RF Power Amplifier
PA Placement
original, fixed (hardcoded), SA 1, QP, QP+SA (Small WireLen, non 0 viol.),
QP+ILP+PrePass (No A* or SA, viol.) comparison with ACS summary, detail
(切換完整版課綱)
Unit 1: AI Computer Vision
CNN (Convolutional Neural Networks) Handwashing
YOLO (You Only Look Once) YOLOv7 list
ViT (Vision Transformers) wafer defects heat map comparison
VLM (Vision Language Models) nutrition analysis
Simulating A Navigation Assistant with sonar
HW#3 Obstacle detection
Unit 6:
Unit 7: AI for Game Designing
Unit 8 Space-Time Evolution
Unit 9 Advanced Prompts
Anthropology
Design
___________________________________________________________
Prompt Engineering Guide
Prompts for scientific accuracy
課堂練習
Deadline: This Saturday at 23:59
Send all the share links to me chang212@gmail.com by email with subject HW#6 ［your id, your name]
任選一題
1. Design a Two-Stage BJT Amplifier according to goals specified using A* (Parameters to optimize RE1, RC1. RE2. RC2)
2. Optimize LM3886 class AB audio amplifier IC using A* (Parameters to optimize: resistors except the load)
A Search Setup*
Cost function (5 components, all minimized):
A* search for optimization, share 4-parameter (for bypassed model c_mu=3 pF, c_pi=22 pF) note: goal reached is actually current step, not current best
revised to OCT model A* search for optimization, share 4-parameter (for bypassed model c_mu=3 pF, c_pi=22 pF)
Lesson 2
Lesson 3
1, 2 任選一題
使用 Claude Sonnet 4.6 推理模式(手動切換，免費用戶額定時間內只能使用三次)
使用 ChatGPT 5 推理模式(自動切換)
使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)
How to publish a Claude artifact
Content share 作業繳交格式
課堂練習
Deadline: This Saturday at 23:59
Send all the share links to me chang212@gmail.com by email with subject HW#4 [your id, your name]
Study Apple C1 Architecture
(a) On Claude, Run NGSpice for the LNA using given parameters. If the results do not match, explain why.
(b) Run an MNA simulation (Modified Nodal Analysis, as used in Cadence Spectre engine). Optimize the IC design to meet specs.
PA Placement
original, fixed (hardcoded), SA 1, QP, QP+SA (Small WireLen, non 0 viol.),
QP+ILP+PrePass (No A* or SA, viol.) comparison with ACS summary, detail
PA Parameter Optimization
original (analytical), Hessian, MNA (post QP+SA)
PA Lessons, PA Verified by by computing the exact KCL residual at its converged solution, not because of blow up of MNA (share)
PA issues system crash, reality check, UI design
LNA reality check in simulator, prompt for spec compliance
LNA v3 (closed-form), V3 pathway to tape-out (advanced, ok omitted)
使用 Claude Sonnet 4.6 推理模式(手動切換，免費用戶額定時間內只能使用三次)
使用 ChatGPT 5 推理模式(自動切換)
使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)
使用 Grok 4 推理模式(自動切換)
How to publish a Claude artifact
Content share 作業繳交格式
課堂練習
Deadline: This Saturday at 23:59
Send all the share links to me chang212@gmail.com by email with subject HW#3 [your id, your name]
1. Study Apple C1 Architecture
(a) make
LNA schematic (TSMC N7 製程的完整設計流程，包含完整元件表、元件值推導公式、die 面積估算) LNA=Low Noise Amplifier
(b) make
LNA Die
(c) Build
LNA Optimizer (SMITH Chart, Frequency Response)
(d) Build a Simplified (closed-form) LNA optimizer on Die, 3.5 GHz LNA TSMC N7.
(e) Build an MNA model (Modified Nodal Analysis, as used in Cadence Spectre engine) to optimize. Must verify your results to meet spec and parameters have to be realistic.
Voronoi start-up A* + Repelling (example illustrated)
使用 Claude Sonnet 4.6 推理模式(手動切換，免費用戶額定時間內只能使用三次)
使用 ChatGPT 5 推理模式(自動切換)
使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)
使用 Grok 4 推理模式(自動切換)
How to publish a Claude artifact
Content share 作業繳交格式
課堂練習
Deadline: This Saturday at 23:59
Send all the share links to me chang212@gmail.com by email with subject HW#2 [your id, your name]
1.
Solve Pz 2 using chain of thought (CoT) reasoning with each step whether correct or incorrect, showing backtracking.
Animate the CoT process in the above synced with river crossing scenarios.
Animate search tree with synced progression (A*/BFS side-by-side)
2.
Solve Pz 20 using chain of thought (CoT) reasoning with each step whether correct or incorrect, showing backtracking.
Animate the CoT process in the above synced with river crossing scenarios.
Animate search tree with synced progression (A*/BFS side-by-side)
Chain of Thought vs. Cognition (concepts in slides with 3 IC design examples, enlarged)
使用 Claude Sonnet 4.6 推理模式(手動切換，免費用戶額定時間內只能使用三次)
使用 ChatGPT 5 推理模式(自動切換)
使用 Gemini 3.0 Pro 免費額度最高 1M tokens (永遠推理模式)
使用 Grok 4 推理模式(自動切換)
Deadline: This Saturday at 23:59
Send all the share links to me chang212@gmail.com by email with subject HW#1 [your id, your name]
How to publish a Claude artifact
Content share 作業繳交格式
任選一題
1.
(a) Solve River Puzzles (Pz 3, Pz 20) by Chain of Thought and Visualize/Animate the process
(b) Verify each step not violating constraints and correctness.
2. Comparing LLMs: Claude 4.6, Gemini 3.0 Pro, ChatGPT 5.0 (三選二) for the following task.
Animate in 3d (in three.js) Yagi-Uda Antenna, AI pipe inspector, or things you choose. （擇一）
The animation must be scientifically accurate.
Hints
推薦 Opus 4.6 搭配思考模式（extended thinking）
我的使用體驗：
一次到位 vs. 反覆修正：這是 Opus 和 Sonnet 之間最核心的差異。Sonnet 在簡單任務上表現不錯，但面對複雜問題——像我們常做的 Three.js 地質動畫、LNA 教學投影片、A* 搜尋演算法視覺化這類高度整合的工作——Sonnet 容易在細節上出錯，需要來回修正好幾輪。Opus 因為推理深度更深，往往第一次就能把架構、邏輯、
Token 消耗的取捨：誠實地面對 Opus 單次消耗更多 token，
「比較一陣子才固定使用」：經過實際比較後的結論，
Prompt in English vs. Traditional English (中文版)
為什麼我的AI跟你不一樣: factors, basic skills，美學
supplemental
提示使用英文的策略 (Word)
