"Stitching" two PCBs with jumper wires

[blastrock]

Hi,

I am designing a project where I want to join two PCBs together. One small expensive 4-layer PCB (similar to a breakout board) and one bigger but cheap 2-layer PCB. Doing everything on a single big 4-layer PCB is too expensive.

So my idea is to stitch them together with some kind of jumpers. It could be something like a U-shaped pin header: https://www.wayconn.com/ph254-1u26/ but I have two problems with that: the plastic part is too thick, I want something as flat as possible, and the pin pitch is too large, I want smaller holes and closer to one another.

Which brings me to my second idea: stitch the PCBs with simple wires. They're small, I can pack them very close together, they're easy to solder and easy to trim on the other side of the PCB. Do you think this is a viable option? And what wire do you recommend? I know absolutely nothing about wire types.

Some additional info: the pins will be used for power (5V, 100mA tops), low frequency gpio, and kind of higher frequency serial signal (including an USB 1.1 connection).

Thanks for your help!

[jmw]

Can you afford castellated holes on the breakout board? Google image search for "castellated hole module". It's neater, flatter, and much faster to solder than a bunch of jumpers.

[Siwastaja]

If extreme mechanical strength is not needed, just SMD pads on top layer as close to the edge as PCB manufacturer lets you, then best case you can just drag solder them together, worst case you need to place short pieces of wire, which is not a problem but takes a few seconds per pin more.

I prefer 1.27mm pitch so that standard ribbon cable can be used. You can strip tens of wires in one go with wide side cutters or crafting knife. Strip only a few mm; then you don't need to twist the wires but can very quickly tin them by dragging with large iron tip. Soldering them to 1.27mm pitched PCB pads is very quick, too. But in your case, if you want the boards right next to each other, no cable is necessary. Any random short piece of wire or clipped component lead allows solder bridging.

But, also consider just making the small 4-layer control board on top, not next to the other board. That way you can use standard pin headers (which also comes in 2.00 and 1.27mm pitches if 2.54mm wastes too much space.)

[blastrock]

Thanks for the replies!

I guess putting them one on top of the other is an option. Castellated holes are way too expensive though. And I'd rather avoid pin headers as they are very thick, I'd really like to avoid those few mm. So even if I put them on top of each other, I'd need something to connect them.

I know that putting them next to each other is a structurally weaker solution, but it could be good enough? I was thinking of cutting a rectangle notch on the side of the bigger board to insert the smaller one, something like 25x40mm, so that I can insert the smaller board and stitch it on 3 edges. The whole thing will be in a 3d printed case, so there won't be much force applied to the PCBs apart from their own weight. And if that's still too leak, I can add a tiny bit of hot glue. Putting them next to each other would allow for a flatter design.

If extreme mechanical strength is not needed, just SMD pads on top layer as close to the edge as PCB manufacturer lets you, then best case you can just drag solder them together

I can actually go as close as 0.3mm from the edge, but I thought that maybe just soldering them together through SMD pads would really be too weak. Couldn't a small impact break the connection?

[liaifat85]

Some possible solutions are, using U-shaped connectors, anchors, ribbon cables etc.
https://gct.co/board-to-board-connector
https://electronics.stackexchange.com/questions/352960/best-way-to-connect-two-adjacent-pcbs-by-their-outermost-hole-rows

[ArdWar]

I'd be wary using those U-shaped connector. It makes the whole connection a 3-part system. One of which not secured to board and all of its contacts are in the same plane and direction. That usually means they'll be at higher susceptibility to vibration.
It's neat solution if you just want to join two already existing boards, but if you're designing from scratch why not use two right angle connectors?
Scratch it, you mean the board will be permanently soldered to each other. That's neat solution on the other hand.

I guess putting them one on top of the other is an option. Castellated holes are way too expensive though. And I'd rather avoid pin headers as they are very thick, I'd really like to avoid those few mm. So even if I put them on top of each other, I'd need something to connect them.

Pin headers without the plastic barrier won't be any taller than castellated edges. Requires manual assembly to remove the barrier however

[Haenk]

Bare, naked wire war not uncommon in the 70s/80s, the more enthusiastic manufacturer used some sticky tape to keep the distance between the wires so they won't short out. Of course there are some nice flex cables as well, which can be salvaged from almost every electronic device. If you keep the distance very short, serial data should work, maybe do some shielding though.

[TomS_]

If castellations are too expensive, then for the expense of a couple of mm more board space you can place a standard 2.54mm pitch connector footprint, with a matching one on the other board.

Place a blob of solder into each hole and capilliary action may very well draw it through to the other hole to join them both together, or feed small wires through to solder from both sides if that isnt working well enough.

You then have an average 1.27mm pitch across the width of it without making things too fiddly to solder, you keep the two boards directly adjacent to each other for minimal height, and its basically entirely free (no additional connectors etc).

[mariush]

You could repurpose a M.2 connector or miniPcie and lock the card at the other end with solder or a screw or something.

For example
M.2 67 positions, 0.5mm pitch, 0.8mm board thickness : https://www.digikey.com/en/products/detail/jae-electronics/SM3ZS067U410AMR1000/4162230 or https://www.digikey.com/en/products/detail/jae-electronics/SM3ZS067U215ABR1500/4880009

pcie  mini 52 positions, 0.8mm pitch, 1.0mm board thickness : https://www.digikey.com/en/products/detail/te-connectivity-amp-connectors/1775862-2/5430585

20 pin right angle header, 1.00mm board thickness: https://www.digikey.com/en/products/detail/amphenol-cs-fci/10056847-101LF/5201796

[tggzzz]

Some additional info: the pins will be used for power (5V, 100mA tops), low frequency gpio, and kind of higher frequency serial signal (including an USB 1.1 connection).

That's useful extra information.

Be aware that the "low frequency gpio" signals will contain "high" frequencies, possibly higher than the USB signals. The maximum frequency in a digital signal is solely dependent on the transition time; the number of transitions per second is irrelevant. A handwaving anthropomorphic explantion is that the signal neither "knows" nor "cares" when the next transition might occur.

FFI theory and demonstration, see: https://entertaininghacks.wordpress.com/2018/05/08/digital-signal-integrity-and-bandwidth-signals-risetime-is-important-period-is-irrelevant/

Best practice would be to have a GND connection on each side of each digital signal. If those were in a ribbon cable with 0.1" pitch, the impedance would be around 100ohms. I would expect short jumpers on 0.1" pitch to be similar.

[MrAl]

Overlap the boards and solder together.

Alternately, copper tape soldered on top and bottom at the edges, all the way across or just two or three places.

[Siwastaja]

Some additional info: the pins will be used for power (5V, 100mA tops), low frequency gpio, and kind of higher frequency serial signal (including an USB 1.1 connection).

That's useful extra information.

Be aware that the "low frequency gpio" signals will contain "high" frequencies, possibly higher than the USB signals. The maximum frequency in a digital signal is solely dependent on the transition time

This is good to remember. "Low-frequency GPIO" is still relevant in one sense: it allows the designer to band-limit the signal, e.g. by RC filtering, which also doubles as ESD protection. Highly recommended. I prefer something like series 100R + old-school high-capacitance TVS (+ extra capacitance in parallel) + another series 100R.

[tggzzz]

Some additional info: the pins will be used for power (5V, 100mA tops), low frequency gpio, and kind of higher frequency serial signal (including an USB 1.1 connection).

That's useful extra information.

Be aware that the "low frequency gpio" signals will contain "high" frequencies, possibly higher than the USB signals. The maximum frequency in a digital signal is solely dependent on the transition time

This is good to remember. "Low-frequency GPIO" is still relevant in one sense: it allows the designer to band-limit the signal, e.g. by RC filtering, which also doubles as ESD protection. Highly recommended. I prefer something like series 100R + old-school high-capacitance TVS (+ extra capacitance in parallel) + another series 100R.

Yes in some cases.

In other cases doing so may erode the T_setup and (especially) T_hold times needed by the receiver.

[blastrock]

Thanks for all the answers! I'll think a bit more about what solution to go for, thanks for all the material. And I'll study more the frequency stuff too...

[Terry Bites]

Mr Samtec has tons of board to board.
You might want to think more in terms of a mezzanine.