Jlcpcb Parts

JLCPCB Part Selection

1. When to use

Use this skill whenever the user is:

• Choosing components for a board that JLCPCB will assemble (PCBA).
• Matching a KiCad schematic or BOM CSV to LCSC part numbers (C#####).
• Comparing candidate parts, finding alternatives, or de-risking a BOM.
• Estimating assembly cost (component cost + loading fees) for N boards.

Do NOT hand-pick parts from memory. Part numbers, stock, prices, and tier status change constantly — always go through the MCP tools below, and re-verify immediately before the user actually orders.

2. Core facts

Part tiers and loading fees

• The fee is per unique part, not per placement: 100 boards × 1 extended part = $3 total; 1 board × 10 different extended parts = $30.
• Tier mapping from search results: is_basic → basic, else is_preferred → preferred, else extended.
• LCSC numbers are written "C" + id (25804 → C25804). Part page: https://jlcpcb.com/partdetail/C<id>.

Economic vs Standard PCBA caveat

The preferred-tier fee waiver applies to Economic PCBA only. If the design forces Standard PCBA (through-hole assembly, parts on both sides, unusual surface finishes, large/heavy boards — see references/heuristics.md), preferred parts are charged like extended ones. When a design smells like Standard PCBA, count preferred parts into the fee risk and say so.

Price breaks

• Prices are quantity-tiered (qFrom/qTo breaks). Always quote the unit price at the quantity actually needed (board_qty × qty_per_board), never the qty-1 price or the deepest break.
• A part that is cheaper at qty 10 can be more expensive at qty 1000 than a rival — re-rank when the user changes board quantity.
• Some sources return a single price with no breaks; treat that price as approximate and flag "price unknown at volume".

3. Selection heuristics (ranking doctrine)

Apply in this order — earlier criteria dominate later ones:

1. Tier first. A basic part beats a preferred part beats an extended part at comparable specs. The suggestion engine scores tier basic +100 / preferred +60 / extended +0 — that dominance is intentional: a $3 fee wipes out any per-unit saving on cheap passives. Only pick extended when no basic/preferred part meets the spec.
2. Stock adequacy. Required stock ≥ 5-10× the production need (need = board_qty × qty_per_board). For passives additionally demand >10k absolute stock — anything thinner can vanish between quoting and ordering. Stock < need is a hard "stock risk" warning; stock ≤ 0 disqualifies.
3. Price at the needed quantity. Compare candidates at the same quantity using their price breaks. Price is a tiebreaker after tier and stock, not the lead criterion.
4. Package consolidation. If the BOM has the same value in multiple packages (10k in 0402 and 0603), propose consolidating to one package — fewer unique parts, fewer feeders, fewer fee-bearing lines.
5. Alternatives. For every critical or single-source line (ICs, connectors, crystals, anything extended), name at least one backup part. A BOM with no second source on its MCU is a schedule risk, say so explicitly.

Component-type guidance:

• Passives: prefer 0603, then 0402 — that's where the basic-tier catalog is densest. Resistors: default ±1% (basic 1% parts cost the same as 5%). Capacitors: for decoupling/general use specify X7R (or better) and a voltage rating ≥ 2× the rail; avoid Y5V. Exact-value fixation wastes money — accept E-series neighbors when the circuit tolerates it (rules in references/heuristics.md).
• ICs — text-match risk. Search is fuzzy text; "LM358" matches dozens of suffix variants with different packages, grades, temperature ranges. Always verify the full MPN AND the package against the datasheet before accepting an IC match. Never accept an IC line on description similarity alone — that is exactly what needs_review status exists for.
• Connectors: verify the footprint's pin pitch, pin count, and mounting style against the candidate's drawing. "USB-C connector" matches dozens of mechanically incompatible parts. Treat every connector match as needs_review until footprint-verified.
• Stock volatility: stock figures are a snapshot. Re-run get_part on every chosen LCSC number right before the user orders; flag any line whose stock dropped below the 5× threshold.

4. Workflow (exact tool sequence)

1. analyze_kicad — pass the .kicad_sch or BOM .csv path. Review the grouped BOM lines: check quantities look sane, note DNP lines and any pre-assigned LCSC fields.
2. suggest_bom_parts — pass the same path (or explicit bom_lines) plus board_qty (ask the user; default 10). Optionally stock_multiple (default 5) and max_candidates (default 5). Read the per-line status:
   • matched — accept the chosen candidate; skim its warnings.
   • preassigned — the schematic pinned an LCSC number. Keep it, but relay any low-stock or extended-fee warnings; suggest a basic swap if one exists.
   • skipped_dnp — do not source, do not cost; list it as DNP in the output.
   • needs_review — see step 3.
   • no_match — see step 4.
3. Resolve needs_review lines. Typical causes: IC matched only by fuzzy text, or the top pick is extended with a stock warning. Run get_part on the candidate to inspect full attributes and price breaks, verify MPN + package against the datasheet, and run find_alternatives to look for a basic/preferred or better-stocked substitute. If still ambiguous, present the top 2-3 candidates and ask the user to decide.
4. Resolve no_match lines. For resistors/capacitors, retry with search_passives (kind, value like "10k"/"100nF", package). For everything else, retry search_parts with progressively looser queries (MPN → MPN family → function keywords), or relax the package. If nothing fits, tell the user the line needs a manual pick or a consigned part — never silently drop it.
5. estimate_assembly_cost — pass the final chosen list as parts: [{lcsc, qty_per_board}] plus board_qty for the authoritative cost breakdown (component total, loading fees, per-board).
6. Present the suggested BOM in the format below, with per-line reasoning, the cost summary, and the risk list.
7. Import into KiCad (optional, on request). When the user wants a chosen part usable in their design, run import_part_to_kicad with the lcsc (one or many) and project_dir (their KiCad project). It fetches the symbol, footprint, and 3D model (STEP + WRL), writes them into the project's libs/lcsc/ libraries, and registers them in the sym/fp-lib-table so they appear in the choosers — the user just reopens the project. Pass a whole accepted BOM's LCSC list to import everything at once; pass include to limit to e.g. symbol+footprint. Requires the kicad-lcsc-manager engine and python3; if it reports the engine is missing, point the user to install it. Do NOT import a needs_review IC/connector before its MPN + footprint are confirmed.

5. Output format

Suggested-BOM table (one row per BOM line; Qty = per board):

• Unit$ = unit price at the needed quantity; Line$ = unit × qty_per_board × board_qty. DNP rows: mark DNP, no LCSC/cost.
• Why = one clause: tier + the deciding factor ("Basic, 30M stock", "Only X7R 0603 in stock", "Extended — no basic equivalent").

Cost summary block:

Boards: 10
Components: $X.XX
Loading fees: $Y (N extended parts × $3; preferred waived on Economic PCBA)
Total: $Z.ZZ  (≈ $Z.ZZ / board, excl. PCB fab + assembly service fee)

Risk list: bullet every warning — low stock lines, extended fees, text-matched ICs pending datasheet check, single-source lines without alternatives, DNP.

6. Worked micro-example

3-line BOM, 10 boards:

Boards: 10
Components: $1.37
Loading fees: $3 (1 extended part × $3)
Total: $4.37  (≈ USD 0.44 / board, excl. PCB fab + assembly service fee)

Risks: U1 is extended and single-source here — alternative: AMS1117-3.3 (C6186, basic, no fee) if the higher dropout and ~5mA quiescent current are acceptable; ask the user before swapping. Re-verify all stock before ordering.