I've moved a little further since then.
I'm making frequent changes right now, because I'm trying to get that
over with as fast as possible so that it doesn't have to change any more.
I have the original design dialed-in and done, but this new version to
try to make a simpler pcb shape is going to be a question mark until I
get the first print.
Latest pcb is 38.2mm x 16.2mm with about 1.7x1.7 chamfer, and the edge
contacts reduced a little making a little more room on the pcb. Now I
could get 3 maybe even 4 traces around the outside of the cpld between
the cpld and the edge contacts, on each side.
Getting rid of both the polarity pin and the notches in the corners
leaves only a very tiny bit of ledge on the pin1 / pin28 corners to trap
the pcb in the carrier. The chamfer provides a good chunk for the pin1
corner, but that still leaves the pin28 corner.
But I think it *just* works out well enough if I dance several fine
lines between various limits, and if so, the payoff is a pcb with a
simpler shape and much less wasted real-estate and also more forgiving
fitment and alignment in the carrier. Without the hole and pin, it's
less of a problem when the fab router cuts aren't *exactly* where the
drawing says. The whole pcb is free to center itself in both directions,
which makes a big difference. I could get that by just making the pin
smaller and loose in the hole, but I just don't like that.
I'm using OpenSCAD now and damn it's really cool. The whole carrier,
actually, 4 different variations of it, is all described in a few K of
text and half of that is comments. And all dimensions are variables
defined at the top, easy to twiddle. Fully parametric. And even dynamic.
I have if() conditions scattered in the file all hinging off a single
variant=foo variable at the top, and a Makefile that overrides that
variable from a commandline option to generate .stl output for all 4
variants from the same source file.
One of those variants takes a full simple rectangular pcb with no
mechanically enforced polarity. I *think* that should be able to take an
old original REX1, possibly with a little simple sanding. So it may be
possible to put old existing REX's (and FigTroniX, but I assume no one
but me ever bothered to make any of those) into carriers if someone
wanted. I have some unmodified original REX1 pcb's to check that out
when I get the test prints.
All of these still have a bottom tray shape that might still be a
problem for your new boards with all the chips on the bottom. The pocket
for bottom-side parts and through-hole legs is still just 12.5mm x 36mm
(that could be extended, perhaps all the way to the pcb edges at 38.2mm)
x 1.2mm.
That pocket shape could possibly be violated, but only by compromising
one thing or another.
You could have the entire width between edge contacts by sacrificing the
side walls. That would make the empty carrier much more flexible, even
floppy, but that might not matter. It may still snap onto a pcb, and the
assembled module might be just fine with the pcb providing all the rigidity.
You could have more than 1.2mm "height" on the back side by having holes
in the floor to allow individual components to poke through. The floor
is 0.8mm right now. Both Shapeways and Sculpteo allow as little as
0.7mm, so if it's really close, you can shave another .1mm, and for a
little more, the pcb could be elevated possibly as much as 0.5mm from
where it is now and still make good contact. I have it way down at 2.0mm
from the socket floor now, and I think it still works ok as high as 2.54mm.
The floor is currently 14.1mm wide, to make it safe so that even if a
user inserted an empty carrier into the socket, there is no risk of them
damaging their socket by pulling the carrier back out. Even the original
carriers didn't have that safety. If you stick an original real carrier
into a socket with no chip installed, it is totally trapped by the
socket pins. If you had components that needed more than 1.2mm
thickness, and they spanned the full 14mm width across, you could have
the floor be as wide as 16mm, maybe even a bit more, and it would be no
worse than the original "real" carriers from a safety standpoint, and
that might leave just enough material to hold the two ends attached to
each other, to allow them to be snapped onto the pcb and hold themselves
there.
Ideal of course, would be to somehow live within that 12.5mm width,
because then the carrier could be a single common form that all pcbs
could use.
Since I think only one guy besides myself has yet ordered any of the
current pcb and carrier design, I'm going to push right ahead and switch
to using this new chamfer version as *the* version for both Teeprom and
REX as quickly as I can verify it. IE hurry up and obsolete the current
design before anyone cares about it.
I have yet another idea for a pcb and carrier combo that provides
mechanical polarity enforcement AND still allows the full use of the
full rectangle pcb shape.
By having a single prong stick out of the end of the pcb near pin1, and
have a matching hole in the end wall of the carrier. Not in the middle
like if it was to fit in the slot in the socket, off-center near pin1,
to fit into a hole in the carrier. That would also provide solid
side-to-side registration of the pcb in the carrier the way the pin did.
The pcb shape would no longer be a simple rectangle. It's no problem to
have the fab router the right shape, but depending on where the fab puts
the panel break-off points, it might be annoying and inconvenient to
sand the edges of the pcb. With all 4 sides being straight lines, it's
dead easy to just swipe the pcb on a flat surface a few times. So, that
idea might have a purpose in life, it's neat to have the option, but I
would try not to need it. But, if you are really filling up a board to
the corners, it's actually a perfectly available option. It's nothing to
add that feature to both the pcb router path and the carrier cad file.
--
bkw
On 8/21/19 2:20 PM, Stephen Adolph wrote:
thanks Brian,
I'll take a close look and see what the extent of the changes would be
to adopt this.
cheers
Steve
On Mon, Aug 19, 2019 at 2:52 AM Brian K. White <bw.al...@gmail.com
<mailto:bw.al...@gmail.com>> wrote:
Steve,
I think I have a pcb & carrier that might work better for you.
I have to wait for the samples still, but here's what it looks like:
https://raw.githubusercontent.com/aljex/pcb2molex8878/cham/pcb2molex8878.png
On this one the pcb is a simple rectangle with no funky steps in the
corners, no polarity hole, and instead a simple chamfer on the
pin1 corner.
Dimensions are just 38mm x 16.2mm
For the chamfer, shorten two sides by 1.7mm
The carrier opening is 0.1mm bigger in both directions, so this is
the
actual pcb dimensions.
--
bkw
On 8/10/19 8:00 PM, Stephen Adolph wrote:
> Just the keys that fit the Molex. 2 on one end and one on the
other end.
>
> On Saturday, August 10, 2019, Brian K. White <bw.al...@gmail.com
<mailto:bw.al...@gmail.com>
> <mailto:bw.al...@gmail.com <mailto:bw.al...@gmail.com>>> wrote:
>
> What do you mean by stubs?
>
> The notches cut out of the corners?
>
> There needs to be something to hold the pcb from sliding
sideways.
> The pin actually would do that well enough on one end, but since
> the pin is far to one end, the other end would move too much (I
> think), and you don't want two pins. So that little bit of
> skinnier pcb trapped between those vertical edges at the ends is
> enough to do it.
> And I think I will simplify that to just be 45 degree angles at
> the corners instead of those little steps the way it is now.
> That job could be done some other way, not just that way. There
> could be a single notch in the middle of the ends, and matching
> ridge in the carrier, and no walls at the corners of the carrier
> and no notches in the corners of the pcb, for one possible
example.
>
> If you didn't already have a developed design with a lot of work
> already in it, and you were aiming to have all components on one
> side anyway for other reasons like manufacuring, then the
carrier
> would be both stronger and effortless to make if the parts were
> all on top instead of bottom. The pcb has to lay pretty close to
> the bottom of the socket and so there is not much room
between the
> pcb and the floor. Plus then you could have the hole for the
> polarity pin as long as you could make room between the traces.
> The pin doesn't extend above the pcb, so you can have chips over
> the hole, but not under it.
>
> --
> bkw
>
>
>
> On 8/10/19 6:58 PM, Stephen Adolph wrote:
>
> Thanks,
> Why do you have the stubs for the molex on this drawing? I
> would have thought that such things would be unneeded
with the
> carrier.
>
> I have to take a close look at length. I think it can work
> but with changes.
>
>
> On Saturday, August 10, 2019, Brian K. White
> <bw.al...@gmail.com <mailto:bw.al...@gmail.com>
<mailto:bw.al...@gmail.com <mailto:bw.al...@gmail.com>>
> <mailto:bw.al...@gmail.com <mailto:bw.al...@gmail.com>
<mailto:bw.al...@gmail.com <mailto:bw.al...@gmail.com>>>> wrote:
>
> Here is a messy pdf
> http://tandy.wiki/images/c/c7/Pcb2molex8878_pcb_dims.pdf
> <http://tandy.wiki/images/c/c7/Pcb2molex8878_pcb_dims.pdf>
>
<http://tandy.wiki/images/c/c7/Pcb2molex8878_pcb_dims.pdf
> <http://tandy.wiki/images/c/c7/Pcb2molex8878_pcb_dims.pdf>>
>
>