A Tale of Two Tools; In which the Saga of Shenzhen Maker Faire unfolds; Drastic capacitor price hikes; HDMI, FOSDEM. and a Libre RISC-V 64-bit SOC design.
With apologies for merging what should have been separate updates into one, this update is slightly longer than usual. Main things: FOSDEM is a go, anyone attending please do get in touch, I’ve got a stand to run and three talks to give: I will need help and also sponsorship for the plane flight cost. HDMI layout: Richard has provided an extremely comprehensive and detailed guide to how to make the critical impedance taper. Clarification is ongoing.
So the connections continue, from Crowd Supply. I love that Joshua is actively interested in what he’s doing. It’s not like the other crowdfunding sites, where there’s virtually no cross-pollination or encouragement to connect and help each other out. In this case, Joshua has been following along and understands the EOMA68 Libre Laptop Project’s need to increase the "units per hour per dollar" metric so has put me in touch with Brent, who has run a successful campaign for a dual-head extruder. Basically, what this does is double the output capacity of any 3D printer design, by sharing the print bed of a Mendel-style 3D printer to do simultaneous identical prints: one on the left, one on the right.
In terms of the "bang per buck per hour" metric this innovation ticks all the boxes. Let’s say it costs $400 USD for one 3D printer, and $100 for the Twoolhead. You could spend $800 on two 3D printers… or you could spend $500 on one 3D printer plus a Twoolhead, and now you have a whopping 60% increase in batch productivity. Of course, it’s not quite that straightforward: you need to take into account that the Twoolhead is an increase in weight, however most people never run Mendel style 3D printers at the kinds of speeds I do, because you get far better quality at the lower speeds (which would be important for any professional considering the Twoolhead as they would be 3D-printing batch jobs for customers), so it is extremely unlikely that they would run into the kinds of problems that I do at 150 to 200 mm/sec travel speeds.
Although the Twoolhead is specifically designed for the Lulzbot, and is the traditional "Wades" geared design on horizontal rods, it will almost certainly fit any similar Mendel 3D printer with the same separation between X rods (70 mm). If you are not familiar with why you should never buy a 3D printer with a vertical X-Rods setup, please review http://reprap.org/wiki/MechanicalRigidity: all you really need to do is grab the end of the extruder on a vertical X-rods printer and see how much it waggles about. Try the same thing on a horizontal X-rods setup and that tells you everything you need to know.
Brent has picked an extremely sound foundation - a Lulzbot Taz 6 - on which to base his invention, and anyone interested in increasing batch productivity would do well to investigate the Twoolhead.
I seem to have two extraordinarily patient friends: one a mechanical engineer based in Hong Kong for nearly three decades, and Mike from <Vastelectronics.com> in the west of Shenzhen. Their patience never ceases to amaze me. My mechanical engineer friend is the one who has followed my advice on the two 3D printers: we spent two days (and nights) trying to rush-assemble a second Riki200. It was good that we didn’t succeed. So, after 10 hours of travelling already, starting at 5am that morning, my friend returned from Shenzhen to pack up, and we embarked on another four hour journey of taxis, trains, and border crossings at 10:30 at night. Surprisingly, or perhaps not, crossing at Luohu is quite straightforward at that time of night, which is a good thing given that we were each carrying 25 kg trolley-bags plus a 15 kg 3D printer each in those dreadfully cheap and colourfully-chequered shopping bags (if you have one you will know it as its zipper will have ripped out of the seam on your very first shopping trip), so as to disguise our wares to not look like "cargo". We arrived at "My Pleasant Hotel" somewhere around 11:30pm at night and zonked out, totally exhausted.
Normally I like to do a heck of a lot more planning: actually go somewhere and scout it out in advance, but this time, because my friend had to be in Shenzhen for work that afternoon, it wasn’t possible to be there and set up: instead we had to wing it. This caused some… stress. We arrived at the Polytechnic in the Nanshang District and had to lug our 80 kg "cargo" half a kilometre through the campus to the booth. Assembly took another hour and a half, because, actually, people were fascinated by the "unfinished" state of the Riki200. In particular, people loved that I raided the cupboard for coathangers, applying the principle of "Permaculture", and the masking tape pipes I have improved by winding some 0.5 mm brass wire in between the two layers so as to stop it from collapsing. More on this in another update (when I am not cut off from YouTube by the Great Firewall of China…) as I found something particularly fascinating about the relationship between the X and Y movement and the rotation of the pipe’s ends, which is particularly important to observe in the layout.
My friend took much less time with the Taobao clone, and I could audibly hear the exhaustion dropping away from him as he began explaining, animatedly, what it could do. Later in the day I was so delighted to hear that he had been approached by someone from Canada who has been selling RAMPS 1.4 and getting absolutely fed up of customers complaining of burn-out. If you are not familiar with why, see http://reprap.org/wiki/RAMPS_1.4 and check the "Thermal Design Flaw" section that I added. This guy from Canada is really interested in RD3D and will drop by again later today. In particular he is interested because all of his dissatisfied customers will be able to recover their failed purchase by transplanting the Polulu steppers and the Arduino, saving them money but also having an upgrade path to TMC2130 steppers, and a Due (32-bit), and later the dual TMC2660 expansion board I am developing.
Also, my friend has had someone approach him for a design commission, because he has an obsolete extremely high quality camera where the lens adapter has broken, and he cannot find a replacement. My friend will design and 3D print one for him. I was extremely relieved to hear about this as I was hoping that my friend would financially benefit from this trip in some way, after all the help he’s provided me.
The Workshop on Frame Mechanical Rigidity went extremely well. It was total chaos getting organised. Mike bought 1,000 colourful drinking straws for me, but on the morning he was to drop them off he had a production issue at the factory. Despite this, he still drove two and a half hours round-trip to spend one minute at the roadside dropping off a cardboard box with… straws in it. I still cannot believe he is so incredibly patient and helpful.
The workshop was really well attended: 18 people, two of them children, more than 50% of them women. What totally surprised me was that none of them actually owned a 3D printer. So I adjusted my expectations accordingly very quickly, and was able to use the principles I was teaching them to make sure that if they ever bought a 3D printer they would know what to look out for, so as not to be bitterly disappointed and shocked by how rubbish whatever it was that they bought really was. Or, at least if they did buy something cheap, how to go about improving it. Three of the attendees - all of them young women - were actually from the Polytechnic, on a 3D Printer Design Course.
One thing that absolutely amazed both myself and my friend about this Maker Faire is quite how many young women attended and were really, really interested. We say in the West that there is a bit of a problem with getting women interested in technology and science in general: I can categorically say that in Shenzhen this is not the case. Not only that but the whole makerspace arena in general is extremely well-funded: http://x.factory has some serious, serious kit there, and there are parties every Friday night to attract and encourage young people to attend. Both myself and my friend agreed that whilst the USA and Europe are going into headless chicken melt-down mode, China is set to wipe the floor with the rest of the world when it comes to advances in engineering and design. It would be nice if they also respected Open Hardware principles, and amazingly I have met someone there who works for a 3D printing company and yes, he actually does want to release designs as Open Hardware. If nothing else this chaotic trip is worth it, just to meet that one person.
So, anyway, after the party we flowed into a taxi, got back to My Lovely Pleasant Hotel and again zonked out, exhausted. Surprisingly, 90% of the aches of walking around 10 km in one day are gone this morning, where I have the PCB Component Sourcing workshop to do later this afternoon.
This was slightly less awkward as all the equipment was there already. We left it on the table overnight on the basketball courts! It was just too much even to move, despite they offered some inside space to store everything. The stand, again, was well attended on both days, and both my friend and I had lots of people coming up and actively being interested in why you can consider making your own light-weight air duct pipes out of masking tape and 1 mm brass wire: the application of the principles of Permaculture where it doesn’t matter, and properly sourcing high grade engineering components where it does. One of the organisers apparently tried to come and say hello, but there were so many people around the booth as we were both doing separate explanations that they simply couldn’t get near. Apparently this contributed to us winning not one but two SZMF Awards.
The last day, we packed up, as best we could in a frazzled state, and got everything back by taxi to The Lovely Pleasant Hotel in Huaqiang North. Travel back to Hong Kong and then finally to Taiwan brought the total number of transfers of the 40 kg worth of luggage to a whopping forty decantations to and from different methods of transport. This is not something that I wish to repeat for FOSDEM without assistance!
One thing I managed to get to do was to speak briefly to Mike. He explained that Apple’s insatiable demand for capacitors has pushed pricing up by a staggering five fold increase. Shenzhen suppliers are hoarding 0.1uF, 1uF, 10uF and 100uF capacitors - the most common types basically. I have already begun the process of replacing some of the 10 µF capacitors with twin 4.7 µF ones: this is generally good practice anyway as the "load" is shared between the two, and it’s well known that "two capacitors are better than one."
Also, the DDR3 x8 RAM ICs - we selected 1066 MHz parts because the 800 MHz ones were becoming hard to get hold of - are now also getting hard to get hold of. So the next revision (2.7.5) will have to have some experimentation and confirmation of different types of RAM ICs: all the way up to 1800 MHz DDR3 which will then be run at the current clock rate of 350 MHz to see if they work.
Basically the available budget is now constrained: any further price increases in components means that those increases simply cannot be accommodated. That means that in order to deliver, lower cost components have to be sourced, one way or the other. That will mean either finding components from different manufacturers, or it means finding components (DDR3 RAM is the primary cost here) that are lower capacity: 1 GB total instead of 2 GB total RAM. This is just how it is, unless someone has some other ideas: I’m open to suggestions (on the mailing list, please).
Richard, bless him, sent an extremely comprehensive and detailed explanation about why an impedance taper is needed, and how to go about creating one: http://lists.phcomp.co.uk/pipermail/arm-netbook/2017-December/014976.html
Basically, signals need to have the same impedance all the way along and at the end of tracks. Imagine a water pipe that ends up jack-hammering and vibrating in the walls: that’s the water trying to bounce back down the pipe, and electrical signals are no different. However, in this particular case, we have to have a change of impedance towards the ends, and there are techniques for layout that ensure a smooth transition that does not result in the signal bouncing back and interfering. Richard finally found the time to explain, with diagrams, how this can be done. I am not surprised that it took him so long to find the time to do it (bear in mind that he is volunteering his time here) as it is extremely complex and comprehensive to do.
As you can see from the mailing list, I have some clarification questions for him, and, given that the cost of capacitors is so insanely high, I am hoping that they will come down somewhat before we go ahead with production.
FOSDEM is a "go" - the organisers approved a stand. Luckily we had a volunteer (thank you to Wim) to come along on Sunday, otherwise the stand would not have been approved. I will still need help - please join the arm-netbook mailing list discussion.
I have put in for three 20 minute talks: one is an augmentation to bunnie’s "The Essential Guide to Electronics in Shenzhen" on how to actually select components in the first place (this talk has been accepted); the second is an EOMA68 progress report; the third is on a proposed design of a libre, 64-bit, RISC-V SoC.
The key thing here is: the available funds on the Crowd Supply campaign are only now sufficient to pay for production of the A20 Cards and Micro Desktops (to be shipped early next year if all goes well), and a few pre-production prototypes of the three Libre Laptop PCBs. What is not available is "funds to pay for living costs, plane flights and accommodation." So, if anyone would like to help sponsor the cost of getting to Brussels, that would be extremely helpful. Also, if anyone has ideas on how to cover living expenses (short-term / part-time contracting included) I am open to ideas.
It was a post on the mailing list that spawned the idea of getting in touch first with the Shakti Team and then with the developers on the lowRISC mailing list. More on this will be given in the proposed talk at FOSDEM - the main thing is that the various RISC-V processors which are coming out right now are either university-driven research projects, 32-bit "embedded," or they are targetted at high-end "server class" such as the recent one from SiFive. The SiFive high-end processor has general-purpose buses - PCIe primarily - which means that it’s definitely not suited for embedded or industrial uses.
In speaking with the leader of the Shakti Team, we got on really well and had some extremely interesting private exchanges, which led to the development of a pinmux that could handle five separate applications and yet still utilise 95% libre and open hard macros for its interfaces. It’s documented here: http://rhombus-tech.net/riscv/shakti/m_class/.
The design concept basically takes in the best features of designs from embedded processors that I’ve seen over the past six years. The idea is to actually get it made, rather than be a "nice-to-have." Any suggestions, or if anyone would like to actively get involved, do get in touch. The idea is to make it a truly libre project (just like lowRISC), and a commercial project at the same time. We may have a sponsor for some aspects of getting it to production: I can’t confirm this or discuss details.
Why not actually help lowRISC to fulfil their goal of getting a crowdfunded SoC into production? Well, I asked them exactly that, as part of this discussion
Apparently, and you can confirm this from the surrounding mailing list messages, they are sufficiently heavily focussed on the extremely important task of ensuring that the main RISC-V core is up and running that for the team to change focus to designing a pinmux and/or tracking down the interfacs and so on would be far too much of a distraction for them, right now. I’ve known for some time that it’s something they want to do… yet I feel that it would upset their primary focus on the core (pun intended) expertise.
So, many apologies! Bit of a monster update, thanks to recovering from Shenzhen Maker Faire: I intended to put out just up until "Day One of SZMF" as its own update, over four weeks ago now.
Main message: whilst I am absolutely committed to fulfilling my promise to all of you to deliver, the available funds are now restricted to purchasing components (and one last pre-production EOMA68-A20 revision). That means that I have an immediate requirement to take on part-time contract work (help needed here), or need additional sponsorship, or need to start an additional crowdfunding campaign (several ideas have been discussed in other updates), and so on. All the while bearing in mind that all such ideas must not conflict with the promises that I have made or, better yet, actively support them.
A most immediate pressing need: I will need sponsorship for getting to FOSDEM 2018. I’ll likely have to go to Amsterdam and get a train to Brussels (I’ve done that before) as it will be significantly cheaper than a direct flight from Taiwan, and I love Holland anyway.
Your ideas and support appreciated.