Don't Route Over Ground Plane Splits — Part 2: Immunity

Flipping the experiment

In Part 1, we measured how much a split ground plane radiates. The signal source fed the board, and the TEM cell picked up the emissions. The split board was 17 dB worse on average, with peaks over 40 dB worse at higher frequencies.

Now we reverse the signal path. The comb generator or pulser feeds the TEM cell directly, flooding the board with RF. The spectrum analyzer connects to the board's SMA port through the shielded tent feedthrough. This tells us how susceptible each board is to external RF.

A board that radiates more will often also be easier to excite, because the same broken return path hurts both directions. I would treat this as a qualitative expectation here, not a promise that the dB numbers must match.

Setup

Same equipment as Part 1:

• Siglent SSA3032X-R spectrum analyzer — now connected to the board's SMA port (the receiver)
• Tekbox TBTC3 TEM cell — now fed by the signal source (the transmitter)
• Tekbox TBST-120/60/60/2-B shielded tent
• Tekbox TBCG4 comb generator The test board sits on the TEM cell septum. One SMA port connects to the spectrum analyzer, the other is terminated with 50 Ω. The comb generator feeds the TEM cell's N-type connector through the tent feedthrough. Tent closed for all measurements.

As noted in Part 1: all measurements were taken with the board in a single, fixed orientation on the septum. DUT placement matters — in a real pre-compliance test, rotate the board and take the worst case.

SA settings for all measurements: 30 MHz to 1 GHz span, RBW 120 kHz, VBW 300 kHz, positive peak detector, reference level 57 dBuV, 0 dB input attenuation, preamp ON, 10 dB/div.

Results

The three raw spectrum captures were taken at different SA settings — the split board picked up so much more RF that the reference level and attenuator had to be changed just to fit the trace on screen. The plot below overlays all three boards on the same axes. Hover for readout values at any frequency.

Same picture as the emissions test, but the deltas are even larger. On average, the split board picks up 27 dB more than the intact board. At the worst frequency points, the delta exceeds 50 dB. Stitching caps recover about 26 dB on average, but their effectiveness drops above ~500 MHz.

Comparison with Part 1

The immunity results point in the same direction as the emissions results from Part 1:

The immunity test shows an even larger delta than the emissions test in this setup. I would not over-interpret that as a strict transmitter/receiver ranking, but the practical conclusion is clear: the split hurts both emission and susceptibility, and the stitching caps help in both tests.