L
Admin
[Technical article / FAQ, rfforum Editorial Team]
Harmonic filters are often drawn as a clean block after the PA. On the bench, distance, interconnect VSWR at 2f0/3f0, and bias-network resonances decide whether the PA still sees the load-pull contour you designed for — or a reflective wall that collapses efficiency and raises memory effects.
1. What placement actually changes
- A filter with good rejection but poor return loss at 2f0 reflects harmonic power back into the PA; efficiency and ACLR can move even when fundamental match looks fine
- Cable / microstrip length between PA and filter rotates the harmonic load; a few cm at sub-6 GHz is already a meaningful fraction of a wavelength at 2f0
- Bias networks and decoupling that look quiet at f0 can resonate near 2f0 and turn "filter VSWR" into a bias-path problem
2. Freeze list for a PA + filter claim
- Measured Γ at 2f0 and 3f0 at the PA reference plane with the real filter and interconnect, not only the filter datasheet
- Physical length / stack between PA drain/collector plane and filter input, on the same sheet as drain efficiency and ACLR
- Whether DPD / memory metrics were taken with that exact interconnect — swapping a "temporary" cable is a method change
- Thermal note: a reflective harmonic load that heats the device can look like a bias or DPD bug
3. Habit
Never freeze "PA efficiency with harmonic filter" without a harmonic load-pull or at least a 2f0 Γ measurement at the PA plane. Rejection alone is not a load-line guarantee.
Discussion: on your last efficiency drop after adding a harmonic filter, was the root cause 2f0 Γ, interconnect length, or a bias resonance — and do you require a harmonic Γ note before accepting the filter placement?
— rfforum 编辑部 / Editorial Team
Harmonic filters are often drawn as a clean block after the PA. On the bench, distance, interconnect VSWR at 2f0/3f0, and bias-network resonances decide whether the PA still sees the load-pull contour you designed for — or a reflective wall that collapses efficiency and raises memory effects.
1. What placement actually changes
- A filter with good rejection but poor return loss at 2f0 reflects harmonic power back into the PA; efficiency and ACLR can move even when fundamental match looks fine
- Cable / microstrip length between PA and filter rotates the harmonic load; a few cm at sub-6 GHz is already a meaningful fraction of a wavelength at 2f0
- Bias networks and decoupling that look quiet at f0 can resonate near 2f0 and turn "filter VSWR" into a bias-path problem
2. Freeze list for a PA + filter claim
- Measured Γ at 2f0 and 3f0 at the PA reference plane with the real filter and interconnect, not only the filter datasheet
- Physical length / stack between PA drain/collector plane and filter input, on the same sheet as drain efficiency and ACLR
- Whether DPD / memory metrics were taken with that exact interconnect — swapping a "temporary" cable is a method change
- Thermal note: a reflective harmonic load that heats the device can look like a bias or DPD bug
3. Habit
Never freeze "PA efficiency with harmonic filter" without a harmonic load-pull or at least a 2f0 Γ measurement at the PA plane. Rejection alone is not a load-line guarantee.
Discussion: on your last efficiency drop after adding a harmonic filter, was the root cause 2f0 Γ, interconnect length, or a bias resonance — and do you require a harmonic Γ note before accepting the filter placement?
— rfforum 编辑部 / Editorial Team
