the firmwares. Without access to these pieces important user contributed features including good support for mesh networking will no longer be possible. A great example of this is Atheros moving all the important bits in its 802.11ac chips into the firmware and is now refusing to release the code. Atheros *was* one of the better companies, but not because Atheros is good. But because we had multiple free software activists inside the company and because of work my company did to push things forward.

We don't need more than 2GB. We want more than 2GB. And that's fine. We can do that. But the first priority has to be to get a basic system out the door and funded. There are already efforts to get code released for more powerful SOCs. In six months you'll be able to replace that 2GB dual-core 32BIT card with a quad core card. And clearly the issue is not with the 2GB of ram. The issue is with bloated desktop environments. That's a solvable problem by simply not utilizing said desktop environments. You simply can't buy a new computer for $65- but that's exactly what you will be able to do. These aren't intended to replace your high end Intel/AMD system just yet. That's farther out.

Re:Because.se one size does not fit all

[lkcl]

Not trying to be rude.

no offense taken: this is the internet... it's slashdot.... what you wrote is actually really helpful as a counterpoint: chris answered i think really well https://news.slashdot.org/comm...

, bill yourself as a fully open and environmentally friendly alt to the raspberry pi and similar.

*deep, deep breath*..... AAAAAAAAAAAgh no :) *shudder* no, no, nOooo, and no.

ok to explain my reaction, there: those are all SBCs (single-board computers). after six months of supernova-style popularity, they're dead. each manufacturer of each SBC has to scramble like mad to bring out *the next* SBC using whatever processor they can get their hands on, and the next, and the next, in a desperate cat-and-dog bitch-fight of popularity and unethical abuse of the word "open". ... am i painting a broad picture here of the *really* stark contrast between the approach taken by the embedded "educational" SBC clone market and what we're doing with EOMA68?

by total contrast we're creating the beginning of a comprehensive eco-system of hardware re-use which *happens* (through direct correlation) to both save money for end-users and also reduce e-waste.

but more than that: if we took the EOMA68-A20 board and turned it into an SBC, it would *INSTANTLY* be perceived as being a tired total waste-of-time banana pi or cubieboard clone... when in fact the irony is that those products exist *because* of the reverse-engineering and persistence that i applied to Allwinner to obtain GPL compliance. the sunxi community then helped take that initiative over, they've been working non-stop now for years to pressurise allwinner, and i've been helping quietly in the background ever since.

this project has a completely different focus in other words, where it succeeds if there is a *huge* compatible eco-system (tablet, laptop, router, camera, gps, media centre, lcd tv, games console - everything you can see on here http://rhombus-tech.net/commun... and many more) and a huge compatible range of EOMA68 Computer Cards (and an FPGA card and a Pass-through Card and a DisplayLink Card) with a wide price-range and crucially a decades-long-term "just plug it in, it will work" *stable* standard. ... massive, massive difference.

Future CPU cards (different CPU architectures)

[mafm]

Currently the CPU in the CPU-cards available in the campaign is an ARM 32 bits ("armhf" for Debian systems).

In the future, if things go well, there are plans to launch other CPU-cards that meet requirements of low power, hw and sw freedom (not requiring proprietary firmware blobs to run), etc. Other CPUs have been already considered, including different architectures, like MIPS. The housing (laptop, micro desktop, etc.) can be reused, it's just a matter of swapping the CPU-card -- that's one of the main points of this project.

I'm hoping that there's enough interest in the project and goes ahead, that the ecosystem thrives and other CPU-cards based on free designs like OpenRISC or RISC-V will be produced in the future.

Re:Future CPU cards (different CPU architectures)

[lkcl]

Currently the CPU in the CPU-cards available in the campaign is an ARM 32 bits ("armhf" for Debian systems).

and a Pass-Through Card just to make sure that people don't get the impression that EOMA68 is restricted to Software Libre, ARM processors *or* processors *at all*... https://www.crowdsupply.com/eo...

In the future, if things go well, there are plans to launch other CPU-cards that meet requirements of low power, hw and sw freedom (not requiring proprietary firmware blobs to run), etc. Other CPUs have been already considered, including different architectures, like MIPS. The housing (laptop, micro desktop, etc.) can be reused, it's just a matter of swapping the CPU-card -- that's one of the main points of this project.

i did a big evaluation here - bear in mind that this evaluation process has been continuous and ongoing for *five years*: https://www.crowdsupply.com/eo...

I'm hoping that there's enough interest in the project and goes ahead, that the ecosystem thrives and other CPU-cards based on free designs like OpenRISC or RISC-V will be produced in the future.

OpenRISC was not designed around a harvard architecture so is extremely unlikely to go beyond around... 500mhz, even if it was in 10nm, due to insufficient pipeline lengths. it would be ultra-ultra-ultra-ultra low-power in those geometries but would never be capable of going beyond those speeds without a total redesign.

RISC-V on the other hand has been designed from the ground up around the lessons learned from generations of RISC development.... it's just that it's going to be about 3-6 *years* before a decent SoC is made based around it... there was one that *almost* had what was needed... i think this was discussed on the mailing list... basically if it has PCIe and the expectation that the CPU can be married with a PCIe-based Graphics Card, it's already too late: that's a Desktop system, minimum power consumption 150 watts.

my preferred approach is to work with e.g. Loongson Leemote MIPS64 -that's a tried-and-tested design that has hardware-accelerated support for 200+ x86 instructions and i hear some ARM ones as well, that allows foreign instruction sets to be executed at SEVENTY PERCENT of the actual MIPS64's own clockrate. which is pretty amazing.

FSF's Respects Your Freedom certification

[mafm]

Probably interesting for many folks around here... there are plans to submit these projects for the Free Software Foundation's Respects Your Freedom program (contacts already started).

Why? Its the economy, Stupid!

[Anne Thwacks]

if a single Software Libre Engineer can teach themselves PCB design and bring modular computing to people on the budget available from a single company, why are there not already a huge number of companies doing modular upgradeable hardware?

Volume is king in electronics. Surely everyone knows that here! In case you had not noticed, a computer is made from -

• Less than a pint of oil (two pints if the case is mostly plastic)
• Less than two pints of sand (Much less if no glass in the screen)
• Not enough steel to make a wing panel for a Fiat Punto
• Not enough copper to make an ashtray
• Just about enough aluminium to make a saucepan
• A couple of kilowatts of energy
• VERY EXPENSIVE machine tools
• HUNDREDS of THOUSANDS of HOURS of VERY EXPENSIVE engineering time
• MASSIVE AMOUNTS of special purpose tooling

The last three items are one-off costs, spread over the entire production volume. If your volume is high, they are negligible, if your volume is low, you are stuffed.

PCB design is a non-issue - if you don't pay the going rate. PCB test, debugging and verification, not so much. Hint: you cannot do your own quality control - no one spots their own errors.

Re:What are the success criteria?

[lkcl]

Sounds interesting, but I'd have to see a complete proposal before I'd chip in. I'd want to see the schedule, the budget, the resources, and the success criteria to know if the project succeeded.

most of the information you've asked for, because this is a *genuinely* open and transparent project, is on http://rhombus-tech.net/crowds... - including the BOM, a full risk analysis, and much more. having been around for a long time, long enough to have seen the openmoko fail, and the pi-top team break their promises, and the shit-storm that resulted from the purism team's deceptive marketing, and the difficulties that the openpandora team had with R.F. and firmware: if you have any specific advice, TELL ME. i WANT TO KNOW. best place to do that is the mailing list because then other people can help evaluate your proposals and questions - http://lists.phcomp.co.uk/pipe...

The summary sounds way too grand, so I think I'd want to see it broken down into pieces that are small enough to understand, too.

it's been five years: that's a lot of time to think, plan and get everything lined up. if you're interested in the background as to *why* i am tackling this, it may help to read the background section (first question) http://lkcl.net/articles/eoma6...

"breaking it down into small pieces" it turns out is extraordinarily difficult. the simplest i've found is, "you know like a pause memory pause card? yeah? well this is a pause computer pause card. same benefits as memory cards except now you move the *whole computer*". but even that often is not conceptually enough. after repeating things about 200 to 250 times at hope2016 (and losing my voice on the first day) there's a live video which you can find at https://www.crowdsupply.com/eo... - i managed to get it down to *only* 3 minutes, to cover *a few* of the benefits. the rest (that i have been able to think of over the past five years) are covered here in the "scenarios" section: http://rhombus-tech.net/whitep...

Also important to make sure nothing is overlooked, such as sufficient testing. Be fine if the same organization that helped check the proposal evaluated and reported on the results (perhaps holding the money, too).

well - it's just me, self-auditing with an "always transparent GENUINELY open approach learned from software libre project management of 20 years" unless other people pop up to help. so, you and everyone else on the mailing list will just have to keep an eye on me. and help out with the testing... because it's a software libre project, and i can't do everything, so *need help*. funny and really cool story: a guy called albert contacted me last month, asked if there was any plans to do a french keyboard. i went (internally), "argh, haven't got time, let's point him at the git repo and tell him about the STM32F072 nucleo boards, see what happens" and surpriiise! turns out he's an embedded hardware engineer... so guess what? he's now joined the mailing list and is helping to do french keyboard firmware and much more! https://www.crowdsupply.com/eo... and you can check the mailing list archives as well.

P.S. I think this is a solution to the general problems with all of the crowdfunding systems that I have examined. No accountability or adequate planning.

you're telling me. i spoke to a battery manufacturer last year: we had a bit of a laugh as he explained that a *FUNDED* project for a head-wearable device contacted them and asked him to violate the laws of physics. they'd used a high

Re:Won't happen

[lkcl]

What you don't realize is that none of that means it won't happen, it just means Joe Average User won't end up buying it.

There is lots of open, modular electronics already. Your boogeyman didn't pop out.

sorry, i don't understand. could you possibly expand on this, perhaps help review the logic analysis behind the modular standards that i've reviewed over the past five years, if you feel that i've missed any or missed anything, please do let me know. the list that i maintain, including comprehensive analysis, is here: http://elinux.org/Embedded_Ope...

the issues to take into account are: it must be absolutely simple, it must absolutely work, and it must not break (due to mechanical or EMI issues). we just saw a report only last month a journalist (who was happy to say that he has "big fat fingers") trying to assemble and upgrade a Gaming PC - he just couldn't cope. and that's a "modular" design, isn't it? exposed electronics, fiddly parts - all modular mind you! - he was absolutely shit-scared of dropping screws and shorting out $1000+ worth of parts.

by total contrast, robust memory-card-form-factor casework is simple, works, and is protected mechanically and is EMI (static) shielded. if however you know of any other modular industry standards that fit these criteria, please do tell me because i will need to evaluate them and adapt accordingly.

Re:But my PC is already modular.

[lkcl]

I have an ANCIENT (>10 years old) Dell XPS desktop machine - and last week, the motherboard failed. Went to Fry's paid $65 for a new motherboard and $120 for a new CPU (which included a new cooling fan). My RAM modules were too ancient to run in the new motherboard - so I spent another $60 for a couple of RAM modules. To my surprise, the original power supply, graphics card, hard drive, DVD drive and case all fitted perfectly - and a simple reboot got me back into Ubuntu as if nothing had happened - I was back up and running in an hour.

Sure, the CPU socket had changed - and my decade-old DDR-2 memory wouldn't work in the DDR-3/4 motherboard - but aside from that, modularity worked 100% perfectly. I could have chosen from a dozen different CPU's and a similar number of RAM suppliers and any one of a dozen motherboards - and the outcome would have been the same.

... you're aware that intel has moved *away* from socketed CPUs and is forcing BGA onto manufacturers, now? you're really lucky to have been able to find a motherboard that suited you which didn't have the BGA-soldered processor on it.

So the desktop PC "standard" is already an incredibly modular system. The problem is that (by modern standards) it's physically huge.

For small systems like IOT devices, the cost of "the computer" including graphics, networking, RAM, long-term-storage is down to $10 or less...so modularity at that scale is just pointless - increasing the cost by adding connectors between the parts is just silly.

... and actually causes huge reliability and manufacturing issues. yeah. you get it. which is great to see.

For systems at the scale of a cellphone, modularity is a tough sell because the physical form-factor has to fit perfectly with the shape of the battery and screen and heat management is a big issue - so making a *usefully* modular phone is challenging.

dave hakkens is *PISSED*, man. like, really *really* disappointed and betrayed by google. all that money and they *claim* open-ness but actually instead they're just strengthening the positions of the existing cartels. what a fucking waste. the sad thing is, i spotted all this years ago, but it's taken everyone else (and dave) quite a while to catch up, because most people are non-technical and do not have a reverse-engineering background https://davehakkens.nl/news/re...

The real issue is modular laptops.

that's why i'm tackling it: to make sure it's done right, and in a transparent way, with a standard that's *genuinely* open.

It's a real pain if you screen gets cracked or your motherboard or power supply fails. But you don't need modularity at the electronics level - it's all about modular cases and connectors.

and having the right *to* repair or 3d-print those modular case parts.

You can buy software, download free software or write your own - and it's pretty simple to make it work on the trifecta of OSX, Windows and Linux - and trivially easy if you can make it web-based. But it's very evident that the business model of most companies these days is to lock you in to buying music/video/apps from their "app store"...that's where the $$$'s are...so expect to see more moves like MS's efforts to lock down Win-10 so you have to buy apps through their store.

one thing that hadn't occurred to me (but i must add to the standard, to protect against), is that DRM is going to be totally ineffective in modular computing. you can't *possibly* agree a DRM standard across a truly open architecture: it just doesn't work when access to all the hardware relies critically on drivers that are released under GPLv2+ licenses.

i mean... they can _try_... but it will be a hell of a job to maintain.

so i think i'm safe in adding "hardware-level DRM simply not

Re:Won't happen

[Aighearach]

And by "move to block them" what do you mean, they'll dress up in a chicken suit and dance in the street? It won't work.

It is a pretty stupid conspiracy theory when all it consists of is an underpants gnome.

Maybe in your country there is some sort of process to "block" companies that anger competitors, but in most of the world it simply can't be done.

Re:But my PC is already modular.

[lkcl]

The problem is that (by modern standards) it's physically huge

Any smaller and it would not hold a DVD drive, an LTO drive and a DAT drive, and still have somewhere to put USB sticks and SD cards, not to mention the place required for SCSI cards.

yeahhh i had to make a decision whether to make the first EOMA standard for mass-volume clients or for mass-volume servers. i figured that with facebook, google and hp and others having the data centre market pretty much sewn up, and them trying to convince everyone that "cloud is good", and having poisoned the word "open" in that area with their "open compute" standard, the chances would be much better if i focussed on "the little guy"... ... that meant using hardware that was simple enough for someone like me to learn, and with a persistent and bloody-minded attitude actually gain access to Reference Designs and so on, and that in turn meant SoCs around the $2 to $8 mark that are designed for "tablets" and "smartphones", not "intel-style PCs", and _that_ in turn meant "SATA and GbE and PCIe are off the interface set". you can't get a QFP-176 SoC at $2.50 that has PCIe or SATA, basically, but you *can* get one that has HDMI, 3x SD/MMCs, 2x USB2 and several other really good interfaces that you can make a fully-functioning low-cost computer from (look up the Allwinner R8 - it's QFP176 and $2.50. absolutely amazing).

As someone who actually saw http://www.youtube.com/watch?v=6v4Juzn10gM EDSAC working, I think the standard tower PC case is about right, and I KNOW that tape will keep my data for 30 years (I have read my own backups 30 years later),

toootally cooool! ahh... don't read them too often, the heads are abrasively-worn by the read/write cycle...

and I know DVDs won't keep my data for even six years.

*sigh* that oxygenation of the metallic compound through the edges of the disc when they thinned down the amount of varnish to save money, it's a bitch, huh? did you happen to mark the DVD with a permanent marker then close it up inside a plastic case? i know of someone who learned the hard way that the fumes from those pens migrate *into* the discs and oxidise the metal.... whoops... ... you can't use HDDs because after 5 years there's a 25% chance that just *powering them up* will cause a head-crash: thermal warping (cool-down) of the HDD when it was last shut down, combined with the bearings seizing as the oil/grease isn't being cycled round during storage.... yyyeah.

you can't use SSDs because the geometries are too small now, you get data loss over time due to quantum effects and radiation... what are we down to now... 20 atoms to store one bit of data??

we're screwed basically :)

I certainly don't want MY data in the cloud.

cloud... that's where you store your data on other people's computers, right? :)