A Systematic Approach

When an RF switch underperforms, the symptoms are usually excess insertion loss or poor isolation, and both are more often caused by the board than by the part. This article presents a systematic procedure for diagnosing Skyworks RF switch designs, using the SKY13373-460LF and SKY13321-360LF as references, so that the cause is found by measurement rather than by guessing.

Step 1: Measure the Switch Alone

Before changing anything, measure the switch in isolation on the board with a network analyser. Terminate the unused ports in 50 ohms, keep the cables fixed, and record the insertion loss and isolation across the band. Compare the result with the data sheet at the same frequency. If the switch alone meets its specification, the problem is elsewhere in the path, and if it does not, the layout around the switch is the place to look.

Controlling the Measurement

RF measurements are easy to get wrong. Calibrate the analyser at the board plane, not at the instrument, so the cable and connector losses are removed. Keep the cables still between measurements, because a moved cable changes the result. Terminate unused ports properly, because an open or unmatched port reflects energy and corrupts the isolation reading.

Step 2: Check the RF Layout

Insertion loss is dominated by the layout of the path. Confirm that the trace is a controlled impedance, that it is short and free of stubs, and that the ground reference is continuous under it. Lossy or high-dielectric material near the path adds loss, and a split in the ground return under the trace raises both loss and radiation. Compare the board against the reference layout, which was designed to reach the data-sheet numbers.

Ground and Vias

The switch's ground pad must connect to a solid ground plane with multiple vias placed close to the pad. Insufficient grounding raises loss and degrades isolation, and it is one of the most common layout mistakes. Check the via count and placement against the reference layout.

Step 3: Verify the Control

A switch that is not fully switched looks like a switch with poor isolation, because the off path is not fully off. Verify that the control voltages reach the levels the data sheet requires, that the control lines are decoupled, and that the timing allows the switch to settle before the radio transmits or receives. Measure the control voltage at the device pin, not at the controller, because trace resistance and loading can reduce it.

Step 4: Confirm the Isolation at Frequency

Isolation falls with frequency as coupling through the board and package grows, so a design that meets the specification at 2.5 GHz may not at 6 GHz. Measure the isolation at the top of your band, and if it is short, increase the physical separation between the paths, keep them symmetric, and add grounding between them. In a multi-antenna design, the antenna isolation also contributes to the total, so measure the complete system, not just the switch.

When to Escalate

If the switch still underperforms after the layout, ground and control are corrected, escalate with the measured data: the switch-alone result, the control voltages and the isolation across the band. BeiLuo's FAE team can review the layout against the reference design and, if needed, characterize the switch in our RF lab so the cause is identified before a board respin.