design-logic-circuit
关于
This Claude Skill designs gate-level combinational logic circuits from functional specifications or truth tables. It implements circuits using basic gates, handles NAND/NOR universal conversions, and builds standard components like adders and multiplexers. Use it to translate Boolean logic into a verified, hardware-realizable network.
快速安装
Claude Code
推荐npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/design-logic-circuit在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Design Logic Circuit
Spec → combinational circuit. Define I/O, derive min Bool expr, map → gate schematic, (optional) convert to universal gate basis (NAND/NOR), verify via exhaustive sim.
Use When
- Bool fn → gate net (physical or sim)
- Std combinational blocks (adders, muxes, decoders, comparators)
- Convert arbitrary → NAND-only / NOR-only for mfg
- Teach/review digital logic spec → schematic
- Prep combinational datapath for build-sequential-circuit or simulate-cpu-architecture
In
- Required: Spec — truth table, Bool expr, verbal I/O desc, or std block name (e.g., "4-bit ripple-carry adder")
- Required: Target gate lib — unrestricted (AND/OR/NOT), NAND-only, NOR-only, or std cell lib
- Optional: Optimize goal — min gate count, min prop delay (critical path), min fan-out
- Optional: Max fan-in (e.g., 2-input only)
- Optional: Don't-cares
Do
Step 1: Spec
Interface + behavior before synthesis:
- Inputs: Names, widths, ranges. Multi-bit → bit order (MSB/LSB-first).
- Outputs: Names + widths.
- Truth table: Every input combo → outputs. Many inputs → algebraic or minterms/maxterms.
- Don't-cares: Input combos that can't occur (BCD 1010-1111) → mark.
- Timing: Prop delay constraints. Combinational = no clock → worst-case gate delay through critical path.
## Circuit Specification
- **Name**: [descriptive name]
- **Inputs**: [list with bit widths]
- **Outputs**: [list with bit widths]
- **Function**: [verbal description]
- **Truth table or minterm list**: [table or Sigma notation]
- **Don't-care set**: [d(...) or "none"]
→ Unambiguous spec, every legal input → exactly one output.
If err: Ambiguous (missing cases, conflicting outputs) → clarify. Don't assume don't-care unless told.
Step 2: Derive min Bool expr
evaluate-boolean-expression skill:
- Single-output: Each bit → min SOP (or POS, whichever fewer gates).
- Multi-output: Shared sub-exprs → factor out. Fewer gates + more routing.
- XOR detection: Checkerboard patterns in K-map. XOR expensive in NAND/NOR-only, efficient in std libs.
- Record: Min expr each output + literal count + term count.
## Minimal Expressions
| Output | Minimal SOP | Literals | Terms |
|--------|-------------|----------|-------|
| F1 | [expression] | [count] | [count] |
| F2 | [expression] | [count] | [count] |
- **Shared sub-expressions**: [list, if any]
→ Min expr each output + shared sub-exprs ID'd.
If err: Non-minimal (more literals than expected) → re-run K-map or Quine-McCluskey. >6 vars → Espresso.
Step 3: Map → gate schematic
Bool → gate network:
- Direct SOP: Product term → multi-input AND. Sum → OR fed by ANDs. Complemented var → NOT (or use NAND/NOR to absorb).
- Gate assignment: Each gate:
- Type (AND, OR, NOT, XOR, NAND, NOR)
- Inputs (name or from other gate)
- Output name
- Fan-in
- Fan-in decomp: Exceeds max → tree of smaller. 4-input AND w/ 2-input → 2 two-input ANDs feeding 3rd.
- Notation: Text netlist or structured.
## Gate-Level Netlist
| Gate ID | Type | Inputs | Output | Fan-in |
|---------|------|-------------|--------|--------|
| G1 | NOT | A | A' | 1 |
| G2 | AND | A', B | w1 | 2 |
| G3 | AND | A, C | w2 | 2 |
| G4 | OR | w1, w2 | F | 2 |
- **Total gates**: [count]
- **Critical path depth**: [number of gate levels from input to output]
→ Complete netlist, every output traceable to primary inputs, no floating.
If err: Dangling wires or feedback loops (invalid combinational) → recheck. Every signal = exactly one driver, every gate input connects.
Step 4: Convert → universal basis (optional)
NAND-only or NOR-only:
- NAND-only:
- AND → NAND + NOT (NAND tied inputs)
- OR → De Morgan:
A + B = ((A')*(B'))' = NAND(A', B')→ NOTs then NAND - NOT →
A' = NAND(A, A) - Bubble pushing: Cancel adjacent inversions. 2 NOTs series cancel. NAND → NOT = AND.
- NOR-only:
- OR → NOR + NOT
- AND → De Morgan:
A * B = ((A')+(B'))' = NOR(A', B') - NOT →
NOR(A, A) - Bubble push cancels inversions.
- Gate count: Before + after. NAND/NOR-only typ more gates but simplify mfg.
## Universal Gate Conversion
- **Target basis**: [NAND-only / NOR-only]
- **Gates before conversion**: [count]
- **Gates after conversion**: [count]
- **Gates after bubble-push optimization**: [count]
- **Conversion netlist**: [updated table]
→ Functionally equiv, redundant inversions eliminated via bubble push.
If err: More inversions than expected → re-examine bubble push. Common: forgetting NAND/NOR self-dual under complementation. De Morgan consistently from outputs back → inputs.
Step 5: Verify via exhaustive sim
Correct for every input:
- Approach: ≤16 inputs (65,536 combos) → exhaustive. Larger → targeted vectors + corner cases + random.
- Propagate: Each combo, propagate gate by gate in topological order.
- Compare: Check each output vs truth table. Don't-cares → 0 or 1.
- Record: Mismatches w/ input + expected vs actual.
- Timing (optional): Count gate levels longest path. × per-gate delay → worst-case prop.
## Simulation Results
- **Total test vectors**: [count]
- **Vectors passed**: [count]
- **Vectors failed**: [count, with details if any]
- **Critical path**: [gate sequence, e.g., G1 -> G3 -> G7 -> G9]
- **Critical path depth**: [N gate levels]
- **Estimated worst-case delay**: [N * gate_delay]
→ All vectors pass. Functional + critical path docs.
If err: Vector fails → trace path gate by gate. First gate w/ incorrect output. Common: wire wrong input, missing inversion, NAND/NOR conversion err.
Check
- All I/O named + widths spec'd
- Truth table covers all legal combos
- Bool exprs min (K-map/Quine-McCluskey)
- Every gate all inputs connected + exactly one output
- No combinational feedback
- Fan-in constraints respected
- NAND/NOR conversion preserves equivalence
- Bubble push eliminates redundant inversions
- Exhaustive sim passes (non-don't-cares)
- Critical path depth docs
Traps
- Combinational feedback: Output → own input chain = latch, not combinational. State needed → build-sequential-circuit.
- Forget inversions in NAND/NOR: Most common err = dropping NOT in De Morgan. Bubble push systematically outputs → inputs, not ad hoc.
- Exceed fan-in w/o decomp: 5-input AND not in 2-input lib. Balanced tree min delay, not linear chain.
- Ignore don't-cares: Unexploited → bigger circuit. Always include when avail.
- Gate vs wire delay: Intro design → gate delay dominates. Real VLSI → wire delay can exceed. Note when estimating.
- Multi-output hazards: Shared gates → change one affects shared sub-expr. Verify all outputs after any mod.
→
evaluate-boolean-expression— derive min Bool expr for this skillbuild-sequential-circuit— add state (flip-flops) → sequentialsimulate-cpu-architecture— use blocks (ALU, mux, decoder) as datapath
GitHub 仓库
相关推荐技能
content-collections
元Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。
polymarket
元这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。
creating-opencode-plugins
元该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。
sglang
元SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。
