USB advantures: galvanic isolation, over fiber-optics, 3.0-to-2.0 translation, and more


This article describes several experiments with USB 2.0 and USB 3.0 that I performed from 2020 to 2022, including three possible methods of high-speed 480 Mbps galvanic isolation, and additional discussions of USB 3.0 over fiber-optics, USB 3.0 to USB 2.0 transaction translation, USB over twisted pairs, and more.


In early 2020s, I was trying to listening to the shortwave radio (HF) with an cheap RTL-SDR receiver, a random-wire antenna, and a homebrew frequency mixer front-end to convert HF band to VHF band that the RTL-SDR was designed to receive. Unfortunately, it only worked on my laptop. If a Raspberry Pi is used instead, the receiver would be immediately plagued by a ton of high-frequency noise that even caused front-end overloading. Various attempts to solve this problem had failed: shielding the Raspberry Pi, filtering the power supply, adding common-mode chokes to the coax, USB, or Ethernet cable, powering the system using a linear regulator, or adding an isolation transformer, all had no effect.

Eventually, I determined that the origin of interference is common-mode noise - I could reproduce the same deafening noise by plugging the laptop into the AC mains. Basically, the ground reference of the antenna and the ground reference of the mains are at different electrical potentials, connecting them together causes a miniscule radio-frequency current to flow and corrupts the ground reference. The problem is likely also exacerbated by the use of a random-wire antenna, which has no well-defined ground by itself - its “ground” reference is defined by whatever external circuits that happens to be connected the antenna, and even random metal objects around the antenna.

I managed to partially work around the problem by isolating the antenna input (including ground) with an RF balun transformer. By the virture of transformer action, common-mode noise is suppressed, allowing only differential-mode signal to pass. However, a few days later I started experimenting a different circuit arrangement of the RF front-end, and the noise problem came back.

Galvanic Isolation

At this point, I started looking for a more systematic solution instead - is it possible to create a separate power domain, that is, an isolated power supply with its own power and ground planes, with no direct connection to the external world, allowing the radio to work in its own quiet island with disturbance?

Technically, this is called galvanic isolation - using transformer, optics, radio, acoustics, and other indirect means of signaling and energy transfer, it’s possible to allow signaling between different subcircuits without anelectric current flow, or creating an isolated power supply without a direct electrically conductive path.

It has wide applications in industrial systems.

First, galvanic isolation protects the computer, control systems, and even human operators from destruction by high voltage transients or faults as the dangerous high-voltage subcircuit is safely quarantined. The most common application is switched-mode power supplies for phone chargers, computers, and other gadgets. Most designs use transistors to switching the AC mains voltage at the primary side, stepping it down via an isolation transformer (only high-frequency AC current can pass such a transformer, DC and 50/60 Hz AC current are blocked), and creating a safe low voltage at the secondary side. To provide a feedback signal to the primary side without compromising isolation, an additional optocoupler is used to transmit the control signal as a beam of light.

In additional to safety, galvanic isolation can also eliminates unwanted conductive electromagnetic interference between different devices. The most common example is a “ground loop” in audio and video systems, causing the notorious 50/60 Hz hum in professional equipment. Or in my example, deafening, wideband noise in a radio receiver (although the actual improvement can be limited, two insulated metal plate is a capacitor, and when the frequency is high enough, even picofarards of capacitance looks like a short circuit to the noise).

Finally, galvanic isolation also enables a “floating” circuit. Galvanic isolation allows engineers to make a floating measurement without compromising safety. The signal ground of an oscilloscope input is usually connected to AC mains ground for safety, and the device-under-test (such as a USB gadget) is often also referenced to ground. In this case, one cannot make a measurement between two arbitrary points in the circuit - connecting the “ground” of the oscilloscope probe to a voltage is effectively a short circuit. To overcome this problem, engineers often “cheat” by disconnecting the protective ground to “float the scope”, but it creates a safety hazard and makes the oscilloscope unsafe - a fault within the oscilloscope can energize its entire chassis. The best practice is to “float” the device-under-test instead of the scope itself, which is accomplished by galvanic isolation.

USB 2.0 High-Speed Challenges

Unfortunately, although

there was no ASIC for High-Speed (480 Mbps) isolation on the open market.

In fact, USB 2.0 Low-Speed (1.5 Mbps) and Full-Speed (12 Mbps) isolators

However, RTL-SDR is a USB 2.0 high-speed (480 Mbps) device