This is a big update: a lot has been going on, with bursts of activity and the need to adapt to changing circumstances.
After returning to Hong Kong, we applied for visas and have been granted dual-entry 30 day ones. China Travel Services inform us that after getting enough dual-entry visas we should qualify for multi-entry ones. What's nice is that CTS can just try each in turn in one application (whereas if you go along yourself to the Chinese Embassy they turn you away and you have to join the 3-hour queue again).
Zhuhai is really nice, and Horizon Cove (the gated community) is fantastic, it's basically 50% jungle. The place even has two tennis courts, and I've found someone who gives tennis lessons. It's been eight months since I played, and I haven't really got enough exercise since. Time to start again. Internet access is, once again, sporadic, more on this later.
The latest pre-production prototype of the Micro-desktop is a "go". The A20 board... not so "go". Just as happened four years ago with the Micro-HDMI connector selected at the time, the layout was wrong. It happens. Another $800 USD mistake. So, Mike suggested doing a two-layer "test" Board, with just the micro-HDMI connector on it. Two-layer PCBs are really cheap and really quick to get made (three-five days instead of four-ten).
Now, from the PCBs that were made up with the incorrect Runde micro-HDMI connector layout, Mike cut the legs off and hand-soldered it down to the PCB so that it would at least be possible to test the new HDMI track layout and check that it works. Ironically, it's been necessary to return to the same layout used three to four years ago, but it still needed testing.
So, he sent that test board to me, and it became immediately obvious that the Runde Micro HDMI connector is not ok to use: it's unfortunately a whole (shock, horror) 0.4mm too high up, pushing it above the line set by the 2mm high MicroSD card and thus reducing the component height limit for the BOTTOM of the PCB.
We went through all this three years ago for the very first EOMA68-A10 Cards. I cannot say it's fun to go through it all again. Speaking to another supplier in Huachiang Road, they have a sample of a Molex mid-mount connector... which has been discontinued. They have exactly the same one as Runde, but it took three days of miscommunication to establish this. Ironically it looks like being in China isn't turning out to be as beneficial as expected for these unusual parts!
Currently under investigation is the JAE DC3 Series, part number DC3R019JA1. A supplier that has 400 of them (a part-reel) has been found; that at least would give the means to complete up to 400 boards. JAE list the part as "active", so it should be possible to obtain... now all we have to do is find out the lead time and see if there's a distributor with a couple of reels of 1500. What I will do is simply overbuy because I do not want to have to go through this again. They're $0.15 each, so it's not exactly a problem!
Anyway, of the two boards that were made up, only one of them started up; the other we suspect has a short (around the HDMI connector). I've asked Mike to replace the 4.7uH inductors with a different type that's known to be a power inductor - identical to the ones that are on prior PCBs - we'll see what happens.
This is literally one of the most complex PCBs I've ever done, and after the initial stressful flurry of development just trying to cram in connections as best I can, I've spent almost two weeks now refining and cleaning up the layout to something approaching acceptability. If you'd like to see the PDFs of the schematics and the PCB layout, you can [download them, and various datasheets] (http://hands.com/~lkcl/eoma/rockchip_rk3288/).
Also, as a test board, to make sure I'm familiar with the RK3288 before going ahead, I've received an RK3288 Firefly courtesy of Firefly and have managed to get a Devuan root filesystem with mainline linux kernel and u-boot up and running. The resources and board bring-up will be documented online.
The RK3288, despite being now almost two years old, is an awesome and really rather well-supported processor. Xiarun Li, a Chinese software libre developer who recently started working at Rockchip, is being extremely helpful. He is also working on a software libre VPU decode engine, adding in mainline support into gstreamer and the linux kernel (as a v4l2 driver) for the RK3288's VPU. By the time the boards are ready, that software could well be sorted out as well, which would be very nice.
Rockchip's Management don't quite understand what all the fuss is about with this whole "software libre community" thing, and the whole "GPL compliance" thing. However, given that Google has a hell of a lot of money, and Asus (also a hell of a lot of money) picked the RK3288 almost two years ago to create a Chromebook around the RK3288, Rockchip seem to listen to them, so GPL-compliant source code has been getting out the door slowly ever since the RK3288's release. The only complaint - if it can be said to be a complaint - is that the prevalence of the RK3288 is so high that the signal-to-noise ratio on "HOWTO install linux RK3288" is correspondingly high, making it actively difficult to get stuff up-and-running. The #linux-rockchip community has therefore been absolutely invaluable. Many thanks for your help.
What possible relevance could the U.S. Election have on this crowdfunding campaign? Well... up until the 8th of November, all the funds were in the sponsor's USA-based bank account, and all the costs are in RMB. My partner pointed out that traditionally, the US Dollar has gone up in value after an election, but this time, thanks to the high probability of Donald Trump being elected, it might actually go down, resulting in the value of the funds raised having something like 10 to 20% wiped out in an instant.
Rather than gamble with people's money, a decision was taken to scramble very very fast and to get the money out of USA accounts and into China (RMB) accounts as quickly as possible. Particular care had to be taken to ensure that the money did NOT arrive in Hong Kong Dollars, because the HKD is exchange-rate-locked to the USD, defeating the object of the exercise.
If you recall: just before the election, Donald Trump promised to bring jobs back to the United States, to the extent of considering putting tariffs onto imported products. China responded that if that was done they would immediately add a whopping 40% tax onto high-value goods such as iPhones, and would respond tit-for-tat proportionately on any further moves made.
Even on election day, various United States citizens apologized on Slashdot to the countries of Canada and Mexico, whose currencies had over 10% wiped off them in minutes due to off-hand remarks made by Donald. So it wasn't "idle talk" or "speculative theory": actual currencies really were adversely affected. Basically we did the right thing: 50% of the funds is now in RMB, in China.
I know in advance when there's going to be a news report that there was yet another massive, world-wide DDOS attack: access to gmail (or any other internet page) slows down to 6k per second and I have to refresh pages five to twenty times for them to load. I've had it with China internet quality. I've installed cyrus IMAP server on my laptop, prayer webmail service, and have hand-patched offlineimap to improve its usage under these extreme conditions. Due to the fact that I have 200,000 messages to pull over it's going to take several days to complete: 36 hours has managed to synchronise 30%. The rest I will complete in Hong Kong or Taiwan.
For a number of reasons I decided to get a new development machine. It was discussed online- I felt that it was very important to be completely transparent about justifying spending RMB 19,000 (around $USD 2,600) on a single laptop. Here's the installation report; it's literally the most extreme computer I've ever owned - with specifications going well beyond anything that I ever imagined owning in my lifetime.
The problems with my three-plus year-old 13 in MacBook Pro began to add up immediately when I first set it up: they included a constant barrage of SATA bus reset errors which were temporarily fixed by setting min_power on the host bus. The sheer extreme way in which I push machines to their limit was beginning to tell. 8 GB of RAM simply is not enough to run two web browsers (for component research purposes and communication), dozens of PDFs, a full x86 VM to run the CAD software and running OpenSCAD for 3D development. But I didn't know that at the time that I bought the macbook pro.
If the Dell XPS 9350 could have been upgraded to 32GB of RAM, I would have gotten it in an instant: it's fully supported by Debian, and Dell are extremely understanding. However, it doesn't. So the risky decision (which luckily paid off) was made to evaluate and buy an Aorus X3 Plus v6 "gaming" machine. It's by no means perfect, but after three days of configuration and adjustment, it's now my main machine.
The MacBook Pro will be kept as a backup, with data synchronized over to it on a regular basis.
Long story, but I have an associate with whom I've been talking for two years, who lives in Taiwan, is developing his own product, and has recently bought an IR PCBA oven so that he can do his own BGA component assembly. He's offered to help with the RK3288 PCBs, so we made a snap decision to go visit him for a month (we have to be out of China by the 20th anyway). I will be helping him out as well. Also, it's the opportunity I've been waiting for to go visit various component suppliers (particularly Amphenol who designed the PCMCIA socket on which the Housings are critically dependent: they designed the world's last remaining affordable SMT PCMCIA socket, which both they and Runde supply).
Now, the general plan is to go and visit Rockchip's HQ, and it turns out that Fuzhou is nearly directly west of Taiwan by around 150 km across the sea. We nearly went to visit last week, but it turns out to be a 900 km train trip that, all told, would have involved 10 hours of travel. With the 30-day visa limit coming up, it was too much.
So, instead what we are considering is going to Taiwan, working with my associate for a month to get a few EOMA68-RK3288 samples ready, then go visit Rockchip HQ and demonstrate the samples to them.
I'm using the mailing list to ask for people's help and advice, which I feel is very very important. The updates are a summary of "what happened", so if you'd like to get involved, please do join the mailing list, you're most welcome: http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook