L
Admin
CW absolute-power habits (thermal vs diode, mismatch factor) break as soon as the DUT is pulsed radar, TDD, or a framed PA. A sensor that was “within 0.2 dB” on CW can lie by several dB on a burst if video bandwidth, gating, and crest factor are wrong. This is about **power meters / sensors on pulsed waveforms** — not spectrum-gated ACLR.
**1. Say which power you mean**
- **Average (true average over a long window):** what a thermal sensor naturally reports; what many link budgets and heat sinks care about.
- **Pulse average (on-slot average):** average only while RF is on — needs a gated meter or a sensor+video path fast enough, plus a defined gate.
- **Peak / envelope peak:** max of the detected envelope — depends on video BW and detector law; easy to invent with a too-slow or too-fast video path.
- Habit: write **which of the three**, duty cycle / gate, and video BW on every plot. “Pout = 40 dBm” with no qualifier is not a measurement.
**2. Sensor family vs pulse**
- **Thermal:** excellent average; **blind to peak** unless you know duty cycle exactly and the pulse is clean. Do not back-calculate peak from thermal + “I think duty is 10%” if the envelope has droop or an ALC spike.
- **Diode / envelope sensors:** can do peak and pulse-average **if** video BW and instrument mode match the pulse width. A diode sensor in CW-average mode on a pulsed DUT is a different (wrong) instrument.
- Habit: for short pulses, verify the sensor’s **minimum pulse width / rise** against your slot; if the pulse is narrower than the sensor spec, you are measuring the sensor, not the PA.
**3. Video bandwidth and crest-factor traps**
- Video BW too low → peak reads low, pulse edges smear, “peak” tracks something closer to average.
- Video BW very high + noisy envelope → peak detector grabs noise spikes; reported peak walks with attenuator and cable.
- Modulated bursts (OFDM, noise-like) have **PAPR**; a CW-calibrated diode correction may not hold. Prefer a sensor/path characterised for your waveform class, or cross-check with a calibrated average × known CCDF/PAPR from a trusted demod.
- Habit: after a peak reading looks good, **change video BW one step** and demand the number stay inside your uncertainty — if it jumps, you were on the edge of the sensor’s pulse response.
**4. Level, mismatch, and the burst crest**
- Burst peaks still hit the sensor diode and any upstream pad. Set the path for the **peak**, not the time-average — same spirit as mixer-level discipline on an analyser.
- Mismatch factor still applies; a pulsed DUT’s Γ often walks during the ramp. A good coupler + average of several pulses beats arguing about 0.3 dB on a single peak grab.
- External trigger / gate delay is part of the measurement — save it with the result.
**5. A short pulsed-power sanity checklist**
1. Name the quantity: average / pulse-average / peak.
2. Sensor mode + video BW + min pulse width vs actual slot.
3. Gate/trigger source logged (hardware TX-enable preferred over “guessed period”).
4. One video-BW step test; one attenuator step test — peaks should move as physics predicts, not randomly.
5. Cross-check: thermal long-average ≈ pulse-average × duty (for clean rectangular pulses); if not, fix duty/gate before you trust peak.
**Discussion**
On your last pulsed or TDD PA bring-up, which number lied first — thermal back-calculated “peak,” diode peak with wrong video BW, or an ungated average that mixed on and off slots — and do you gate the power meter from a hardware TX-enable or from the RF envelope?
— rfforum 编辑部 / Editorial Team
**1. Say which power you mean**
- **Average (true average over a long window):** what a thermal sensor naturally reports; what many link budgets and heat sinks care about.
- **Pulse average (on-slot average):** average only while RF is on — needs a gated meter or a sensor+video path fast enough, plus a defined gate.
- **Peak / envelope peak:** max of the detected envelope — depends on video BW and detector law; easy to invent with a too-slow or too-fast video path.
- Habit: write **which of the three**, duty cycle / gate, and video BW on every plot. “Pout = 40 dBm” with no qualifier is not a measurement.
**2. Sensor family vs pulse**
- **Thermal:** excellent average; **blind to peak** unless you know duty cycle exactly and the pulse is clean. Do not back-calculate peak from thermal + “I think duty is 10%” if the envelope has droop or an ALC spike.
- **Diode / envelope sensors:** can do peak and pulse-average **if** video BW and instrument mode match the pulse width. A diode sensor in CW-average mode on a pulsed DUT is a different (wrong) instrument.
- Habit: for short pulses, verify the sensor’s **minimum pulse width / rise** against your slot; if the pulse is narrower than the sensor spec, you are measuring the sensor, not the PA.
**3. Video bandwidth and crest-factor traps**
- Video BW too low → peak reads low, pulse edges smear, “peak” tracks something closer to average.
- Video BW very high + noisy envelope → peak detector grabs noise spikes; reported peak walks with attenuator and cable.
- Modulated bursts (OFDM, noise-like) have **PAPR**; a CW-calibrated diode correction may not hold. Prefer a sensor/path characterised for your waveform class, or cross-check with a calibrated average × known CCDF/PAPR from a trusted demod.
- Habit: after a peak reading looks good, **change video BW one step** and demand the number stay inside your uncertainty — if it jumps, you were on the edge of the sensor’s pulse response.
**4. Level, mismatch, and the burst crest**
- Burst peaks still hit the sensor diode and any upstream pad. Set the path for the **peak**, not the time-average — same spirit as mixer-level discipline on an analyser.
- Mismatch factor still applies; a pulsed DUT’s Γ often walks during the ramp. A good coupler + average of several pulses beats arguing about 0.3 dB on a single peak grab.
- External trigger / gate delay is part of the measurement — save it with the result.
**5. A short pulsed-power sanity checklist**
1. Name the quantity: average / pulse-average / peak.
2. Sensor mode + video BW + min pulse width vs actual slot.
3. Gate/trigger source logged (hardware TX-enable preferred over “guessed period”).
4. One video-BW step test; one attenuator step test — peaks should move as physics predicts, not randomly.
5. Cross-check: thermal long-average ≈ pulse-average × duty (for clean rectangular pulses); if not, fix duty/gate before you trust peak.
**Discussion**
On your last pulsed or TDD PA bring-up, which number lied first — thermal back-calculated “peak,” diode peak with wrong video BW, or an ungated average that mixed on and off slots — and do you gate the power meter from a hardware TX-enable or from the RF envelope?
— rfforum 编辑部 / Editorial Team
