Tr5LC VCO 相噪机理LC VCO phase-noise mechanisms
相噪从哪来,
拆开给你看。Where phase noise
comes from, in parts.
VCO ISF Explorer 1.3.0 可从测试台开始,也可读取已完成的 VCO 仿真。沿工作点、噪声调制和 ISF 追踪相噪,再比较工艺角或变量扫描中的器件贡献与优化指标。VCO ISF Explorer 1.3.0 starts from a testbench or a completed VCO simulation. Trace phase noise through the operating point, noise modulation and ISF, then compare device contributions and optimization metrics across corners or variable sweeps.
转动选频偏Turn to pick an offset
演示片Film
从波形,
追到相噪来源。From waveforms
to the sources of phase noise.
从小电流脉冲怎样改变振荡相位讲起。影片用虚构电路演示双器件设置,用自带的合成示例浏览结果和扫描汇总,最后说明怎样用 ISF 解释仿真变化。配音随页面语言切换。See how a small current pulse shifts oscillator phase. A demo circuit shows two-device setup, and the included synthetic example shows the result views and sweep summary. The film ends with how to use ISF to explain a simulated change. Narration follows the page language.
界面Interface
核心数据与扫描结果。Core data, sweeps and result views.
00 核心数据页汇总振荡频率、相噪和闭合差值。扫描工艺角或变量时,同页逐点列出结果,并绘制相噪随扫描变量的变化。选中一个点后,各结果视图会同步切换。可在 03 相噪导航器对比计算与仿真。View 00 summarizes oscillation frequency, phase noise and closure. For a corner or variable sweep, it shows one row per point and plots phase noise against the swept variable. Select a point across views 00–08, or compare calculated and simulated phase noise in the view 03 navigator.
1.3.0 桌面截图,结果来自自带的合成 LC 示例。点击截图查看原图。Desktop screenshots from 1.3.0. Results use the included synthetic LC example. Select an image to see it full size.
为什么做Why
一个相噪数字,
说明不了原因。One phase-noise number
does not explain itself.
仿真给出的相噪曲线只告诉你结果。要改电路,得沿着输入频带、噪声调制和 ISF 查到器件,并分清相干路径与独立源的功率。A simulated phase-noise curve only gives the result. To change the circuit, trace input bands, noise modulation and the ISF back to devices, while keeping coherent paths and independent source powers distinct.
流程Workflow
准备测试台,追踪相噪。Prepare the testbench and trace phase noise.
在 Testbench setup 准备新测试,或从 Setup 读取已完成的 Maestro 仿真。Jobs & preflight 提供预检、存储空间估算与归档预览,已保存的任务可跨会话重开。Prepare a new test in Testbench setup or read a completed Maestro simulation from Setup. Jobs & preflight provides preflight checks, storage estimates and archive previews. Saved jobs can be reopened across sessions.
选择 A、B 或 CPick entry A, B or C
A 读取 Design Lab 项目导出。B 在已有 Maestro 设置中添加 ISF 测试,确认前可以查看保留和修改的内容。C 从选定的新单元配置测试台。三个入口均调用 VCO TB Generator。A reads a Design Lab project export. B adds an ISF test to an existing Maestro setup and previews what it keeps or changes before confirmation. C builds a testbench around a selected new cell. All three use VCO TB Generator.
逐器件核对Check each device
双器件分析中手动选两只 MOS,指定互补核心或非对称器件对,逐个检查估计与闭合。用对称近似外推之前,须核对实测端子波形的半周期一致性。For two-device analysis, select both MOS occurrences and choose Complementary core or Asymmetric pair. Check estimates and closure for each device. Symmetric-pair extrapolation requires measured half-period terminal agreement.
比较匹配的结果Compare matched results
按器件、机理、工况与频偏匹配改动前后的结果。在 08 查看优化指标,用 HBNOISE 核验相噪。ISF 验证并解释仿真变化,单看基线无法预测调参方向。Match devices, mechanisms, operating points and offsets across the before-and-after runs. Inspect optimization metrics in 08 and qualify phase noise with HBNOISE. ISF verifies and explains simulated changes. A baseline alone cannot predict tuning direction.
测试台生成Testbench generation
A、B、C 三个入口都调用 VCO TB Generator 1.0.0,在你的 Virtuoso 和 Maestro 环境中准备 L-d 级 ISF 测试。Entries A, B and C call VCO TB Generator 1.0.0 to prepare L-d ISF tests in your Virtuoso and Maestro environment.
源码与安装Source and installation
启动桌面,
先检查环境。Start the desktop,
check the environment.
安装后打开 Tk 桌面,点击 Load synthetic example 即可离线回放。Testbench setup 的原生操作需要 VCO TB Generator、本地 Virtuoso、Maestro、Spectre 和自备 PDK。Open the Tk desktop after installation and select Load synthetic example for offline replay. Native Testbench setup requires VCO TB Generator, local Virtuoso, Maestro, Spectre and your own PDK.
# 在源码目录安装并启动桌面界面# install and start the desktop UI from the source directory $ python3.11 -m venv .venv $ . .venv/bin/activate $ python -m pip install '.[analysis,images]' $ vco-isf --workspace ./isf-workspace # 比较已保存的结果# Compare saved results $ vco-isf metrics compare before.json after.json --point lc_ideal --offset-hz 10000 --json
- 运行环境Runtime
- Python 3.11+ 与 Tk 8.6。实时提取需 NumPyPython 3.11+ with Tk 8.6; NumPy for live extraction
- Maestro
- 在 .cdsinit 里载入 cadence/vco_isf_menu.il,从 Plug-in 菜单打开Load cadence/vco_isf_menu.il from .cdsinit and open it from the Plug-in menu
- 配置Config
- PDK 位置、器件名、工艺角由你配置。configs/ 只有虚构示例PDK location, device names and corners are your configuration; configs/ has fictional examples only
- CLI
- 用
vco-isf metrics读取或比较结果,用vco-isf optim的evaluate、gate和calibrate批量评估与判定。Usevco-isf metricsto inspect or compare results andvco-isf optimwithevaluate,gateandcalibratefor headless evaluation and qualification. - 许可证Licence
- MIT
限制Limits
- 测试通过不等于任何 PDK 或设计的数值签核。导入的数据一律标为回放、非签核。Passing tests is not numerical sign-off for any PDK or design; imported data is always marked replay, non-signoff.
- 拓扑识别需要所选 history 里可读的生成网表。不解码 schematic OA。自带运行器使用 HB、HBNOISE 和 HBXF。PSS/PNOISE 需要单独的适配器。Topology detection needs a readable generated netlist in the history; schematic OA is not decoded. The bundled runner uses HB, HBNOISE and HBXF. PSS/PNOISE requires a separate adapter.
- 建模范围为所选 gm 器件的闪烁和热噪声,以及受支持的无源 LC 热噪声。阵列开关及未支持的相关源保留为未建模。全 VCO HBnoise 保留这些来源。旧 thermal 预算缺少当前证据时仅作未验证诊断,合计显示 N/A。Modeled sources include flicker and thermal noise from the selected gm devices and thermal noise from supported passive LC loss. Array switches and unsupported correlated sources remain unmodeled and stay in whole-VCO HBnoise. Legacy thermal budgets without current evidence are unverified diagnostics, and the combined result is N/A.
- 谐波配对阶 k 与独立输入频带 n=|m| 分别展示。Qk 先相干相加,各独立频带再按功率相加。热噪声为有限频带预算,需检查阶数收敛。Harmonic pairing order k and independent input band n=|m| have separate views. Qk terms add coherently, while independent bands add in power. Thermal results use a finite input-band budget and need order-convergence checks.
- 指标与质量页保留数值告警和相位候选证据。thermal 频带页按源和 |m| 查看功率。缺失分量、同口径参考和未验证旧预算均有提示。窗口在任务运行中拒绝关闭。停止会等待当前频偏完成,已有结果可续跑。Quality retains numerical warnings and phase-candidate evidence. Thermal bands show power by source and |m|. Missing components, same-scope references and unverified legacy budgets are flagged. The window refuses to close during an active job. Stop finishes the current offset and preserves results for resume.
其他工具Other tools