On Thu, Aug 13, 2015 at 11:57 AM, Wookey wookey@wookware.org wrote:
+++ Luke Kenneth Casson Leighton [2015-08-12 22:44 +0100]:
great - they might be interested to start with the laptop casework i designed, instead of trying to retro-fit something.
I sit in an office with people working on this. The plan was to retrofit, precisely to avoid the hassle for doing the mech-eng work, which we all know is hard.
yes - 6 months and counting... but i've done that 6 months already (for a client that wants to sell libre laptops)
And we all like our lenovo boxes.
... i know - they really are good. back in 2009 i considered doing a drop-in board into some suitable legacy casework. i was still seriously considering it by about 2011... and the more i learned, the more i realised it was a bad idea to use pre-existing casework.
... but, y'know... if you haven't had the experience, it sounds like a really good idea, and y'know, maybe the team you're working with will pull it off.
the rest of the answers, below, i've written on the assumption that the team you're working with would be interested to have a glimpse of some of the kinds of questions and issues that they will need to consider, if the project is to be a success.
I did show someone your case, but quite a lot of people have been stockpiling x220 cases for a while now, so this probably won't change for the 1st iteration.
well, do invite them to consider what happens if the project becomes popular, and there are not 100 people but 500 or even 2000 people who want arm64 laptops, especially over the next 3-5 years.
also, invite the team to consider if they could source say qty 50 x220 cases in a couple of months, in order to fulfil an order by thinkpenguin, gluglug or one of the other libre-hardware suppliers.
also, wookey, do invite them to consider learning from the mistake that i made, which was to have the casework [of the tablet] designed *BEFORE* doing the component selection and PCB design [i did tell the person we paid $10k to do both at the same time, but he didn't listen].
also, here's some immediate questions to consider:
(1) can you get hold of enough LCDs? (2) are they still in stock? (3) are they affordable? (4) if not, can you find a replacement that fits *exactly* in the available space, including routing the cables correctly? (5) can you get hold of enough keyboards? (6) can you get hold of enough mouse trackpads? (7) can you get hold of the batteries? (8) have you reverse-engineered the battery connections? (9) have you checked to see if there's some DRM or communications hardware in the battery which, if you don't monitor properly, would prevent it from charging, operating, or would even cause it to blow up if you got it wrong? (10) have you checked that the battery charging IC compatible with that *very specific* battery is still available? (11) if the battery's not available, have you considered that you would need to reverse-engineer the battery casework to create a replacement?
that's just what i can think of off the top of my head. the number of questions that you have to be absolutely, absolutely positive of the answers, is just... it's overwhelming.
one mistake [one component that's absolutely critical and you absolutely absolutely have to have it because otherwise it won't fit in the pre-selected case] and everything you've done up until that point will have been completely wasted.
so it costs.... upwards of $4k to $10k (depending on how you go about it) to get a PCB made that fits *exactly* in the (pre-selected) casework....
.... and you made one mistake, have to abandon the pre-selected case, and all that money spent on PCB design is gone. wasted. you have to start again, with a totally new design that fits precisely and exactly into the next casework selected, and you have to check that aaaalll the components of the newly-selected casework also fit, and are available, and are affordable.
and that takes months - literally. all that checking.
...and by that time, the SoC - the core of the project - goes end-of-life because it's not popular enough.
so you get 1/2 way through the *second* $4k to $10k spend on PCB design and prototyping... and you have to abandon the 2nd board, select a new SoC and start entirely again.
We are actually stalled on availability of the AMD part so nothing much has moved for 6 months. We may have to move to another part.
yep... welcome to the insanity of AMD. i got absolutely no response - whatsoever - out of them, either.
repeat that for N manufacturers, fast-forward 5 years, and you begin to understand why i've been making available commonly-available (usually china-sourced) SoCs.
basically i've picked SoCs where the manufacturers either already provide the full source code, full datasheet, as well as either a full Reference Design or at the absolute very least the RAM-to-SoC connections already done.
doing DDR2/3/4 layout is an amazingly specialist task: if you're planning to support SO-DIMMs the job *may* be slightly easier.
but please, if you don't get a response from an SoC manufacturer, move on. and if it's not hugely popular already, move on. chances are that if they're silent, they either consider you to be a waste of time (too small), or they have had so little interest in the SoC that they're embarrassed to admit it was marketed in the first place, and have, incredibly, already written off the $40m investment to get to the point of having working silicon.
now, if you were going to order a million of them, they'd be knocking on your door! but... 100? they won't even give you the time of day - not even if you have contacts at ARM and are a team of prominent Debian Developers.
apologies if you've considered all these things already, and have answers. apologies also if there's so much that it seems overwhelming... but this is what you've chosen to take on, so you do really really need to be considering all of these things, and much much more besides.
There should be a 64-bit arm laptop prototype at debconf, which will also be interesting. No-idea if that will actualy see the light of day, or when...
there's another project... openlunchbox or something... they've set a very high barrier to entry.
l.