Larger via diameter in future revision?

[Tommy_2Tall]

Hi!

I just want to drop this idea to see if I might be on to something other DigiSparkers would like as well.

Would it be a terrible idea to slightly increase the through-hole diameter so that we could use snap-off SIP sockets instead of those relatively "huge" stackable headers?

Some considerations would have to be made when working with shields that have a component height larger than ~2mm but in a lot of cases the SIP headers seem to be an excellent low-profile option so that your DigiSpark and 2-3 layers of shields doesn't look like the tower of Pisa.

The SIP sockets I've found so far seem to mate with a 0.7 - 0.9 mm (~0,034") diameter PCB through holes.

Just comparing the sockets I ordered with the DigiSpark header holes I think that it should be possible to use a suitable diameter without making it too much of a "loose fit" for other types of trough hole components.
 I am a complete n00b in this area so I may have missed something completely basic like thermal design or common component heights..

Any thoughts and reflections on the subject from our more experienced members?

[digistump]

I haven't used SIP sockets for anything but sockets - but I will give the reasoning behind the huge stackable headers - I saw many tiny arduino projects that used special headers and I hated that - so a founding principal of the Digispark design is that it uses common components, at least when it comes to its interfaces - I don't want to ever have to buy a special part to be able to plug into my MCU.


That said - this is an interesting idea because 1) things do get a bit tall and 2) I'm always open to any idea that is compatible with the existing setup and this sounds like it might be. If someone were to make this work using the reference design that is available then we'd certainly consider it - but it would be essential that the standard headers still work perfectly as we'd continue to distribute them with those.




Thanks for sharing the great idea!

[Tommy_2Tall]

Hi!

Thanks for the swift response. Excellent support, just as we've all gotten used to by now. 

Regarding the backwards compatibility I don't think this would change anything for the worse;
the SIP socket headers use the same 0,1" (2.54mm) spacing between the center of the holes,
 it's just an almost insignificant change in the hole diameter that would be needed.

I'm not sure how to "make this work in the reference design" so I'll just supply some detailed measurements and hope I can convince you with those numbers and some solid reasoning instead. 


It's a little embarrassing but I should have done some actual measurements back at home rather than referring to measurements from a datasheet that proved to be for some other sort of (solid moulded DIP) socket.. or maybe they where listing radii but mistakenly put a diameter symbol there.. never mind...
I just assumed it would be near enough but the measurements I mentioned were WAY off so I hope I didn't lure anyone into getting their caliper out to verify those clearly incorrect pin width measurements.


I did some more accurate measurements when I got back home yesterday.


DigiSpark power and I/O vias:
I had some problems measuring the exact inner diameter of the plated vias for the power and I/O pins and I don't think the Eagle file are public(?).
I tried to measure by eye and caliper (holding it over the opening) and by measuring wall-to-wall between the I/0 holes on a proto-shield but these measurement could be more precise.. for example if I found a suitable needle that fits snuggly in those holes.
I'm getting a measurement of approximately 1.3 mm inner diameter with the plating applied.
 Is that fairly accurate?

Do be able to get a tight but usable fit with SIP sockets that would have to be increased to.. 1.4 ~ 1.45mm?

DigiSpark headers:
0.6 x 0.6 mm square pins, which equates to a "diameter"(/diagonal) of 0.848 mm.
That's a lot of air already between the pins and the inner diameter of the vias..
If you push one of the corners of the pins to one of the via walls  when aligning/soldering them there's approximately 0.45 mm of "air"/solder on the opposite side of the pin. Quite a loose fit. 
The plastic body of the female header (same for stackable headers) has the standard 2.45 mm by (2.45 * pin count) PCB foot print and a height of 8 mm.

SIP snap-off sockets:
As mentioned above, the spacing/pitch is still 2.54 mm (0.1").
The plastic part of the SIP sockets has an approximate PCB footprint 2.45 mm by (2.45 mm * pin count) and a height of 3.0 mm.
The thicker part of the SIP socket pin is approximately 1.36 mm diameter and extends ~1.25 mm below.
That's the part I want to fit inside the vias for a tight and well-aligned fit (precision machined cylinder inside a slightly larger  precision  machined cylindrical hole).
The rest of the pin below that section should mate reasonably well even if the cone-shaped parts aren't pressed tightly against each other.
My sockets seem to "lock on" with just 1 - 1.5 mm of the thin pin inserted into the socket.
When stacked as tightly as possible the "stacked height" is approximately 4.25mm (plastic body + "through-hole" part of the pin).
The corresponding measurement for the standard headers is ~11mm for a female + male pin (8mm plastic body + 2 mm plastic spacer on male pin +  ~1 mm PCB) or 17 mm for stackable female headers

DigiSpark component height:
All 3 of my DigiSparks have a maximum component height of of approximately 2.54 mm (PCB surface to the top of the voltage regulator).
With SIP sockets that would still leave 0.46 mm of air between the top of the voltage regulator and the bottom of the first shield.
With a standard female headers there would be 5.46 mm of air between the top of the regulator and the first shield.
You could fit 3 voltage regulators on top of each other in within those 8 mm, though with very limited air cooling. :-)
I did have the 8bit port expander shield soldered up with DIP socket and all but I forgot to measure the height of a DIP IC + DIP socket. :-S
I'm guessing you would need an extra layer of SIP sockets to clear that height.. :-S


Possible side effects:
Increasing the via hole from ~1.3 mm  to 1.4 mm would give the standard headers an additional 0.05 mm of slack in either direction from the center of the 2.54 mm spaced holes.
I think that's negligable considering I haven't seen any complaints yet about the 0.225 mm slack in either direction those headers have with the current via hole diameter.

Component height might become an issue, for example if you have DIP sockets + DIP ICs on your shields.
Adding a "raiser" layer of DIP sockets may look a bit weird but it would give almost the same height (~ 8.5mm) as a standard female header (8 mm).

Pin compatibility is obviously the major issue.
Personally I would solve that by using SIP sockets on some of my DigiSparks and standard headers on some of them, depending on what I would want to use them for.
It's always good with extra options that don't have any unavoidable side effect. ;-)

For example I would probably use standard headers on one or two "prototyping"/breadboarding DigiSparks and when I have a project that I would like to make more permanent and more presentable I would check if it would be possible convert that build onto a DigiSpark (and shields) with SIP sockets to get a more proffessional/compact look and smaller enclosure.

I haven't done any extensive tests (yet) but logic dictates that the 0.6x0.6 mm square pins and the 0.5mm diameter SIP pins do not mate that well in their normal form..

The SIP pins would fit in a standard female header but it would probably not be a tight/secure fit.

Trying the opposite, square peg in a smaller round hole, would most definitely end in tears and frustration.

While snapping of a few SIP socket segments I actually separated the machined metal part from the plastic and it occured to me that it might be possible to solder together some sort of Frankenstein-ian "square pin to SIP" adapter pins (sanding/filing down a standard male header and securing it in the SIP socket with some solder) but that would only be suitable for prototyping purposes.

OK.. time to hit the brakes.. this post is getting huge.
Hope my "rant" makes sense and that I might be able to convince you. :-)

[Bluebie]

I wish there was some kind of pin socket which could take regular wires. I find most of the little wires I use to hook things up when soldered aren't thick enough to reliably connect in to the pin header sockets used on arduino and supplied with digispark, so I have to break apart pin headers and solder headers on the end of every little wire. It'd be rather nice if I could just stick the wires straight in to something which could still accept a pin header, like the sort of sockets on breadboards. It doesn't bother me much that the digispark gets a bit tall with stacked headers, but I agree it would be a little nicer if it was a bit flatter.

[Tommy_2Tall]

Hi Bluebie!

In that case the SIP sockets seem like something for you...
Geez.. I'm starting to come off as a very persistant infomercial-host. :-)


It won't be as simple as sticking the wire into breadboard but it's certainly easier than breaking apart standard female headers or soldering wire to the short end of standard male headers.

You could use the funnel (and possibly the hole for the thinner part of the pin) in the same way that the R/C people attach their (HXT) bullet connectors to their battery packs:
fill the "funnel" with solder, dunk the wire(s) in there, remove heatsource.



Since the female part of the socket is clearly exposed there is no dismantling necessary and you should be able to desolder the wire(s) if needed.. That socket would be filled with solder though but it could be reheated to accept another wire later on.
The thinner part of the pin is 0.5 mm so without the solder the hole in the bottom of the "funnel" is probably a poor fit for magnet/signal wires.

[Bluebie]

That does sound like an improvement anyway. If it were an option I think I'd rather order my digisparks with these thingies instead of the usual header pins.

[MichaelMeissner]

One source of getting those headers is dipmicro: http://www.dipmicro.com/store/HDR40X1FM  From the picture, it looks like this would mate with that: http://www.dipmicro.com/store/HDR40X1MM


An alternative might be to attach a system that would allow you to connect and disconnect pins.  There are several different format that I've seen, though most might make the digispark bigger, which might be a problem.

• Dipmicro sells a wafer system with male and female headers, and some crimp pins.  The combo pack is out of stock at the moment: http://www.dipmicro.com/store/SET-HDR.
• Pololu sells various crimp connectors: http://www.pololu.com/catalog/category/70.  Note, you will also need male crimp pins: http://www.pololu.com/catalog/product/1931 and female crimp pins: http://www.pololu.com/catalog/product/1930 or wires with pre-crimped terminals: http://www.pololu.com/catalog/category/71
• Pololu also sells terminal blocks that have 0.1" spacing that you can put wires in without soldering (you have to solder the terminal block, but to attach a single wire, you don't need to solder it).  Note, these have two rows of pins for stability, and for a digispark, you likely would need to let one pin hang off into the air.  Adafruit and Sparkfun also sells these, though Pololu seems to sell them at a better price point: http://www.pololu.com/catalog/category/117

[Tommy_2Tall]

OK.. so now have managed to find the DigiSpark source package that was published on the order status page, installed Eagle CAD, clicked a round a bit and managed to generate a "drill legend" for the DigiSpark.

Apparently my new and improved measurements were not that accurate... :-)
The largest holes (= power and I/O vias) are 1.03 mm according to the drill legend.
So 1.4 mm holes are possible with plenty of room to spare with the 2.54 mm pitch but it's quite a significant change in hole diameter... bummer! 

[Mark]

@bluebie
 
I have used cat 5 cable (solid) in the past.
Not exactly the perfect fit, but with some 'kinking' of the wire, it fits well.
 
The other oprion is to use one end of the jumpers as per these.
http://yourduino.com/sunshop2/index.php?l=product_detail&p=94
 
These stackable headers don't resolve the height issue, but do fit the sockets, without resorting to bending to 60 deg. http://yourduino.com/sunshop2/index.php?l=product_detail&p=220
Yes they aren't the right size, but other sizes must be available.
If there is enough interest I'm happy to email Terry and ask.
 
 
re SIP
You may wish to look at the specs for the number of insertions.
This could possibly be low as they aren't designed as a plug/socket.
 
Mark

[bjh]

I love cat5 Ethernet conductors, stranded for hookup wire and flexible applications and solid for breadboarding and prototyping, but I've had poor luck getting good connections with headers. 


For those I split old ribbon cables down to anywhere from 2 to 8 lines and solder headers to each end to make short busses to bridge to a breadboard and use Ethernet wires from there.  Now that I've got a few in my kit I haven't had to make more in a while.  I also sprung for a package of premium breadboard jumpers for when I need one-offs.


For more permanent projects, I usually solder to headers or right to the board.


I've played with machine pin sockets some because I scored a few tubes of them at surplus stores super cheap.  I've cut them up and used them header-style to make modular laser shows and all sorts of things, and haven't had trouble with them wearing out, though the little pins bend much more easily than true headers and break if they get bent more than a couple times.  They're also great for soldering things with thin or short legs or soft leads into little breadboarding modules.

[bjh]

Ooo, now that I see these next to each other, I realize that cat5 wires connect quite well into machine pins and little "adapters" of headers topped with machine pins would be much more compact than the ribbon cables. To the soldering cave!


(Though they won't replace the ribbon cables when I want to be able to extract and re-attach the microcontroller)