6
\$\begingroup\$

I can not find a true pass-through KVM Switch for USB-C (Thunderbolt) to buy, so I am thinking about maybe building one myself. Currently I am just plugging and unplugging the Thunderbolt (USB-C) cable from my docking station to the laptop I want to use at the moment.

Why is there no Thunderbolt KVM switch that does exactly that? Is this technically not feasible or what is the reason none seem to be offered by the industry?

Some active USB-C KVM switches are offered. What I understand is they do not offer full Thunderbolt bandwidth and they are also rather expensive at ~$150-250.

I am thinking about just buying three Thunderbolt cables and having them switched with a mechanical switch. I mean, it works if I just unplug one laptop and plug the other in. Why isn't it available already in a box I could buy? Is there a technical barrier for that? What mechanical device should/could I use for such switch (the actual switching) if I were to build one myself?

Ideally the switch will keep the power delivered (maybe from a second external USB-C power adapter, I have them available) and just emulating plugging/unpluging the cord for the data lines.

So the questions are:

  1. Is this is feasible at all?

  2. What could I use to mechanically switch the Thunderbolt signal cables?

\$\endgroup\$
5
  • 2
    \$\begingroup\$ Dual host docking stations for thunderbolt exist. But thunderbolt is very fast. Switching must be performed by a bridge IC since it is basically PCI express over a cable. \$\endgroup\$
    – Jeroen3
    Commented Feb 17, 2020 at 8:32
  • \$\begingroup\$ @Diego I'm interested in something like this too. Did you ever make any progress? I'm looking at using 3 24-pin USB C break-outs with 6 4PDT analog switches, but I have no idea if that would actually work. \$\endgroup\$
    – Jonah
    Commented Jan 22, 2021 at 2:53
  • \$\begingroup\$ No, I ended up just manually switching the laptops. Good thing they only need one single thunderbolt cable to connect. \$\endgroup\$
    – Diego
    Commented May 19, 2021 at 21:11
  • \$\begingroup\$ @Jonah I have exactly the same idea as yours. Have you tried implementing it? I don't see any reason why it should not work \$\endgroup\$
    – dbdq
    Commented Jul 12, 2021 at 8:12
  • \$\begingroup\$ @dbdq No, I haven't gotten around to accomplishing this. It's not ideal, but right now I'm making do with a Level1tech Displayport KVM \$\endgroup\$
    – Jonah
    Commented Jun 3, 2022 at 3:13

1 Answer 1

1
\$\begingroup\$

I am thinking about just buying three Thunderbolt cables and having them switched with a mechanical switch. Is this is feasible at all?

Technically yes, it is feasible, but not that easy.

What could I use to mechanically switch the Thunderbolt signal cables?

TBT3 is a two-lane high-speed differential bus running at 10Gbps/lane. So to switch 8 signals you will need 8 RF-grade relays with 50-Ohm coaxial construction, like COTO or Standex-Meder reed relays, plus similar ones for several sideband signals. Relays of this kind run for about $8-$9 each. And you will need a very carefully crafted PCB layout to maintain 80-100 Ohm differential interface for each pair of signals. The electromechanical approach is frequently used in production test equipment.

Alternatively you can try to multiplex (2:1) all super-speed signals with an analog switch IC, like TMUX646 from TI, it should be good for 10Gbps rate,

enter image description here

You will probably need to implement a full-scale CC-PowerDelivery controller on each downstream port, because the host will not configure itself for TBT without knowing the cable characteristics and operational mode of link partner, but it might be possible to get along with the existing PD function in PC host.

\$\endgroup\$

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.