3
\$\begingroup\$

This question extends upon "Connectorless USB on a PCB". Our requirements are for a smaller connector than standard USB, and also 5 leads instead of USB 1.0's 4. The connector will be used perhaps twice in the lifetime of the board, for ICSP / JTAG programming / firmware updates. The product is a low-cost throwaway device somewhat like the DigiSpark.

Edit: Also, the connector will be inaccessible without breaking open the enclosure, so it is production-access-only. If end-customers break open a $5 toy, well, they have earned the right to destroy or reprogram it!

Cost and added height of adding a connector of some sort, such as a microUSB socket is highly undesirable.

We would like to do our programming / update using a standard USB to MicroUSB cable such as is shipped with many modern smartphones, plugged into a configuration & testing board (not an actual USB port).

Edit: [28-Oct-2012]

  • It turns out that microUSB is not viable, if only because of the thin PCB required.
  • USB 1.1 / 2.0 are not viable due to having only 4 contacts on the USB-A end: Thanks @DaveTweed!
  • Top 5 contacts of USB 3.0 Type B cable (image) appear PCB friendly, so question remains open

The questions:

  1. Has anyone had success using such a connector etched onto a PCB? Could you share a link please?
  2. What thickness of PCB would we have to go into production with, for such a connector to work?
  3. Is there any available reliable footprint we could use for such a connector?
  4. What else would we need to keep in mind - such as board cutting/milling tolerances, stiffness of PCB, rounding of "connector" corners, ...

SOLUTION: Added here for reference.

The solution that worked out best for us was a 3M SOIC narrow test clip, which has 7 contacts to a side, and a conventional IDC 0.1" cable going back to our programming board.

No additional milling is required, half of a standard SOIC pad footprint close to the edge of the board, with solder bumps on the pads, works out perfectly, and the clip grips the board firmly. Varying PCB thicknesses have been tested, they all work fine.

A bit of a convenience hack we added: Positioning two of the nearby components equidistant from either edge of the SOIC "programming pad footprint" ensures that the clip can be attached very quickly and perfectly aligned to the pads.

In laying out the programming pad pins, we ensured that a misalignment of the clip would not cause any problems either due to shorts or problematic signal injection. Needing only 5 of the 7 pads simplified this reshuffle.

Hopefully this solution would help others with similar requirements.

An observation: Pogo pins suitable for 0.05" spacing worked out way too expensive compared to the SOIC test clip approach.

\$\endgroup\$
  • 2
    \$\begingroup\$ A couple of comments: First, if you make it look like USB, people will plug USB into it. Make sure that won't damage anything. Second, the "standard USB" connector at the other end of the cable isn't going to have your fifth contact. Third, at some point, the incremental cost of the PCB to do the milling, etc. required to create your "connector" will rival the cost of an actual connector. \$\endgroup\$ – Dave Tweed Oct 28 '12 at 11:53
  • \$\begingroup\$ The "disposable" device ships with no external ports, the "port" is blocked by the enclosure. Good point on ensuring no damage on plugging regular USB in, thanks. \$\endgroup\$ – Anindo Ghosh Oct 28 '12 at 12:27
  • \$\begingroup\$ @DaveTweed Thanks for the useful insight, it got us to re-examine the standard cables at hand. It turns out that USB 3.0 might still be an option, and the upper 5-contact section of USB 3.0 Type B is still a bit smaller than regular USB-B. Please see the notes I have added to the question. \$\endgroup\$ – Anindo Ghosh Oct 28 '12 at 13:02
3
\$\begingroup\$

If this is for production access only, then here's a trick I used once. The PCB (standard 0.062" thickness) was designed with SMT pads along one edge (both sides) so that a 2×7 0.100" JTAG header could be soldered to it, with the pins sticking out from the edge. This was useful for debugging prototypes, which needed more or less constant access to the JTAG (and weren't in cases).

In production, the connector is omitted, and I used a slightly modified DIP clip to access the pads for the connector. I just needed to bend the contacts inward a bit to get adequate friction and contact pressure.

Instead of the DIP clip, you could use a standard card edge connector as well.

\$\endgroup\$
  • \$\begingroup\$ Usually the "maximum number of insertions" spec of card edge connectors precludes them from production test use. \$\endgroup\$ – markrages Oct 28 '12 at 22:50
  • \$\begingroup\$ @DaveTweed The DIP clip connecting to bare pads at the edge sounds promising... ordering a few to test with. Thanks! \$\endgroup\$ – Anindo Ghosh Oct 29 '12 at 5:02
  • \$\begingroup\$ The DIP clip works out brilliantly, and the overall cost between PCB fab and fixture turns out to be an incredibly tiny fraction of the pogo pin or (gasp!) Tag-Connect approaches. Thank you, @DaveTweed! \$\endgroup\$ – Anindo Ghosh Oct 30 '12 at 3:20
5
\$\begingroup\$

Using pogo pins in a test fixture, as Dave Tweed mentions, is an industry standard technique for programming PCBs in production. Having a custom fixture made can be cost effective for large volume runs. Using a larger number of spring loaded pins to test a board is sometimes known as "bed of nails" testing.

For a couple of recent projects, I used cables from Tag-Connect in conjuntion with a set of pads laid out on my PCB. The cables have a header conenctor on one end (to connect to a JTAG programmer). The molded connector on the other end of the cable has a set of pogo pins, along with three alignment pins, that match up with test pads and alignement holes on the PCB.

Tag Connect Plug

I have found the cables useful for debug and prototyping. The same pads can then be used with a custom fixture for production test.

\$\endgroup\$
  • \$\begingroup\$ Thanks for the input. We did consider pogo pins, but the cost for 0.05" compatible is prohibitive in our part of the world. Tag-Connect does not even have a source in India that we could locate, besides being even more expensive an approach than pogos. A $11 (including shipping to India) SOIC test clip gave us the tiny footprint we wanted on the board, at a fraction of the cost. \$\endgroup\$ – Anindo Ghosh Oct 30 '12 at 6:52
4
\$\begingroup\$

The "standard" way to get production-only access for programming a chip on a board is to build a mechanical test fixture for the board that holds it securely, and use "pogo pins" positioned to access pads laid out on the board for this purpose. This gives you the most flexibiltiy in terms of the layout of the board.

Of course, you can add additional pogo pins to the fixture to aid in functional testing of the board as well.

\$\endgroup\$
4
\$\begingroup\$

If you have space, edge card connector fingers would be fairly simple. If you use only one side, you can modify a receptacle with a shim to account for your apparently thinner board. This may not be quite as suitable for very large volume as a custom pogo pin fixture, but it is likely cheaper to rig up, and the parts are more widely available.

For low volumes, a very simple choice (provided your programming operation is no more than a few seconds and includes a verify) can be to use a single row of plated through holes for a .100" header connector. Instead of soldering a header into the board, you solder one to the programming cable (or plug it a female socket on the cable), insert the header pins into the board, and then tilt them sideways, holding pressure with your hand until the programming is complete.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.