On Thu, Jul 9, 2015 at 3:44 PM, Wookey wookey@wookware.org wrote:
+++ Luke Kenneth Casson Leighton [2015-07-09 15:01 +0100]:
http://rhombus-tech.net/community_ideas/laptop_15in/news/
just created a 5min video describing the casework: the parts for the base are now completed, i will move onto the screen next.
That looks really nice, luke.
thanks. took long enough :)
How do the joins work (e.g along the top edge)? Are they screwed or slotted or clipped?
it's.... complicated :) there are eight pieces involved in the construction: left, centre and right vertical parts, then front and back edges, two each, horizontally, and then the touch panel tray.
the left and right edges have a bolt hole at each end, you can see a threaded bar sticking out horizontally from all four corners: these will be M2 phillips screw-headed bolts (or allen key hex-bolts) in the real thing.
in each of the front and back parts, a nut is placed into a channel where it can be pushed down into the right place. i'm having a leeetle difficulty with that, the channel is over 15mm deep.
the front of the vertical edge has, around the bolt hole, an indentation, and then just above it is a sticking-out bit.
the end of the horizontal edge has a corresponding sticking-out bit (bolt hole goes through that, leading to the nut channel), and it has a corresponding indentation to lock the sticking-out bit from the left edge.
both of these together lock the left and front edge absolutely solid when the bolt is in place. both sets of indentation-plus-sticking-out-bit are designed to be injection-molded using only 2-part molds.
the back edge is a little more complicated: same principle, but there has to be a channel running down the middle of the protrusion that locks into the left edge, because that's where the cable goes.
the middle part is held with M2 8mm screws. 4 at the back to hold the back-left and back-right edge. there is "stepping" in the two edges, so the screws towards the back actually go down deeper by about 4mm than the ones further towards the front. the middle part therefore has a couple of "wings" with corresponding "steps" in it. i'm going to alter this slightly to be more like the touchpanel.
there are another two screws that are *horizontal* and these go into a couple of flying wings attached to the middle part, screwing into the front edge, left and right. this was a bit of a complicated arrangement because the wings are off the ground (important when you think about 3D printing), you can't make the wings too thin because the screws would break..
the touchpanel part is screwed down in *seven* places. one is in the middle (from top) 4 are along the front edge (again top). the front of the touchpanel has a hidden strut all along its length: this fits into slots in the front edges. the idea is not to pack out the entire touchpanel with plastic (or the front edges) but to leave _just_ enough in each to give some strength so that things don't bend about due to being weak. saves the screws a bit from shearing sideways in their holes.
another 2 horizontal screws into flying wings lock the touchpanel into the front edges so that both front edges can't rotate about their main axis.
that's just the main parts - there's some internal parts for the keyboard tray, battery support, PCB support.
which is why it's taken since.... january to get this far!
You know what looking at that case made me think of: project ara, with a nominally similar concept of a skeleton you put various components into (CPU, display, batteries etc). Looking into using the bus they've devloped to make bit-swapping more modular might be something to think about.
grrr. i'm extremely disappointed and angry with google for using - and extending - a cartel there, based around MIPI. MIPI LCDs are *already* cartelled: MOQs are at least 100k units, possibly even higher: nobody wanting to do a base unit will even get the *slightest* opportunity there, despite several executives being jailed a few years ago. they just got a bit more subtle about it.
... and then google went and damn well extended MIPI with this UniPro stuff. go look up how many manufacturers of chipsets there are, there. first one in - with all the patents - will be the *last* one in.
luckily there are a couple of companies just coming out of stealth mode who have much better architectures than what google came up with. heck, the first was buglabs and that was *years* ago.
It's all in the kernel already. The fancy magnetic hot-swap connectors will be hard to get currently no doubt, but any connector will do to start with.
yes i figured using 8-pin SIM card connectors would do the job. 8 pins is just enough for USB2, power, I2C and a couple of pins GPIO. and SIM card connectors are designed to be slim, spring-loaded, you can PCB-print the pressure-points and gold-plate them... perfect.
with spring-loading you can even do a side-slide-in module, but i'd prefer a magnetic latch.
The major design difference is that there is a core microcontroller on the ara design to manage the hotplugging, so that the CPU modules aren't special.
well, USB2 is hot-pluggable. and it doesn't mind being disconnected. I2C is ok for hot-swapping if you keep it on a separate bus.
Anyway this is probably a distraction as you already have a plan, but just thought I'd mention it.
as you can see above, i have... :) the CPU module form-factor i want to use is CompactFlash. connectors are available, easily. have a 3D printer, can make casework.
i've even got a first suitable processor, for the first version: the Ingenic M150 - it has 128mb of LPDDR2 RAM *on board* in the same IC package. i'm not so bothered about the fact that it's only a 1ghz MIPS, nor about the fact that 128mb of RAM is such a small amount these days: the space in CompactFlash is unbelievably tight and i at least want to get _something_ out the door successfully on the first run that's affordable.
... but i'm still concentrating on EOMA68 at the moment.
l.