For my project I'm searching for chips specialized in USB galvanic isolation. It looks like there is no real product on the market. I found Silanna and AdvancedPhotonics that respectivly have products for USB2 high speed and USB3 superspeed.

However, Silanna stopped the production in 2019 and there is not much information about AdvancedPhotonics reliability.

Do you know chips that can do isolation for such speed? Any idea to achieve the same with discrete components?

  • 1
    \$\begingroup\$ Maximum I've seen is 12 Mbps but, on a different type of data I've made mag isolators run at 600 Mbps (a very bespoke job). \$\endgroup\$ – Andy aka Feb 4 at 11:48

Do you know chips that can do isolation for such speed ? Any idea to achieve the same with discrete components?

Since USB 2 is a bidirectional bus, the USB2 HiSpeed isolation is actually pretty challenging. It's possible there's a silicon solution out there, but I've yet to see it.

I'm not aware of any USB3.0-capable isolators; it's interestingly probably technically easier to isolate the SuperSpeed lanes, as they are unidirectional, on the one hand, but on the other hands, bandwidths beyond 15 GHz are really challenging.

No, you cannot reasonably build that from discrete components. (I mean, sure, you can. It amounts to building a least-latency optical transceiver system. You should probably apply to a company that builds network optics afterwards.)

I know that there's USB SS-via-fiberoptics adapters. Maybe that's an option to you? It's not "just an IC", but a relatively complex thing. It, due to the very nature of not supporting all legacy things, isn't compatible with all devices, and costly (as in: often cheaper to buy 2 PCs, equip them with network cards and fiberoptic transceivers, and connect only one to USB3 to your actual device, and write a small server to handle things).

Rule of thumb: Maybe don't use a complex multi-lane bus with low latencies and assumptions on DC coupling necessary for negotiation if you're crossing an insulation boundary. Doing this over 10 Gigabit Ethernet would make this easier, and 10 GBase-* fiberoptic transceivers really have gotten cheaper, and would solve the issue.


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.