[Arm-netbook] video of laptop casework demo available

[Luke Kenneth Casson Leighton]

On Mon, May 4, 2015 at 5:29 PM, Paul Boddie <paul at boddie.org.uk> wrote:
> On Monday 4. May 2015 12.32.47 Luke Kenneth Casson Leighton wrote:
>> http://rhombus-tech.net/community_ideas/laptop_15in/news/
>
> It looks rather promising, and I can see how the limitations of the printer
> might make prototyping difficult: I've also had people print me things, and it
> has required a lot of "post-processing" on their part to get mechanically-
> acceptable results.

 yeah tell me about it.  i learned only recently that you have to
switch on "antiooze retract".

 one of the problems i'm encountering is that the inner walls are
really quite thin.  repsnapper isn't too smart, so it goes "i'll draw
that inner part - 1mm thick - as a square, and right next to it i'll
draw *another* inner part buuuut, because they're separate, i have to
go zip-zip-zip (from 1st square to 2nd and back) to get the bit of
oozing plastic off of the nozzle".

uuunfortunately, that "bit of oozing plastic on the nozzle" can be as
much as 3-4mm long.  it accumulates *per layer* on the exterior of a
wall 1mm thick that's being built up 0.2mm at a time.

you can tell what's going to happen, can't you... :)  yep you guessed
it: that "bit of oozing plastic" makes the corner of the inner wall
eeever so slightly higher by 0.2mm than the surrounding sides being
built up... by the time you get 10mm high the corner is a whopping *4*
millimetres higher than the height where the printer head is
travelling!!

that ends up with a corner pushing the bed down, causing no end of
problems including the entire part breaking free of the heated bed.

by switching on antiooze retract, the zip-zip-zip which causes 3-4mm
of plastic to be dumped in a corner is replaced with "bzzz-zip-bzzz"
which is:

* z-axis going backwards a couple of millimetres
* x-y positioning to the new location
* z-axis going forwards a couple of millimetres

and it carries on from there.  it seems to work and produces much
better quality work.

*if* however the walls on the parts i was printing were, say, the
quotes normal quotes thickness - most people print 3D parts with walls
*at least* 5mm thick, and *especially* if they were being printed as
solid parts (which i am not), then there would *not* be the problems
that i'm encountering.

looking at the parts that make up the mendel90, almost all of them are
completely solid.  holes are at least 8mm in diameter.  any "interior
walls", the smallest thickness i can see on any of the 100 parts is
about... 4mm.

by contrast, i'm printing *hollow* parts where the walls are in some
places only 0.7mm thick.  certainly, the interior support struts are
only 2 runs of the print head back-and-forth.

so it really really does depend on what you're doing, paul.  i can say
that now, after working with this mendel90 for... five months? :)


> I imagine that this would be adapted to more scalable processes once
> prototyped, however, which would infuriatingly eliminate some of the precision
> issues experienced during prototyping. Or would you first go through an
> iteration or two of higher-specification 3D printing first?

 right.  ok, the plan is as follows:

 * phase 1: get 3d printed parts working (as +ves)
 * phase 2: split all parts into two -ves
 * each 1/2 part will be subtracted from a "block"
 * this "block" will fit inside a metal container
 * the two 1/2 -ve parts will be fitted inside the metal container
 * low-viscosity injection mold plastic will be used to create a prototype part

 phase 3:

 * the two 1/2 -ve parts will be printed out on flash-thermal resin (on a DLP)
 * the two 1/2 -ve  parts will then be flash-cast to metal.
 * the resultant 1/2 *METAL* parts will then be used for *production*
injection molding.

the reason for using high-accuracy thermal resin to cast the molds in
metal is that - i hope - they will be much more accurate, so will not
require polishing.

 if they *do* require polishing (i.e. not look good enough) then what
i will do instead is investigate some casting methods that give the
cast parts a "sandy surface" look, meaning that any blemishes simply
won't show up as significant.

 $USD 20k on polished injection molds when they're only going to last
2 to 3 thousand units before they wear out due to abrasion - it's
silly.  there has to be a better way.

l.