How Do You Design a Galvanically Isolated USB Interface?

Detailed Explanation

Adding galvanic isolation to a USB interface is a different design problem from isolating a slower interface like UART, SPI, or I2C — for those, a general-purpose digital isolator or optocoupler can often sit directly in the signal path. USB Full Speed's 12 Mbit/s differential signaling has tight eye-diagram, timing, and impedance requirements that passive or general-purpose isolation components cannot meet, so isolated USB interfaces are built around a dedicated USB isolator IC instead.

Why USB Needs a Purpose-Built Isolator

A USB isolator IC does not simply pass the D+/D− differential signal through an isolation barrier the way a digital isolator passes a UART TX/RX line through. Instead, it terminates the USB link electrically on each side, decodes and re-times the signal, transmits it across the isolation barrier (typically using the same capacitive or inductive coupling technology as the vendor's general-purpose digital isolators), and re-drives a clean, correctly-terminated USB signal on the far side. This architecture is why isolator vendors publish dedicated USB isolator parts — Analog Devices' ADuM4160, Texas Instruments' ISOUSB211, and Skyworks/Silicon Labs' Si86xx USB isolator variants are commonly used examples — rather than expecting a designer to build isolated USB from a general-purpose digital isolator and discrete signal conditioning.

Isolator Placement and the Two Grounds

The isolator IC sits directly between the host-side USB connector (or host-side USB PHY, if the device is not simply a passthrough) and the isolated-side USB PHY or connector, with the isolation barrier running straight through the middle of the D+/D− differential pair. This creates two electrically separate ground domains — a host-side ground referenced to the USB host and PC chassis, and an isolated-side ground that must have no conductive path to the host-side ground anywhere on the board, including through shared shielding, mounting hardware, or a stray copper pour that bridges the barrier. PCB layout for the isolation barrier follows the same creepage and clearance principles as any other isolated interface — see PCB creepage and clearance distances for how the required physical spacing across the barrier is determined, and note that a medical device's patient-connected isolation requirements under IEC 60601-1 are typically more stringent than a general industrial isolation barrier.

Powering the Isolated Side

Isolating the data lines only solves half the problem — the isolated-side circuitry (the isolator IC itself, plus whatever USB device or host circuitry sits beyond it) still needs power, and that power cannot come from the host-side USB VBUS through a direct conductive connection without defeating the isolation. The standard solution is an isolated DC-DC converter: either a small transformer-based isolated regulator module, or a discrete flyback/push-pull isolated supply, deriving isolated-side power from host-side VBUS (or a separate isolated-side source) with no conductive path crossing the barrier. Isolator IC datasheets specify the part's own supply current on each side; the isolated DC-DC supply must be sized for that current plus whatever the isolated-side USB device circuitry itself draws, with margin for the worst-case combined load.

Full Speed vs High Speed Isolation

Nearly all isolated USB designs for embedded products operate at USB Full Speed (12 Mbit/s) or Low Speed (1.5 Mbit/s) — the speed tiers the common isolator ICs (ADuM4160, ISOUSB211, Si86xx) support. Isolating USB High Speed (480 Mbit/s) is a substantially harder signal-integrity problem — the isolator itself becomes a much more demanding, and more expensive, design — and is rarely justified for typical embedded use cases like CDC-ACM virtual serial ports, HID interfaces, or firmware-update (DFU) links, all of which run comfortably at Full Speed. See USB 2.0 vs USB 3.x vs USB4 for the broader speed-tier decision; if a product's actual data-rate requirement pushes toward High Speed or USB 3.x, isolating the interface may not be practical, and a different isolation strategy (isolating a lower-bandwidth control/status link instead, or isolating power only) is worth considering.

When Isolation Is Actually Required

The most common driver for isolated USB is a medical device with a patient-connected circuit on one side of the interface and a PC or USB hub on the other — IEC 60601-1 defines the applied-part isolation requirements that make galvanic isolation mandatory in this case, and the required isolation voltage rating and creepage/clearance distances depend on the specific applied-part classification. See Electrical Safety: AS/NZS 3820 and IEC 62368-1 for the general hazard-based safety framework this sits within. Industrial USB isolation is typically driven by a different concern — protecting a PC-facing USB port from high-voltage transients, ground loops, or electrical noise on plant-side equipment — where the isolation voltage requirement is usually lower than a medical applied part but still a deliberate design decision, not a default.

Design Considerations

• Check the isolator's Full Speed vs Low Speed support against the actual USB device class needed. Not every USB isolator IC supports both speed tiers, and a device class that requires Full Speed (most CDC-ACM and HID implementations) will not enumerate correctly through a Low-Speed-only isolator.
• Size the isolated-side DC-DC supply for the isolator's own current draw plus the isolated-side device load, with real margin. Under-sizing this supply is a common source of intermittent enumeration failures that look like a signal-integrity problem but are actually a brownout on the isolated side.
• Treat the isolation barrier's creepage and clearance as a hard PCB layout constraint, not a routing afterthought. For medical applied-part isolation under IEC 60601-1, the required distances are considerably larger than typical low-voltage PCB spacing and must be planned into the board outline from the first layout pass. Zeus Design designs isolated interface PCBs, including medical and industrial USB isolation, from component selection through to layout and compliance-ready documentation.
• Confirm the isolator's data rate is sufficient before committing to the architecture — if the product genuinely needs USB High Speed or USB 3.x bandwidth across the isolation barrier, isolated USB may not be the right approach at all, and an alternative such as isolating a lower-bandwidth companion link or isolating only the power path is worth evaluating early.

Common Mistakes

• Attempting to isolate USB data lines with a general-purpose digital isolator or optocoupler. These parts are not designed for USB's Full Speed timing and eye-diagram requirements and will not produce a reliably enumerating link.
• Under-sizing the isolated-side power supply, producing intermittent enumeration failures, USB reset loops, or device brownouts that are easy to misdiagnose as a signal-integrity or firmware problem rather than a power budget shortfall.
• Allowing a stray conductive path to bridge the isolation barrier — a shared shield, a mounting screw through both ground planes, or a copper pour that crosses the barrier gap — silently defeating the isolation without any obvious symptom until an insulation or dielectric-withstand test fails.
• Under-speccing creepage and clearance for the actual isolation voltage and safety classification required, particularly for medical applied parts under IEC 60601-1, where the necessary spacing is substantially larger than general low-voltage PCB practice.

For general isolation technology selection outside of USB specifically, see optocoupler vs digital isolator.