EMC Troubleshooting Part 2 — Narrowband Footprints & Harmonics

Scope: How to read narrowband features on the RE plot (100–1000 MHz is where they often pop up) and trace them back to clocks, harmonics, and reflections. Part 3 will cover semi‑narrowband and broadband cases.

1) Two spectral families to watch

• Narrowband rows: tall, thin lines at discrete frequencies. Typically clocks, their harmonics, spurs, or resonances tightly tied to a periodic source. This usually are both common-mode energy or differential (e.g. bad layout).

• Broadband humps: wide "bell" shapes (often common‑mode energy). We covered these in Part 1.

In this part we focus on the narrowband rows.

2) Make a quick harmonic map

Before you start: have on a spreadsheet the list of all your clocks.

When you get the lab’s RE plot, list the strongest narrow lines (e.g., 150 MHz, 200 MHz, …). Try to relate them to a known fundamental.

Example: If your design contains a 50 MHz clock:

• 3rd harmonic → 150 MHz

• 4th harmonic → 200 MHz

• 5th harmonic → 250 MHz

A neat table helps you see patterns immediately (odd‑only vs. mixed odd/even, missing orders, etc.).

Example of what kind of fottprint we are looking for

3) Odd vs. even harmonics — what they mean

A perfectly symmetric, 50%‑duty ideal square wave contains odd harmonics only:

x(t)=∑k=0∞4π(2k+1)sin⁡(2π(2k+1)f0t

Real life breaks the ideal in two main ways:

A) Dominantly odd harmonics

Likely cause: edges are fast, periodic source is clean and symmetric, good matching impedance. The spectrum rolls off ~1/n but stays rich in odd orders.

What helps:

• Edge‑rate control: add a small series resistor at the source (preferred)

• tiny C  to ground (between the driver and the above resistor)( Both reduce di/dt and trim high‑order content.

  
  

⚠️ Caution with shunt capacitors (e.g., 4.7–12 pF near the clock source): they low‑pass the edge and can increase jitter, distort duty cycle, and stress the driver. Use only when timing margins are ample!

B) Even harmonics become strong

Likely cause: 

1. reflections/ringing,

2. duty‑cycle distortion (≠ 50%)

3. asymmetric rise/fall.

Any of these inject even orders.

What helps (reflection‑driven cases):

• Source series termination placed as close as possible to the driver.

• Receiver end damping only when you can....

• Tame ringing  the above resistor helps reducing the ringing as well.

• 
  

4) Fast diagnostic flow at the lab

1. List narrow lines (e.g., 150, 200, 250 MHz…).

2. Guess the fundamental (here 50 MHz) and tag each as odd/even.

3. If odd‑heavy → try edge‑rate control first (add small R at source; optionally tiny C). Re‑scan and log dB change per harmonic.

4. If even present/strong → check for ringing on the scope at the source/receiver. Add source series R, and verify duty cycle at the receiver. Re‑scan.

   
   

5) Practical component values & checks

• Series R (clock/data): start 10–220 Ω; tune to flatten overshoot/undershoot without violating rise‑time budget.

• Shunt C (only if needed): 2.2–12 pF NP0/C0G at the source; re‑check timing and duty cycle. Prefer driver‑integrated slew control if available. I usually leave space for an 0603 or 0402

• Anti-Snubber (C–R) (Not R-C) at source: begin around R = 10–47 Ω, C = 2–18 pF;

• Measurement sanity: use a low‑inductance probe (spring ground) when checking edges; wrong probing can create phantom ringing.

  
  

6) Example: 50 MHz clock shows 150 MHz & 200 MHz

• 150 MHz (odd) lines up with the 3rd harmonic → edge‑rate driven.

• 200 MHz (even) indicates asymmetry/reflections.

Trial fixes:

• Add 100Ω at the source; re‑measure RE → even harmonic typically drops with ringing.

• If odd still high, try small C (4.7 pF) at source or increase series R slightly. Confirm that setup/hold are still met and slew is within spec at the receiver.

  
  

7) What to log (so you can prove causality)

• Harmonic table (orders, odd/even, amplitudes).

• Oscilloscope shots at source & receiver (rise/fall, overshoot %, duty cycle).

• Termination values tried and resulting dB change per harmonic.

• Any cable lengths and enclosure states used in the scan.

• 
  

Coming next — Part 3

Semi‑narrowband vs. broadband in the 150–500 MHz region: SMPS interaction with cable resonances, PoE specifics (magnetics balance, CM chokes), and how to tell when the board vs. the cable is in charge.