Counterintuitive EMC Fix: Pass around 200 - 230 MHz by raising the clock
- Francesco Poderico
- Oct 2
- 3 min read
What I’m trying to show you today is counterintuitive.If you’re failing radiated emissions around 200–230 MHz by just 3–4 dB, there’s a simple way to flip that fail → pass: increase your clock frequency.Yes—increase it.
Before we get there, please grasp one key point about the limits we’re testing against:
For EN 55011 Class B, Group 1 at 3 m (small-size EUTs), the QP limit is 40 dBµV/m from 30–230 MHz, and 47 dBµV/m from 230–1000 MHz.There’s a +7 dB jump right at 230 MHz. If your offending harmonic sits just below 230 MHz, nudging it just above 230 MHz instantly buys you up to 7 dB of margin.
First: prove the cable is the radiator (2-minute sanity check)
If your EUT has cables, they’re very efficient antennas in roughly the 100–250 MHz band. So when you see a peak in that region, the first thing to check is whether a cable is exporting your noise.
Clip a ferrite onto the suspect cable.
If the peak around 200–230 MHz drops, you’ve proved the cable is acting as the antenna.
If your customer doesn’t want a ferrite on that cable—fine. We’ll fix it another way.
Example: failing at 225 MHz (about +3 dB over)
You don’t have a 225 MHz clock, so where does it come from? Make a quick list of possible fundamentals by dividing the failing frequency by small integers (harmonic orders):
225/9 = 25.0 MHz (9th harmonic)
225/6 = 37.5 MHz (8th harmonic)
225/5 = 45.0 MHz (7th harmonic)
225/4 = 56.25 MHz (6th harmonic)
225/3 = 75.0 MHz (5th harmonic)
…and so on.
You notice the board does have a 25 MHz clock. Bingo—225 MHz is likely its 9th harmonic.
Why is the 9th showing up?
An ideal 50 % square wave has strong odd harmonics (3rd, 5th, 7th, 9th…).
A long(-ish) 25 MHz route, a layer change without nearby ground vias, or other discontinuities can happily launch that harmonic.
As a feel-good check: the quarter-wave at 225 MHz is about 33 cm; PCB traces can resonate at fractions of that (e.g., 1/8), so 15–20 cm of effective path or a messy reference change can be enough to light it up—then the cable takes it and radiates.
The counterintuitive move: raise the clock so the harmonic crosses 230 MHz
We want the same harmonic order to land just above 230 MHz.
Here the order is n = 9 (225/25).
The minimum new fundamental to push the 9th past 230 MHz is:
f0≥230 MHz9≈25.56 MHz f0 =230/9 ≈25.56 MHz
Add a little guard band (e.g., aim ≥ 26.0 MHz).
Two simple ways:
If allowed, swap 25 MHz → 26 MHz XO/clock. The 9th moves to 243 MHz.The cable “antenna gain” is pretty similar, but the limit jumped by 7 dB—that alone often flips fail → pass.
If you have a PLL/DFS, nudge the effective base (e.g., 26 MHz → 9th at 234 MHz).
That’s it. You’ve changed nothing on the cable, nothing on the board layout—just pushed the same harmonic into a friendlier part of the standard.
Important cautions (read these)
Don’t break interfaces. Some peripherals are picky (e.g., Ethernet 25 MHz ref ±ppm, USB 48 MHz, audio rates). If a global XO change isn’t possible, see if you can retime only the noisy domain or use a PLL tweak.
Re-scan everything. After shifting one harmonic, make sure you didn’t accidentally create a new trouble spot elsewhere.
Confirm the source. A quick near-field probe pass over the 25 MHz route and any layer jumps will usually sing back at 225 MHz if it’s your culprit.
Fast lab checklist (5 steps)
Ferrite on cable → did the 225 MHz peak drop? (Radiator proven.)
Identify harmonic → failing frequency ÷ n; match to real clocks (here, 25 MHz, 9th).
Probe the board → near-field around the 25 MHz path, especially at layer changes without ground stitching.
Nudge frequency → PLL tweak to ≥ 25.6 MHz, or swap to 27 MHz (if allowed). Target harmonic > 232–235 MHz for comfort.
Re-scan (QP) → confirm margin; check the rest of the band for surprises.
If you can’t change the clock: try source series R on the driver, a touch of slew control, and make the return under that 25 MHz net continuous (add ground vias near the transitions).
Comments