L
Admin
[Technical article / FAQ, rfforum Editorial Team]
CW or long-burst ACLR can pass while the first symbols after a long idle fail. GaN PA thermal and trapping memory shift bias and gain on a millisecond-to-second scale; a pulsed ACLR test that only windows mid-burst never sees the idle-to-burst transient that real TDD or burst traffic hits.
1. What memory owns after a pulse
- Self-heating and trap states from the previous burst (or long OFF) move gain, phase, and ACLR on the next ON edge
- A PA that meets ACLR after 10 ms of ON can fail in the first 100–500 µs; CW acceptance never exercises that edge
- Case temperature and supply sag under the real duty cycle change the settling curve; a desk CW soak underestimates both
2. Freeze list before you sign pulsed / TDD ACLR
- Idle time and ON duration used in the pulse train (not only "pulsed")
- Whether the analyzer gate included the leading edge or only mid-burst symbols
- Case temperature and supply under the same duty cycle as the claim
- Bias / sequencing table — memory often looks like "PA ACLR" when bias recovery is late
3. Habit
Never gate a bursty GaN PA release on CW ACLR alone. Archive an idle-to-burst gated plot next to every "ACLR OK" traveler. If DPD is in the path, also note whether coefficients were frozen cold or after the same idle/ON profile — memory + stale DPD is a common false "PA bad" call.
4. Practical triage (10 minutes)
- Gate leading-edge vs mid-burst ACLR on the same pulse train; if only the edge fails, chase thermal/trap/bias recovery before buying another PA
- Lengthen idle by 10×; if ACLR worsens only after long OFF, document idle time as a release gate
- Repeat at product case temperature; a desk-ambient pass that fails at hot soak is a memory/bias problem dressed as ACLR
Discussion: on your last pulsed ACLR surprise, did the fail sit in the leading edge, mid-burst, or after a long OFF — and do you require an idle-time note on every pulsed ACLR sheet?
— rfforum 编辑部 / Editorial Team
CW or long-burst ACLR can pass while the first symbols after a long idle fail. GaN PA thermal and trapping memory shift bias and gain on a millisecond-to-second scale; a pulsed ACLR test that only windows mid-burst never sees the idle-to-burst transient that real TDD or burst traffic hits.
1. What memory owns after a pulse
- Self-heating and trap states from the previous burst (or long OFF) move gain, phase, and ACLR on the next ON edge
- A PA that meets ACLR after 10 ms of ON can fail in the first 100–500 µs; CW acceptance never exercises that edge
- Case temperature and supply sag under the real duty cycle change the settling curve; a desk CW soak underestimates both
2. Freeze list before you sign pulsed / TDD ACLR
- Idle time and ON duration used in the pulse train (not only "pulsed")
- Whether the analyzer gate included the leading edge or only mid-burst symbols
- Case temperature and supply under the same duty cycle as the claim
- Bias / sequencing table — memory often looks like "PA ACLR" when bias recovery is late
3. Habit
Never gate a bursty GaN PA release on CW ACLR alone. Archive an idle-to-burst gated plot next to every "ACLR OK" traveler. If DPD is in the path, also note whether coefficients were frozen cold or after the same idle/ON profile — memory + stale DPD is a common false "PA bad" call.
4. Practical triage (10 minutes)
- Gate leading-edge vs mid-burst ACLR on the same pulse train; if only the edge fails, chase thermal/trap/bias recovery before buying another PA
- Lengthen idle by 10×; if ACLR worsens only after long OFF, document idle time as a release gate
- Repeat at product case temperature; a desk-ambient pass that fails at hot soak is a memory/bias problem dressed as ACLR
Discussion: on your last pulsed ACLR surprise, did the fail sit in the leading edge, mid-burst, or after a long OFF — and do you require an idle-time note on every pulsed ACLR sheet?
— rfforum 编辑部 / Editorial Team
