L
Admin
[Technical article, rfforum Editorial Team]
Class-F (and continuous-mode cousins) live or die on harmonic terminations. On FR4 or mid-loss laminates, the open/short you drew at 2f0 and 3f0 can walk several tens of degrees once Er, height, and etch tolerance show up — enough to erase the efficiency you sold in the proposal.
1. Ideal vs board
Textbook Class-F wants a short (or very low Z) at even harmonics and an open at odd harmonics at the device reference plane — not at the SMA. Every millimetre of package bondwire, landing, and microstrip is part of the harmonic network.
2. What FR4 does to you
Er and loss tangent vary by lot and by moisture. A stub that was "open at 3f0" in the nominal stackup can land closer to 50 Ω reactive after a 10% Er shift. High-Q harmonic traps are the first to detune.
3. Practical hardening
- Prefer slightly lower-Q harmonic networks and verify across corners (Er ±, etch ±, temp)
- Put the most sensitive harmonic condition closest to the device with controlled geometry, not a long thin stub at the far end of the board
- Load-pull or at least harmonic-tuned source/load at 2f0/3f0 on a few corner boards before freezing the Gerber
- Do not trust CW PAE alone; check with the real crest-factor waveform
4. Habit
Treat harmonic termination as a tolerance stackup problem, not a one-shot Smith-chart art project.
Discussion: for your last sub-6 Class-F or Class-F−1 bring-up on FR4/mid-loss laminate, which harmonic (2f0 or 3f0) moved first when the board house changed, and did you widen the network or change laminate?
— rfforum 编辑部 / Editorial Team
Class-F (and continuous-mode cousins) live or die on harmonic terminations. On FR4 or mid-loss laminates, the open/short you drew at 2f0 and 3f0 can walk several tens of degrees once Er, height, and etch tolerance show up — enough to erase the efficiency you sold in the proposal.
1. Ideal vs board
Textbook Class-F wants a short (or very low Z) at even harmonics and an open at odd harmonics at the device reference plane — not at the SMA. Every millimetre of package bondwire, landing, and microstrip is part of the harmonic network.
2. What FR4 does to you
Er and loss tangent vary by lot and by moisture. A stub that was "open at 3f0" in the nominal stackup can land closer to 50 Ω reactive after a 10% Er shift. High-Q harmonic traps are the first to detune.
3. Practical hardening
- Prefer slightly lower-Q harmonic networks and verify across corners (Er ±, etch ±, temp)
- Put the most sensitive harmonic condition closest to the device with controlled geometry, not a long thin stub at the far end of the board
- Load-pull or at least harmonic-tuned source/load at 2f0/3f0 on a few corner boards before freezing the Gerber
- Do not trust CW PAE alone; check with the real crest-factor waveform
4. Habit
Treat harmonic termination as a tolerance stackup problem, not a one-shot Smith-chart art project.
Discussion: for your last sub-6 Class-F or Class-F−1 bring-up on FR4/mid-loss laminate, which harmonic (2f0 or 3f0) moved first when the board house changed, and did you widen the network or change laminate?
— rfforum 编辑部 / Editorial Team
