So the idea is to use a 3D printer that’s as good as or better than the one used to prototype the laptop’s parts (the Mendel90) but without costing $600 to $700 or above, because it is necessary to buy ten of them in order to have the parts 3D-printed within a reasonable timeframe (i.e., less than three years). Hence I will have to do something like place an order to some parts supplier which literally starts with "Hi, can I have 10,000 M3 hex nuts, please?"
I am however having one of those existential moments, even beyond the "10,000 hex nuts" surreality. To recap, I now have two Chinese 3D printers: $200 + shipping for an Anycubic with a Melzi 2.0 controller, and $150 + shipping for a pile of parts that has me deeply embarrassed that I even thought that it would work. I am so embarrassed that I did not follow my own advice that I have begun writing up a special page on Mechanical Rigidity on the reprap wiki, so that it burns into my brain and I never forget, ever again.
The Anycubic is a combination of good and bad bits: injection-molded X-ends that had warped as they came out of their molds, creating side-loading of the bearings and bending the Z rods. This I can tolerate. Just. The all-metal carriage arrangement is… okay. However it’s one of these "direct-drive" motors with no gearing whatsoever, where the filament gear is mounted directly onto the NEMA17. After seeing how good a Mendel90 is with its 4:1 gear ratio and hobbed bolt I really do not like non-geared direct-drive printhead arrangements, but for $200 + shipping I was prepared to try it out. The NEMA17 itself is low torque, and the controller board can only deliver 1.3 A and is maxed out. Even at 50 mm/sec it skips with this familiar "tkk, tkk" sound that results often in an entire layer being skipped or not printed properly… meaning that at that particular point in the object the entire part could literally snap in two under even the slightest pressure during use. I have to sit here literally for two to three hours, listening out for that "tkk tkk" sound, responding immediately and carefully pushing the filament by hand - as the printhead is moving rapidly back and forth - into the extruder.
The Anycubic is basically what I’m using - because it’s here now - to try to rescue the "pile of parts" bought on Taobao. I’ve taken the opportunity to turn the X-carriage rods from a vertical to a horizontal arrangement, placing the nozzle as close to the centre between the two rods as i dare. What that means is that gravity takes care of keeping the carriage down (yes there is travel-play in the bearings: about 0.25 mm) and also any rotation of the carriage (caused by the travel play in the bearings) is reduced, not amplified. With a vertical rods arrangement, that travel play of "only" 0.25 mm is amplified by the fact that the nozzle is often a good 50 to 80 mm away from the centre line of the two rods. Thus that play of "only" 0.25 mm can easily often be amplified to a whopping 2 mm! I simply do not understand why people do not understand basic mechanics. Vertical carriage arrangements are fine if you use high quality rails, not cheap-ass linear bearings that rattle around with up to 0.5 mm of travel under rotation!
Why am I even spending my time on this? Well, it’s so that I can give this pile of parts to my associate here in Taiwan, but also because I want to get back into the flow of doing 3D design again, after taking a break for over a year. I have to begin adjusting the libre laptop casework again to fit the new PCBs at some point, and I would like to be "in the zone" when it comes to it, as they like to say.
And that brings me to Existential Moment Number One. I returned recently to the reprap forum after a break for a year, and it feels like it’s a sleep-walking ghost town. Sure, there are good people there, and it’s really nice to see some of the familiar experienced faces helping out the newcomers, not least because I can learn things from those people as well. But I didn’t really know what I was feeling was wrong until I learned yesterday that reprappro - the company that was started by one of the founders of the entire open 3D reprap community - shut down only last year whilst they still were in the black.
This is extremely relevant to the EOMA68 Project and to Open Hardware in general: it’s a warning shot across the bows. Basically, reprappro shut down because the market is so saturated with crowdfunded efforts and Chinese clones that their expertise, on which those crowdfunded efforts and Chinese clones critically depend, is completely drowned out. You can see evidence of this in the number of threads along the lines of "I Bought A China 3D Printer Clone Has Anyone Else Also Got One I’m Designing Totally New Parts To Replace The Rubbish I Got" and "I Bought A China 3D Printer Controller Board Clone, Its Firmware Is Illegal Has Anyone Reverse-Engineered It?"
You bought a $20 PCB with illegal, copyright-violating software on it and you’re asking people who normally charge $200/hour for reverse-engineering expertise to spend WEEKS of their time… for free?
The lesson here for all of us is a repetition and thus a reminder of my sponsor’s business model (ThinkPenguin). ThinkPenguin funded the libre laptop because they want their customers to have the option to be able to buy a laptop that doesn’t spy on people. That easily cost them around $50,000 USD in my time (actually, the rent on the apartment in Den Haag plus living costs). The latest illustration that this is not nonsense, that it’s a genuine threat to your privacy in a completely undetectable manner, was reported on two days ago.
ThinkPenguin’s business model is to do all the research into hardware that "just works" so you don’t have to. Sure, you can go buy cheap clones and knock-offs, or you can buy the exact same part (the one that ThinkPenguin funded from their profits to make sure its firmware is entirely libre) from someone else at a cheaper price, but in doing so you undermine your own chances of being able to buy libre hassle-free computers and peripherals in the future. The classic example is the AR9271 USB-WIFI dongle. This is the one for which ThinkPenguin spent two years walking Atheros through the process of releasing the firmware. Buying from their website supports their long-term efforts to free up 802.11ac. Buying from their sponging competitors does not.
A high-quality 3D printer I would love to be able to use for the job of doing all 5,000+ parts for the laptop, is the Ultimaker-2. The build quality is so superb on this printer that it is easily capable of handling 150 mm/sec speeds and still produce good enough quality parts, thus ensuring that the time taken to print all the parts (if a single printer is used) is only three years if we assume an eight-hour day.
You may need to blink a couple of times on that one, and re-read that sentence. Yes, really, I did say three years for a single printer. So that means it would be necessary to get 10 printers in order to print all the parts in a reasonable timeframe. If you’ve not seen how much an Ultimaker-2 costs, it’s around $1000. So that would be $10,000 USD in 3D printers. I cannot possibly justify such expenditure. The actual filament itself needeed is only somewhere around $1500 to $2000 USD.
Now the experiment in buying a China Taobao wannabe knock-off for $150 plus $40 shipping to Taiwan starts to make sense, doesn’t it? Except precisely because it’s a clone made by people who "copy copy copy" with absolutely zero engineering design expertise whatsoever, and then unfortunately tinker, or worse, ship stuff that you didn’t actually ask or pay for (but you’re not a Chinese national so have no legal recourse to recover your money), what I’ve received is basically… 80% junk.
So, I am at this weird cross-point where in order to fulfill the pledges that people have entrusted me to do, I am - somewhat unbelievably - going to have to return to designing a 3D printer - one that has the advantage of the lower-cost of the parts that can be sourced from China, without the huge turn-around times and costs of shipping to the USA or to Europe, but with the engineering expertise from a Western mindset and with the approval and input from the reprap community and other sources, all of which is completely lacking from the average Chinese "clone-shop." Times ten. Hence the surreality.
Now, if anyone has any better ideas, now is a really, really good time to raise them (on the mailing list, so that they can be discussed). Just the fact that ten printers are needed is enough, for example, to justify starting an entirely new crowdfunding campaign just for the 3D printer. Ideas here include providing it to people across the world, so that they can help out with the actual printing of the laptop’s parts, but more than that they will become established and reputable suppliers for replacement parts, and could register on, for example, 3DHubs to make some money from selling 3D printing services. Some of those people could be subsidised from the available funds, but it would be nice if people could actually buy the printer themselves.
The key here is that both the quality of the prints and the speed have to be really above-average, and using high-quality PLA is essential. The parts are really quite complex (some of them are 260 mm long and have to be printed across the diagonal of a 200 x 200 mm bed) and there are lot of them to do (5,000+ in total). Even if ten printers are used that’s still almost FOUR MONTHS of eight hours per day 3D printing per printer.
And no, using a network-3D-printing house is not okay, because the quality of the PLA from such places simply cannot be trusted. It’s Faberdashery’s PLA or nothing. I’ve shared some of the nightmare horror stories of low-quality PLA with people on the list already.
A huge surprise to me is to learn that these updates I’ve been writing do actually seem to be inspiring people, both to contact Crowd Supply and set up their own open hardware project and also to learn about some of the things to avoid, that I in turn have learned from various other open hardware projects over the past decade and more. For example, the ZeroPhone developer (hello!) has been reading these updates and so has deliberately sourced parts that are easy to obtain and commonly available in China. Yay!
Also, amazingly, we have not one new EOMA68 card that people are planning to develop, but two! One using the Nextthing.co GR8 (which is actually an Allwinner A13 die combined with a 512 MB DDR3 in what’s known as a "Multi-Chip Module"), and the other using the Freescale/NXP iMX7, which has tamper-resistance detection built-in.
Then also I have been asked by my sponsor to work with a Chinese router company to investigate creating a QCA9531-based libre router (the QCA9531’s firmware is entirely libre). This has its own page and since that page was last updated I have made the decision to split this PCB down into two parts. The first will be pretty much the QCA9531 reference design, as-is, except with a feedback loop on a header to get the USB, UART, and some GPIO off (and then back again, for a stand-alone router). The second will actually be the planned "Mini Desktop," this time with a four-port USB hub and an HDMI converter IC instead of VGA. The interesting bit will be when both PCBs are put (stacked) into a single box, with a 14-pin header connecting the two. That then becomes a really interesting and powerful product. Actually, three products, all of which could go onto the next crowdfunding page.
Lastly, there is a status update on the mailing list: basically, I am "busy waiting" for the factory to deliver new versions of the PCBs that need pre-production testing. But, finally, please, I do have to emphasise: I really do need help with the 3D printing. It’s a big committment: it’s somewhat impractical to set up ten 3D printers in an apartment here in Taiwan. Do join the mailing list and make yourselves known.