Project update 45 of 75
We need to move the anticipated scheduled delivery dates. Put simply: I cannot (or more specifically will not) put into production let alone ship items that have not been 100% confirmed as fully-functional! Even a change of one single component (the Micro-HDMI connector for example) requires going back to the beginning of the cycle. Each round of testing of pre-production prototypes takes anywhere between three to six weeks to design, cross-check, order, assemble, receive, test and debug. I receive regular questions, "when is it going to be ready to ship" and the answer is: once there is a 100% working pre-production prototype confirmed. That happens when it happens - not on a "set date". In the meantime the preliminary shipping dates are being moved forward by 10 weeks, but please note: by definition they simply cannot be absolute.
As outlined in the previous update, I’m staying as a guest here in Taiwan of a friend who would like to set up an entirely independent product design and manufacturing workshop that can be replicated to scale. Open Hardware is therefore really important to him, hence why I am here to help him with that.
So it was rather important to him to prove that the small-scale equipment that he’s setting up can handle 0.5 mm pitch BGA ICs and 600-pin 0.6 mm pitch BGA, as it was for me to have at least one 100% working EOMA68-RK3288 "golden sample" with which to be absolutely confident that a new campaign can launch on time, overlapping with the planned shipping date for the EOMA68-A20 Cards.
The actual PCBs had arrived over two weeks ago. Chinese New Year got slightly in the way of ensuring that all the components were on-hand. As my host is running a business, we have to schedule some time to use his equipment for an entire day. There was one spare day yesterday, so for 13 hours straight we set up and picked-and-placed the components for five boards, in three rounds, because only 28 separate types of components can be done in one go. We did bottom-only components first, then set up for top-and-bottom, then top. Also we decided to try putting on the RK3288 and the amazingly expensive DDR3 RAM ICs onto two of the PCBs, to see if the IR oven would cope, and it did.
This morning I tried powering up one of the successfully-assembled PCBs. One of the reasons why I’m here is because my host has soldering equipment, heatgun, 4-channel oscilloscope, boxes with SMD components, and so on. The ACT8846 needed re-seating, and once that was done all the voltage levels came up as correct. No magic smoke genie left any of the ICs, and the RK3288 got a little bit warm but nothing excessive. However, it didn’t appear on the USB bus either.
Checking around with the multi-meter I found that, puzzlingly, the USB port marked "2nd USB Host" had a 3.3 V reading on the D+ line. Also, there was no sign of activity on the eMMC clock line or the microSD card clock line. Embarrassingly, a bit of research showed that USB OTG signals to Host that it’s operating as a full-speed device by… yes, you guessed it: putting 3.3 V onto the D+ line. I’ve accidentally swapped the OTG 1 and Host 2 designations. So, tomorrow’s task will be to desolder the USB cable and put it onto the correct port, and try again.
Oh - as hinted at above, the PCMCIA breakout PCBs arrived and I am using one of them to test the EOMA68-RK3288 Card. Anyway, the point is, my host can now announce on his web site PCB Assembly services including small-pitch BGA, and I am very very close to being able to move to the next phase for the RK3288.
This one was designed in seven days, ordered a few weeks ago, and assembled about two weeks ago. However, ironically I don’t have a Micro-HDMI Type D to Micro-HDMI Type D cable to test it! One has been ordered on AliExpress, which should be arriving shortly. I can then configure the EOMA68-A20 HDMI to output a fixed 800x600 resolution and see what happens. I still have yet to bring up the STM32F072 but needed the breakout board in order to pull a reset and re-program (boot mode) on it.
Patience is a virtue. I tell myself that as a way to remember to be persistent, calm, and not scream and go find a short pier to take a long walk off of. The two revision 2.6.2 pre-production PCBs on which the (incorrect) Runde Micro-HDMI connector was to be tried, were made using pairs of v-cut PCBs. The v-cuts are so close to the internal tracks that, on snapping and separation, it can be clearly seen that the internal copper layers are shorting out with each other.
This is just going to be one in a long list of things "Not To Do." Checking the revision 2.4 PCBs, which Mike kindly did as well, I note that the edges are perfectly clean. That means that a different manufacturing technique (and probably a different supplier) was used. Now, I don’t want to call $1700 USD on assembling the 2.6.2 revision "a waste", instead choosing to call it an "expensive learning exercise". My host is going through quite a lot of "expensive learning exercises" as well, and from what I can gather, this being Engineering, it’s pretty much par for the course.
Anyway what we will be doing is - now that my host’s equipment has been proven to be able to handle large BGA ICs - doing the next revision pre-production EOMA68-A20 here in Taiwan. It’ll not be a v-cut twin pair of boards, each PCB will be separate. It’ll also have the JAE DC3 Micro-HDMI connector, which, thanks to the exposed pins, can just be put into the IR oven just like all the other components. As that new connector has to go on anyway, the 2.4 PCB although it was a failure cannot be said to be a "waste" since a new revision had to be done anyway. The point is, it’ll be done here in Taiwan, not in China, where I can keep an eye on it.
Since writing the last update, which had to go into Crowd Supply’s schedule / pipeline, despite still recovering from yet another virus, a lot has happened. I’m planning what to put on the RK3288 campaign: perhaps I will pre-test the 15" laptop’s PCB1 in advance by using it (as-is) as the basis of an "LCD / Computer Monitor". If you put in a Pass-through Card, it would be a 15" LCD Monitor with built-in three-port USB, two watt stereo audio, microphone, and headphone jack. If you pop that out and put in an A20 or RK3288 EOMA68 Computer Card, it’s now a wall-mountable, stand-alone Desktop/All-in-One PC. It’s also very straightforward to prepare in a reasonable time-frame, and gets me 1/3 of the way towards the Laptop production as well. All I need to do is create a separate power PCB and I can take the power circuits from the Micro-Desktop for that. Very straightforward.
I mentioned on the mailing list that the A20 Cards would be going out without
systemd. The Libre Tea Parabola card, being based on archlinux, I had already replaced
sysvinit. The Devuan Card already doesn’t have
systemd, and the Debian Card has a mixed Debian/Wheezy and Debian/Testing set of packages that works very well, based on 12 years of experience of running my own laptops off of Debian/Testing without ever doing an
apt-get dist-upgrade. Now, the result of mentioning that plan was that a torrent of criticism was unleashed that, in its own right tells me that staying away from
systemd is the right thing to do. It took a week before we realised that because the A20 Card has to be released with the Sunxi 3.4.104 kernel (because it’s the only kernel that’s stable and works with all hardware), it’s simply not possible to deploy
systemd anyway (which will make things slightly awkward for the few Fedora backers), but by the time that had been established, several people had already become extremely distressed and annoyed.
Basically I’m going to have to do a special update about this.
Systemd has so many "red flags" stuck in it from so many sources that it’s even overwhelming me. However, the problem is that most people are of the type "Give Me One Good Reason". As in, "Give me one good reason why I should listen to what you’re saying about
systemd" and I simply can’t do that. I can however give them FIFTEEN low-probability reasons. Unfortunately most people cannot cope with that. So it is going to be necessary to explain, in the context of Bob Podolski’s "Bill of Ethics," what is going on here and what decisions need to be made, and why.
Last thing: it looks like we will be renting an apartment, here in Shenken. All signs indicate that this is the best all-round place to be. After confirming the latest (planned) revision A20 Card as 100% working, I can take a trip across to China to keep an eye on the production, there, of the Micro-Desktop and A20 Cards, then come back to Taiwan. Renting here near Taipei city, I can get equipment set up using Open Hardware resources and designs. I’m looking forward to it.