O
Member
【Editorial Discussion】
Noise-power-ratio testing with a notched noise loading is a useful way to stress a wideband amplifier, but it is not automatically a substitute for the multi-carrier traffic waveform the product will actually carry. When NPR looks clean and a later multi-carrier or ACLR/EVM test fails, the gap is often in how the two tests were set up, not in a single “bad” number. Before you retune the PA, it is worth writing down which difference you have actually checked.
1) Align the occupied bandwidth, the notch, and the drive. NPR is sensitive to the noise bandwidth, the notch width and depth, and whether the average power in the loading is the same as the average power in the traffic waveform. A notch that is too narrow, a loading that does not fill the same band, or a drive set to a convenient lab level rather than the intended average can all make NPR look better than the later traffic test. Record the loading bandwidth, notch width, and the power at the same reference plane.
2) Crest factor and statistics are not the same. A well-behaved noise loading has a different peak distribution from a multi-carrier traffic waveform with a defined PAPR, a defined number of carriers, and often a defined idle or burst pattern. If the PA, a driver, or the instrument is set up for the noise-loading average and then clips on the traffic peaks, the traffic test will fail while NPR still looks clean. Compare CCDF or at least peak-to-average at the same plane, and confirm the instrument range was set for the peaks of both signals.
3) Thermal and bias state should match the intended duty. A continuous noise loading soaks the device differently from a bursty or unevenly loaded multi-carrier waveform, and the opposite can also be true. If you measure NPR after a long soak and then run traffic with a different duty, or the other way around, gain, phase, and memory terms can move. Repeat the two tests after the same warm-up and at the same case or flange temperature before you conclude the methods themselves disagree.
4) Do not treat a passing NPR as a clearance of the later demodulated metrics. NPR answers a notched-noise question; ACLR, EVM, or a carrier-specific mask answers a different one. If the two disagree, write down which term you think moved — notch and drive alignment, peak statistics, or thermal state — before changing the matching or the DPD.
Open question: the last time NPR with a notched noise loading looked clean and a multi-carrier traffic waveform failed, what actually closed the gap — the notch and drive alignment, peak statistics, or the thermal / duty state? Please share the kind of PA and which term you think was dominant; no customer names or unverified numbers needed.
Noise-power-ratio testing with a notched noise loading is a useful way to stress a wideband amplifier, but it is not automatically a substitute for the multi-carrier traffic waveform the product will actually carry. When NPR looks clean and a later multi-carrier or ACLR/EVM test fails, the gap is often in how the two tests were set up, not in a single “bad” number. Before you retune the PA, it is worth writing down which difference you have actually checked.
1) Align the occupied bandwidth, the notch, and the drive. NPR is sensitive to the noise bandwidth, the notch width and depth, and whether the average power in the loading is the same as the average power in the traffic waveform. A notch that is too narrow, a loading that does not fill the same band, or a drive set to a convenient lab level rather than the intended average can all make NPR look better than the later traffic test. Record the loading bandwidth, notch width, and the power at the same reference plane.
2) Crest factor and statistics are not the same. A well-behaved noise loading has a different peak distribution from a multi-carrier traffic waveform with a defined PAPR, a defined number of carriers, and often a defined idle or burst pattern. If the PA, a driver, or the instrument is set up for the noise-loading average and then clips on the traffic peaks, the traffic test will fail while NPR still looks clean. Compare CCDF or at least peak-to-average at the same plane, and confirm the instrument range was set for the peaks of both signals.
3) Thermal and bias state should match the intended duty. A continuous noise loading soaks the device differently from a bursty or unevenly loaded multi-carrier waveform, and the opposite can also be true. If you measure NPR after a long soak and then run traffic with a different duty, or the other way around, gain, phase, and memory terms can move. Repeat the two tests after the same warm-up and at the same case or flange temperature before you conclude the methods themselves disagree.
4) Do not treat a passing NPR as a clearance of the later demodulated metrics. NPR answers a notched-noise question; ACLR, EVM, or a carrier-specific mask answers a different one. If the two disagree, write down which term you think moved — notch and drive alignment, peak statistics, or thermal state — before changing the matching or the DPD.
Open question: the last time NPR with a notched noise loading looked clean and a multi-carrier traffic waveform failed, what actually closed the gap — the notch and drive alignment, peak statistics, or the thermal / duty state? Please share the kind of PA and which term you think was dominant; no customer names or unverified numbers needed.
