3
\$\begingroup\$

I am building a wearable prototype where I need two custom PCBs to interface between, and power, a system-on-a-chip on one end, and two sensors on the other. The PCBs will be situated about 0.5—1 m (1.5—3 ft) apart, in different housings. To make the prototype boards as simple as possible (I have very limited PCB design experience) I would need to connect:

  • 24 V @ 3 A,
  • 3.3 V,
  • ground,
  • four data signals, and
  • three differential pairs of data (@960 Mbps/lane)

between the boards.

So at the very least 12-13 lines are needed, likely more to accommodate for the current of the 24 V and ground.

Initially, I had DE-15 (HD-15) D-sub connectors in mind for this, but it seems like that would require 4-layer boards to make the routing work (not a deal-breaker, but would be nice to avoid), I am a bit worried about the integrity of the differential pairs — and most importantly — such a connector (and cable mounting) would take considerable space on the sensor side, which needs to be as small, nimble, and light-weight as possible.

The cable does not have to be removable by users, and I would imagine that PCB-mounted connectors would make for a smaller and/or nimbler prototype than panel-mounted connectors.

So I am now considering using (full-featured) USB-C connectors and cable, since they have enough lines, are built for differential pairs, and crucially are considerably smaller than DE-15's. However, it seems needlessly complicated to conform to the USB-C standard for a prototype.

Beside the issue of using a standard port which would likely fry anything else that would be connected to it (I am thinking of modifying the cable and plug housings in some way to mitigate this), is there any other reason why this is a bad idea? I know for instance that (some?) full-featured USB-C cables have integrated electronics, does that prevent non-standard uses of the cable itself?

\$\endgroup\$
6
  • \$\begingroup\$ Besides it being the worst idea ever for the reason you say, how are you going to mitigate the connector flip issue, as you must be sure you use the correct/same orientation on both ends, or your data/supply wires will be crossed? \$\endgroup\$
    – Justme
    Aug 21, 2023 at 12:36
  • \$\begingroup\$ @Justme I was thinking that I could make asymmetrical 3D-printed fittings for each connector house and cable terminal to make them unflippable. I am aware that this is not ideal, so I am open to other suggestions. \$\endgroup\$
    – JohanPI
    Aug 21, 2023 at 13:12
  • \$\begingroup\$ 960 Mbps per lane? \$\endgroup\$
    – Smith
    Aug 21, 2023 at 14:08
  • \$\begingroup\$ @Smith Yes, each differential pair carries 960 Mbps \$\endgroup\$
    – JohanPI
    Aug 21, 2023 at 14:41
  • \$\begingroup\$ What bandwidth are the differential pairs? Are they just straight Unipolar encoding with ~1GHz bandwidth or higher, or is there a more complex linecode in play? \$\endgroup\$ Aug 21, 2023 at 20:35

1 Answer 1

3
\$\begingroup\$

You originally said "board-to-board", but "board-to-board" has a specific meaning: the two PCBs are connected together directly (no cable). Instead, you say that what you need is a cable between two boards.

Connectors could be classified as:

  • Connectors for specific protocols (e.g. USB)
  • General-purpose connectors (e.g. DE-15)
  • Custom connectors

In your case, you want a general-purpose connector because a) Connectors for specific protocols should only be used for the protocol they are designed for and b) custom connectors are for companies that make 1 million products.

What you didn't say is whether the cable will be accessible by the general public.

  • If not, you need an interior connector
  • If so, you need an exterior connector

I will now recommend connectors that you can buy ready-made cables for:

For interior connector, I recommend rectangular connectors with ribbon cable between them:

For exterior connector, I recommend connectors with round cable between them.

  • Circular connectors, such as M12. Cables
  • D-shell connectors, such as D-sub. Cables
\$\endgroup\$
7
  • 1
    \$\begingroup\$ I have added in my question that the boards will be in separate enclosures, and that the cable does not have to be removable by users. I agree that a round cable should be best, but I think that a PCB-mounted (what I guess is an "interior") connector should allow for a smaller/nimbler prototype. Regardless, circular connectors seems to be at around the same size as the D-subs, so it does not really help me at this time, but the answer would have been very useful for me a few weeks ago! It does not really answer why using USB-C would be a bad idea however, especially for a prototype. \$\endgroup\$
    – JohanPI
    Aug 21, 2023 at 13:45
  • 1
    \$\begingroup\$ "It does not really answer why using USB-C would be a bad idea however, especially for a prototype" -- I believe I did: "Connectors for specific protocols should only be used for the protocol they are designed for". Today you're the only one who work on that prototype. But tomorrow? You don't know. Someone will plug that USB cable into a USB device. \$\endgroup\$ Aug 21, 2023 at 14:09
  • 1
    \$\begingroup\$ > circular connectors seems to be at around the same size as the D-subs, -- An M12 circular connector is rather small, certainly not as wide as a D-Sub connector. \$\endgroup\$ Aug 21, 2023 at 14:11
  • 1
    \$\begingroup\$ Yes that is a valid reason, but I believe that employing custom housings around the connectors would mitigate that sufficiently, especially if they are hidden within the enclosures, and are marked, for these proof-of-concept prototypes. Yes, the D-subs are wider than the M12, but height-wise they are about the same, sorry, I should have specified that. It seems to me that the considerable reduction of footprint size, physical size and cable termination size, along with providing more than enough lines, including differential pairs, makes it ideal in all ways but for the protocol breach. \$\endgroup\$
    – JohanPI
    Aug 21, 2023 at 15:00
  • 1
    \$\begingroup\$ If the differential pairs are GHz bandwidth signals, is untwisted ribbon cable appropriate? One thing that I suspect makes USB-C cabling attractive for this purpose is the 4 twisted differential pairs specified for 2.5GHz. \$\endgroup\$ Aug 21, 2023 at 20:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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