DPD for GaN Doherty PAs - model complexity vs linearization?

Started by CSharp_Dev_Tony · 2026-11-12 03:36 UTC · 38 views · 4 replies
C Member
Doherty PAs are efficient but nasty to linearize. How complex does your DPD model need to be for 100 MHz 5G carriers?
L Admin
GaN Doherty 的 DPD,编辑部更认「记忆深度跟着包络带宽与热时间常数走」,而不是模型阶数越高越好。常见拆法:

1) 复杂度: 100 MHz 级 5G 载波通常要记忆多项式/GMP 一类,阶数与记忆深度按 ACLR/EVM 边际加,而不是一次拉满。Doherty 的效率峰与回退区非线性形状不同,分段或双输入模型有时比硬堆阶数更稳。
2) 训练: 盯 PAPR、预失真饱和与观测接收机失真;热漂移要用慢自适应或温度表。只训 CW/窄带再上宽带,经常 ACLR 假好看。
3) 验收: 报载波配置、PAPR、模型结构(阶/记忆)、迭代次数,以及 ACLR/EVM/功耗三点一起。

欢迎贴 GaN 漏压、瞬时带宽、你们的模型族和最终 ACLR。有「加深记忆前后」对比特别有价值。

— rfforum 编辑部 / Editorial Team
R Member
For a 100 MHz 5G carrier on a GaN Doherty, I would start much simpler than most people expect and add complexity only where the ACLR or EVM actually fails. A memory polynomial or GMP with moderate nonlinearity order — typically 7th to 9th order — and a short memory depth of 3 to 5 taps is usually enough to get within a few dB of the target, provided the training signal matches the real waveform in PAPR and bandwidth. Training on a CW tone or a narrowband signal and then switching to a 100 MHz carrier is the classic trap: the ACLR looks good in training and falls apart on the real signal.

The Doherty-specific problem is that the main and peaking devices turn on at different drive levels, so one global polynomial has to fit two different nonlinear shapes. I have had better results with a segmented or dual-input model that treats the low-power and high-power regions separately, rather than pushing a single model to very high order. Extra order mostly fits measurement noise and the observation receiver's own distortion, and it makes the coefficients unstable when the signal changes.

Memory depth should follow the envelope bandwidth and the thermal time constant, as LunaWei noted. Fast memory from bias and matching covers a few samples; slow thermal drift needs a separate slow adaptation or a temperature table, not deeper memory. My acceptance check is always the same: report carrier configuration, PAPR, model order and memory depth, iteration count, and ACLR, EVM and power consumption together. If adding memory buys less than 1–2 dB of ACLR, I stop there — the extra coefficients are not worth the risk on the bench.
R Member
From a field perspective: make sure monitoring is in place before deployment.
L Admin
@RF_Bench_Notes 对 100 MHz 5G 载波的 GaN Doherty,从中等阶数 GMP/记忆多项式起步、训练波形必须匹配真实 PAPR/带宽,这个顺序很对;Doherty 主峰管分区导通,用分段/双输入模型往往比一味抬阶更稳,多出来的阶数常在拟合观测接收机失真。记忆深度跟包络带宽与热时间常数,慢热漂用慢自适应或温度表,不是堆更深记忆。

编辑部台账三联:
1) 载波配置、PAPR、模型阶数/记忆深度、迭代次数,与 ACLR/EVM/功耗同页。
2) 训练信号 = 验收波形族;CW/窄带训完再上 100 MHz 一律作废。
3) 加记忆若 ACLR 增益 <1–2 dB 就停;系数不稳优先于「再深一点」。

欢迎贴:你们最终用的阶数/深度、分段是否必要,以及观测接收机失真何时开始污染系数。

— rfforum 编辑部 / Editorial Team
广告 AdRenesas 瑞萨微控制器芯片产品照片Renesas Electronics微控制器、模拟与射频/时钟方案 · MCUs, analog, RF and timing访问官网 →免费展示 · 资料来自公开信息 · 如需撤下请联系 ads@rfforum.com
广告 Ad广告位招商 / Advertise hereads@rfforum.com · 刊例 Rate cardT2 · 728×90
Log in or register to reply.