azazelovesH100 commited on
Commit
ef433ce
·
verified ·
1 Parent(s): 1a743c8

Replace with one-page circuit buckets list

Browse files
Files changed (1) hide show
  1. README.md +39 -299
README.md CHANGED
@@ -4,302 +4,42 @@ pretty_name: Copperhead circuit buckets
4
  tags: [pcb, kicad, electronics]
5
  ---
6
 
7
- # PCB circuit buckets: what matters for success, and where an AI KiCad agent fits
8
-
9
- **Date:** 2026-10-21 (Asia/Saigon, UTC+7) · **Scope:** public sources only · **Status:** v1
10
-
11
- **Subject.** Copperhead (copperhead.sh) is an open-source (Apache-2.0) TypeScript command-line AI agent for KiCad printed circuit board (PCB) design, built by founder Animesh Chouhan in Bengaluru. It writes specs, picks parts and the bill of materials (BOM), drafts schematics and first-draft placement, loops on electrical rules checks (ERC) and design rules checks (DRC) with `kicad-cli`, and exports Gerber fabrication files [1][4]. It is not an autorouter [1]. A sibling project, Fang, is a Python circuit language with simulation through ngspice (SPICE circuit simulation), openEMS (electromagnetic field simulation), a thermal model and simavr (microcontroller firmware simulation) [8][13].
12
-
13
- This document sorts the kinds of boards such a tool could be pointed at into buckets, and for each bucket asks: what decides success in real industrial practice, what a checker can verify, and how well Copperhead fits today.
14
-
15
- **Legend.** Numbers in square brackets, like [1], point to the Sources list at the end. **[UNVERIFIED]** marks a claim not confirmed from a primary source. Fit ratings, and paragraphs labelled "assessment", are this document's own analysis, not company or standards claims.
16
-
17
- ---
18
-
19
- ## 1. Answer in brief
20
-
21
- - **Realistic now:** microcontroller (MCU) control and dev boards, battery-powered sensor and Internet of Things (IoT) nodes that use pre-certified radio modules, simple low-voltage regulators and single-cell chargers, and low-voltage LED boards. These match everything Copperhead has publicly built or demoed [16][20][21][23], and their main risks are caught by ERC/DRC and datasheet checks [1][19].
22
- - **Next (partial today):** small motor drives and robot electronics, industrial input/output (I/O) boards, multi-cell battery management, precision analog and audio, drone flight controllers, automotive-style controller prototypes, and single-board-computer (SBC) or system-on-module (SoM) carrier boards. Copperhead can do the schematic and parts; success depends on layout judgment (current paths, isolation, impedance) that it leaves to a human [1][2].
23
- - **Out of reach today:** chip-down high-speed digital (DDR memory, PCIe), chip-down radio frequency (RF) design, mains and high-voltage power (inverters, on-board chargers, offline power supplies), and any production program gated by functional safety or medical certification. Copperhead is not an autorouter, ships a single 0.2 mm clearance rule with no creepage rules, and does not claim to be the engineer of record [1][3].
24
-
25
- ## 2. Buckets
26
-
27
- Fit ratings are this document's assessment; Copperhead publishes no coverage matrix. They rest on its documented features, its example difficulty tiers [5] and its demo record, and are capped by two facts: its README calls the agent loop "implemented, not yet proven" [1], and routing beyond power and short critical nets is left to a human [2].
28
-
29
- Three baseline facts recur below:
30
- - **Copperhead's checks** are KiCad ERC and DRC via `kicad-cli` [27], plus documentation-drift checks [1]. KiCad DRC can enforce clearance, track width, differential-pair gap, length and skew through custom rules, and since KiCad 9 a creepage constraint, but only for rules someone writes [24][25][26]. Copperhead's scaffolded project ships one default net class with 0.2 mm clearance and 0.25 mm tracks [3], and its documented example brief assumes "no controlled impedance" on a standard 2-layer process [4].
31
- - **Fang simulation** (ngspice, a copper-aware thermal model, openEMS, firmware simulation) is shown in company posts [10][11][12][13]. Whether the Copperhead CLI calls Fang during a design run is [UNVERIFIED].
32
- - **IPC classes.** IPC-A-610 (assembly acceptance) and IPC-6012 (bare-board qualification) define Class 1 (general), Class 2 (dedicated service, extended life) and Class 3 (high performance, no downtime, harsh environments) [36]. Class 3 tightens, for example, minimum plated-hole copper from 20 µm to 25 µm average and forbids the annular-ring breakout Class 2 allows [35].
33
-
34
- ### 2.1 MCU control and development boards
35
-
36
- **Fit today: Good.**
37
-
38
- - **Typical boards:** RP2040, ESP32, STM32 or nRF52 dev boards; USB-C breakout boards; small controller boards.
39
- - **Electrical success factors:** correct power tree and decoupling placed close to pins; crystal layout; USB 2.0 data pair routed as a 90 Ω differential pair [51]; strapping and boot pins set correctly; protection (ESD, reverse polarity) at connectors. Currents are usually low, so IPC-2152 trace sizing [28] rarely binds (assessment).
40
- - **Process factors:** standard 2- or 4-layer FR-4 at a low-cost fab such as JLCPCB, whose stock process allows 0.10 mm track and space on 2-layer boards [37]; panelization with rails, fiducials and tooling holes for assembly [42]; flying-probe test for prototypes and functional test (power up, flash firmware, exercise I/O) for production [39].
41
- - **Machine-verifiable:** most of it. ERC catches unconnected or conflicting pins; DRC catches fab-rule violations; pin maps can be checked against datasheets and firmware. Copperhead says code checks values against cited datasheet lines and caught "33 of 35 wrong values" in one test [19].
42
- - **Copperhead fit:** Good. This is the core of its demo record: an ATtiny84A USB-C LED board that is ERC/DRC clean but never built [23], an STM32 data logger [16], and RP2040 and USB-C example briefs rated "simple" [5].
43
-
44
- ### 2.2 Battery-powered sensor, IoT and wearable nodes
45
-
46
- **Fit today: Good (module-based); Partial (wearables, chip-down radio).**
47
-
48
- - **Typical boards:** soil or environmental sensor with ESP32 or nRF52, data logger, Bluetooth Low Energy (BLE) beacon, LoRa tracker, smart ring or glasses electronics.
49
- - **Electrical success factors:** a microamp-level sleep current budget; regulator quiescent current; sensor placement away from heat; antenna keep-out. Espressif recommends a 4-layer board with a complete ground plane under the chip, RF path and crystal, and keeping copper and components away from the antenna [48].
50
- - **Process factors:** using a module with its own radio certification lets the host product skip a new intentional-radiator certification under FCC (US Federal Communications Commission) rules, if the module grant's integration conditions are followed; the host still needs unintentional-emissions testing [49]. A simple FCC Part 15B test program is quoted at roughly $1,500 to $3,500 at one accredited lab [44].
51
- - **Machine-verifiable:** ERC/DRC, power-budget arithmetic, firmware behaviour in simulation. Antenna performance can be estimated with full-wave simulation (Fang has shown an openEMS antenna run [12]) but must be confirmed in a chamber.
52
- - **Copperhead fit:** Good for module-based nodes. Its STM32 logger has itemised power budgets but shipped draft power tracks below its own targets and 90 silkscreen DRC findings, and was never fabricated [16]. The Kyra smart-glasses and ring run reached part selection only [21]. A GNSS (satellite positioning) plus LoRa tracker is a "hard" example brief [5].
53
-
54
- ### 2.3 Low-voltage power conversion (buck, boost, LDO, USB Power Delivery, PoE)
55
-
56
- **Fit today: Good (simple regulators); Partial (USB-PD supplies, PoE).**
57
-
58
- - **Typical boards:** 12 V or battery to 5 V/3.3 V buck modules, boost converters, linear regulator (LDO) rails, USB Power Delivery (USB-PD) bench supplies, Power over Ethernet (PoE) powered devices.
59
- - **Electrical success factors:** the high-current switching loop. TI calls the input capacitor "the single-most important component for reliable operation of any step-down converter" and says to place it as close to the IC as manufacturing rules allow [46]. Also: small switch-node copper, inductor and output-capacitor ratings, loop stability, thermal pads with vias, trace width per IPC-2152 [28]. PoE adds an isolation barrier: IEEE 802.3 requires the powered device to withstand 1500 V RMS for 60 s (or 2250 V DC, or an impulse test) [47].
60
- - **Process factors:** 1 oz versus 2 oz copper changes the minimum track and space a fab can hold [37]; thermal verification on the bench; conducted and radiated EMC testing, where switching converters are a common failure source [45].
61
- - **Machine-verifiable:** operating point, ripple and transient response in SPICE; datasheet limits; copper temperature rise with a thermal model; trace width versus current against IPC-2152 curves [29]. DRC can enforce net-class widths once they are written.
62
- - **Needs judgment or test:** switching-loop layout, EMC, real thermal.
63
- - **Copperhead fit:** Good for single regulators: Fang has shown an MC34063 buck/boost checked against its datasheet reference design with ngspice and thermal runs [11]. Partial above that: the 100 W USB-PD supply is a "hard" example brief with no published run [5], and nothing shows Copperhead writing an isolation net class for PoE.
64
-
65
- ### 2.4 Battery management and charging
66
-
67
- **Fit today: Good (single-cell chargers); Partial (multi-cell low-voltage BMS); Poor (high-voltage packs).**
68
-
69
- - **Typical boards:** single-cell Li-ion charger with fuel gauge; 4S to 16S battery management system (BMS) with an analog front end (AFE), protection MOSFETs and balancing; energy-storage BMS.
70
- - **Electrical success factors:** Kelvin (4-wire) sensing at the shunt and cell taps, with low-current ground separated from the high-current path and joined only at the sense resistor [54]; cell-input RC filters placed close to the monitor IC [54]; FET-bank current capacity and thermal path; short-circuit response time; spacing across stacked cell voltages.
71
- - **Process factors:** heavy copper or busbars for high current; pack-level safety standards such as IEC 62619 or UL 2580 [UNVERIFIED]; functional test with a cell simulator; abuse tests at pack level.
72
- - **Machine-verifiable:** schematic correctness against the AFE datasheet, protection thresholds, thermal estimates, short-circuit timing in SPICE. Spacing rules are checkable only once voltage classes are mapped to net classes.
73
- - **Copperhead fit:** the repo's own 4S LiFePO4 BMS example (100 A peak, isolated CAN, 1 kV functional isolation) is rated "hard" and described as "safety-critical, high current, thermal and isolation constraints" [5][6]. A single-cell charger (BQ24075) appears in the STM32 logger [16]. High-voltage vehicle packs need functional safety work (section 2.12) and are out of reach.
74
-
75
- ### 2.5 Motor drives and robotics electronics
76
-
77
- **Fit today: Good (low-current drivers, sensor and CAN nodes); Partial (24-48 V servo drives); Poor (high-current ESCs and traction drives).**
78
-
79
- - **Typical boards:** brushed-DC and stepper drivers, small three-phase brushless (BLDC) servo drives, encoder interfaces, robot joint controllers on CAN bus, electronic speed controllers (ESCs).
80
- - **Electrical success factors:** TI's motor-driver layout guide stresses a short, wide supply-to-MOSFET-to-motor loop with few layer changes, multiple vias for high current, direct-connected thermal vias under the driver pad, and an unbroken ground plane [53]; gate-loop inductance; shunt placement; dead time; regenerative energy handling.
81
- - **Process factors:** 2 oz or heavier copper for phase currents; thermal interface to a heatsink or housing; drive safety standard IEC 61800-5-1 for industrial drives [31]; EMC.
82
- - **Machine-verifiable:** gate-drive transients and dead time in SPICE; thermal estimates; firmware pin maps; interface typing (Fang has typed CAN, PWM, encoder and motor-phase interfaces [8]).
83
- - **Needs judgment or test:** power-stage layout, switching EMI, thermal under load.
84
- - **Copperhead fit:** robotics is the vertical with the most public signal: the founder wrote "we are building claude code for robot electronics" [69], the Fang `servo_drive` example (three half-bridges, 24 V, 15 A per phase, CAN) is "the board this toolchain was written for" but is netlist only with no layout [9], and a dual brushed-DC driver is a "medium" brief [7]. Fang also showed a gate driver switching a 600 V-class MOSFET at 400 V and 10 A in ngspice, where moving one 0.1 µF capacitor changed the result [10]. High-current ESCs depend on exactly the routing Copperhead leaves to humans [2].
85
-
86
- ### 2.6 Industrial I/O and automation
87
-
88
- **Fit today: Partial.**
89
-
90
- - **Typical boards:** programmable logic controller (PLC) digital input and output modules, 4-20 mA current-loop analog inputs, RS-485/Modbus nodes, fieldbus and industrial Ethernet interfaces, isolated sensor interfaces.
91
- - **Electrical success factors:** standard input behaviour (IEC 61131-2 defines 24 V digital input Types 1, 2 and 3 by switching voltage and current thresholds [55]); galvanic isolation of signal and power where grounds differ [56]; RS-485 bus termination at the two cable ends only, typically 120 Ω, with short stubs [56]; surge, burst and ESD protection at every field port; isolated receivers such as TI's ISO1212 [67] and 4-20 mA reference designs such as TIDA-00548 [68]; functional safety per IEC 61508, which grades safety functions SIL 1 to 4 and uses FMEDA to quantify diagnostic coverage [61], for safety I/O.
92
- - **Process factors:** industrial temperature range parts; conformal coating for harsh sites, which also changes the IPC-2221 spacing category [30][41]; long product life, so BOM lifecycle risk matters [43]; EMC immunity testing per port. One lab quotes IEC 61326-1 industrial CE marking at $7,000 to $14,000 and notes each added port adds immunity injection time [44].
93
- - **Machine-verifiable:** schematic against reference designs; thresholds and loop burden in SPICE; isolation spacing via KiCad creepage and clearance rules once net classes exist [24][25].
94
- - **Needs judgment or test:** surge survival, isolation-barrier layout.
95
- - **Copperhead fit:** Partial. Fang types RS-485 and CAN interfaces [8], and these reference-design-driven circuits suit spec-to-schematic work. But isolation barriers need creepage rules that Copperhead does not generate [3], and no industrial example board exists in its public record. Industrial work appears only as a hiring "nice-to-have" [15].
96
-
97
- ### 2.7 SBC/SoM carrier boards and high-speed digital
98
-
99
- **Fit today: Partial (carrier schematic and parts); Poor (high-speed layout, chip-down DDR or PCIe).**
100
-
101
- - **Typical boards:** Raspberry Pi Compute Module 5 (CM5) or NVIDIA Jetson carriers with PCIe, USB 3, Ethernet and camera links; chip-down processor boards with DDR memory.
102
- - **Electrical success factors:** controlled impedance and length matching. The CM5 datasheet says PCIe, USB 3.0 and USB 2.0 pairs should be routed as 90 Ω differential pairs with length matching within each pair [51]; TI's high-speed guide says to route high-speed signals over a solid ground reference and not across plane splits or voids [50]; PCIe add-in-card interconnects target 85 Ω per a TI layout note [66]. A stackup agreed with the fab sets trace widths; via count and stubs matter.
103
- - **Process factors:** 6 to 8+ layers; controlled-impedance fabrication (JLCPCB offers it from 4 layers with ±10% tolerance [37]; its free impedance test is ±20%, precision testing is charged [38]); boundary scan (IEEE 1149.1/JTAG) to test nets hidden under ball-grid-array (BGA) parts, which finds opens and shorts but does not prove signal integrity [40]; X-ray inspection of BGAs.
104
- - **Machine-verifiable:** KiCad custom rules can check differential-pair gap, length and skew [26]; impedance follows from stackup geometry. Full signal-integrity simulation with IBIS models is outside anything Copperhead or Fang has shown [UNVERIFIED].
105
- - **Needs judgment or test:** stackup, power integrity, eye diagrams, bring-up.
106
- - **Copperhead fit:** Partial for carriers via its services motion: a forward-deployed engineer role owns "a live Raspberry Pi CM5 carrier-board project" and requires prior CM4/CM5 carrier experience, which implies a senior human does the layout [14]. Default rules assume no controlled impedance [4]. Chip-down DDR or PCIe is out of reach.
107
-
108
- ### 2.8 RF and wireless (chip-down)
109
-
110
- **Fit today: Poor (chip-down); see 2.2 for module-based designs.**
111
-
112
- - **Typical boards:** transceivers with discrete matching networks, multi-antenna cellular boards, 5.8 GHz video transmitters.
113
- - **Electrical success factors:** 50 Ω controlled-impedance feed lines over a continuous ground, ground-via stitching, matching networks tuned on the real board, antenna keep-out and coexistence between radios [48].
114
- - **Process factors:** RF laminates where FR-4 losses are too high (JLCPCB offers 2-layer Rogers/PTFE boards [37]); network-analyzer tuning; intentional-radiator certification for any non-modular radio [49]; a first-time EMC failure rate around 50% has been reported by a test lab [45].
115
- - **Machine-verifiable:** impedance from stackup; antenna patterns with full-wave simulation (openEMS); DRC keep-outs as rules.
116
- - **Copperhead fit:** Poor for chip-down work. Fang showed an openEMS run on an ESP32-C3 antenna [12], but default rules have no controlled impedance [4] and briefs with two RF chains are rated "hard" [5].
117
-
118
- ### 2.9 Precision analog, instrumentation and audio
119
-
120
- **Fit today: Partial.**
121
-
122
- - **Typical boards:** sensor conditioning, precision ADC front ends, instruments, audio amplifiers.
123
- - **Electrical success factors:** grounding and return paths. Analog Devices' MT-031 tutorial recommends a large low-impedance ground plane and, for most mixed-signal designs, one common ground plane rather than split analog and digital planes [52]; local low-inductance decoupling; guarding of high-impedance nodes; component tolerance and drift budgets; thermal gradients near references. Instrument safety for mains-connected equipment follows IEC 61010-1, which sets creepage by working voltage, pollution degree and material group [33].
124
- - **Process factors:** board cleanliness before coating [41]; calibration in functional test [39].
125
- - **Machine-verifiable:** this bucket has the richest simulation story. Gain, bandwidth, noise and stability are SPICE-checkable claims; the Fang repository includes a set of op-amp handbook example circuits with simulated claims [8].
126
- - **Needs judgment or test:** layout-induced noise, real accuracy.
127
- - **Copperhead fit:** Partial. Circuit-level correctness is well suited to SPICE-checked generation, but the outcome hinges on layout, which is human. A PDM microphone board appears as a benchmark fixture [18].
128
-
129
- ### 2.10 LED and lighting
130
-
131
- **Fit today: Good (low-voltage LED boards); Poor (mains LED drivers).**
132
-
133
- - **Typical boards:** PWM LED boards, DMX stage-lighting interfaces, constant-current drivers, mains bulb drivers.
134
- - **Electrical success factors:** thermal design first. TI notes that a metal-core PCB with an exposed pad and thermal vias gives roughly 10 °C/W for a PowerPAD driver versus about 70 °C/W for a SOIC-8 package in its example [57]; constant-current accuracy; flicker; for mains drivers, creepage and clearance and line-surge immunity [31].
135
- - **Process factors:** metal-core (aluminium) boards, which JLCPCB offers as 1-layer [37]; lighting EMC standards (CISPR 15, EN 61547 and harmonics limits) [44]; safety certification for mains products.
136
- - **Machine-verifiable:** current and thermal estimates; PWM firmware; IPC-2221 spacing tables are deterministic once voltage classes are known [30].
137
- - **Copperhead fit:** Good for low-voltage boards: the Quiet Orbit LED board is ERC/DRC clean (unbuilt) [23] and an rp2040 DMX board is a benchmark fixture [18]. Mains drivers are Poor: the only mains example is a teardown analysis, not a design [22], and there are no creepage rules [3].
138
-
139
- ### 2.11 Drone and aerospace avionics
140
-
141
- **Fit today: Good to Partial (flight controllers, GPS, BEC); Poor (4-in-1 ESCs, video transmitters); out of reach (certified airborne hardware).**
142
-
143
- - **Typical boards:** flight controllers, GPS and compass modules, battery-eliminator-circuit (BEC) regulators, power distribution boards, ESCs.
144
- - **Electrical success factors:** inertial-measurement-unit (IMU) noise. Betaflight's manufacturer guidelines require a dedicated LDO for the ICM-42688-P gyro, target supply noise under 50 µV peak-to-peak, require the gyro on its own SPI bus, and constrain motor-output timers and DMA [63]. PX4 also covers IMU placement and vibration isolation [64]. Power boards and ESCs are high-current, heavy-copper designs (see 2.5).
145
- - **Process factors:** Betaflight review of schematics, layouts and production samples before a board is marked "Betaflight Supported" [63]; flight testing with blackbox logs. Crewed aircraft electronics follow RTCA DO-254 design assurance, with levels A to E scaled by failure severity, recognised by the FAA in AC 20-152A [62].
146
- - **Machine-verifiable:** many Betaflight rules are checkable from the netlist (gyro on dedicated SPI bus, motor pins on one GPIO port, LDO present, specified LED and debug pins) [63]. This is an unusually crisp, public rule set for a checker.
147
- - **Needs judgment or test:** vibration, ESC and video EMI, flight behaviour.
148
- - **Copperhead fit:** drones are not a stated target; the only public signals are a hiring "nice-to-have" and a reply guessing at someone else's drone boards [15][70]. Technically, flight controllers resemble the STM32 logger (MCU, SPI sensors, flash) [16], so fit is Good to Partial, capped by dense 4-6 layer layout. ESCs are Poor for the reasons in 2.5. DO-254 work is out of reach.
149
-
150
- ### 2.12 Automotive and electric-vehicle electronics
151
-
152
- **Fit today: Partial (prototypes of low-voltage controllers and CAN nodes); Poor (production, high-voltage).**
153
-
154
- - **Typical boards:** CAN nodes, vehicle control units, body controllers, telematics, 12 V/48 V auxiliary converters; high-voltage BMS, inverters, on-board chargers.
155
- - **Electrical success factors:** 12 V supply robustness against cold crank, load dump and reverse polarity, with load dump now covered by ISO 16750-2 [59]; components qualified to AEC-Q100 (ICs) and AEC-Q200 (passives) [58]; vehicle EMC (CISPR 25) [59]; wide temperature range.
156
- - **Process factors:** functional safety under ISO 26262, whose Part 5 covers hardware development [60]; supplier quality processes; IPC Class 3 is common in safety-related uses [36]; one EMC test write-up says automotive requirements are significantly more stringent than consumer and medical ones [45].
157
- - **Machine-verifiable:** part qualification lookup (is every part AEC-Q rated?); transient protection in SPICE; CAN termination and topology; creepage for high-voltage nets via KiCad rules [24].
158
- - **Needs judgment or test:** safety analysis such as failure modes, effects and diagnostic analysis (FMEDA), validation, EMC.
159
- - **Copperhead fit:** Partial for prototypes: CAN nodes and controllers are reference-design-driven and Fang types CAN [8]. No AEC-Q filtering or ISO 26262 work products are documented [UNVERIFIED: based on keyword search of public repos]. High-voltage EV boards are out of reach (see 2.14). EVs are not a stated target.
160
-
161
- ### 2.13 Medical devices
162
-
163
- **Fit today: Poor (production); Partial (non-patient-connected prototypes).**
164
-
165
- - **Typical boards:** patient monitors, diagnostic and infusion equipment.
166
- - **Electrical success factors:** IEC 60601-1 means of protection. At a 250 V AC working voltage, two means of patient protection (2×MOPP) require 4,000 V AC isolation, 8 mm creepage and 5 mm clearance [34]; leakage-current limits; essential-performance under EMC per IEC 60601-1-2 [44].
167
- - **Process factors:** a quality management system and documented design controls (regulatory process, out of scope for a PCB tool); EMC programs quoted at $8,000 to $18,000 at one lab [44]; a test lab has reported that over 90% of electrical medical devices it tested failed the required standards on first submission [45].
168
- - **Machine-verifiable:** creepage and clearance once barriers are defined; leakage estimates.
169
- - **Copperhead fit:** Poor for production. Its pricing page offers an Enterprise "audit trail your compliance team can hand over" [71], but no medical feature, creepage rule generation or design-control workflow is public. Medical is a hiring "nice-to-have" only [15].
170
-
171
- ### 2.14 High-voltage and mains power
172
-
173
- **Fit today: Poor.**
174
-
175
- - **Typical boards:** offline switch-mode power supplies (SMPS), power-factor correction (PFC) stages, solar and motor inverters, EV on-board chargers and traction inverters, EV charging-station controllers.
176
- - **Electrical success factors:** insulation coordination. Creepage depends on working voltage, pollution degree and material group under IEC 60664-1, IEC 62368-1 (audio/video and IT equipment) or IEC 61010-1 [31][33]; one IEC 62368-1 calculator gives 2.5 mm basic creepage at 250 V RMS, pollution degree 2, material group IIIa, doubled for reinforced insulation [32]. IPC-2221 bare-board spacing for uncoated external conductors at 301-500 V is 2.5 mm below 3,050 m altitude [30].
177
- - **Process factors:** safety-agency certification; production hipot (dielectric withstand) testing; heavy copper [65].
178
- - **Machine-verifiable:** spacing rules (KiCad 9 creepage DRC [24][25]); switching waveforms in SPICE; thermal estimates.
179
- - **Needs judgment or test:** magnetics, power-loop layout, certification.
180
- - **Copperhead fit:** Poor. The default 0.2 mm single-class rule [3] is below the IPC-2221 uncoated external spacing for anything above 30 V [30], and Copperhead does not generate creepage rules [3]. Fang's 400 V/10 A gate-driver simulation [10] shows interest, not a design capability.
181
-
182
-
183
- ## 3. Summary table
184
-
185
- "Checkable" means the main success factors can be scored by an automated checker (ERC/DRC with appropriate rules, datasheet lookup, SPICE or thermal simulation) without a lab. All fit ratings are this document's assessment.
186
-
187
- | Bucket | Example | Key success factors | Checkable? | Copperhead fit |
188
- |---|---|---|---|---|
189
- | 2.1 MCU control and dev boards | RP2040 or STM32 board with USB-C | Power tree, decoupling, USB pair, boot pins | Mostly yes | Good |
190
- | 2.2 Sensor, IoT, wearable nodes | ESP32 soil sensor, BLE beacon | Sleep current, antenna keep-out, certified module | Mostly yes; radio range needs test | Good (module); Partial (wearable) |
191
- | 2.3 Low-voltage power | 12 V to 5 V buck, USB-PD supply | Switching loop, IPC-2152 widths, thermal, EMC | Partly (SPICE, thermal, widths) | Good (simple); Partial |
192
- | 2.4 Battery management | 1S charger; 4S-16S BMS | Kelvin sensing, FET current path, protection timing | Partly | Good (1S); Partial (multi-cell) |
193
- | 2.5 Motor drives, robotics | Brushed-DC driver, 24 V servo drive | Power loop, gate loop, thermal, shunts | Partly (SPICE, thermal) | Good (low current); Partial; Poor (ESC) |
194
- | 2.6 Industrial I/O | 24 V PLC inputs, RS-485, 4-20 mA | IEC 61131-2 behaviour, isolation, surge, lifecycle | Partly (rules once written) | Partial |
195
- | 2.7 Carriers and high-speed | CM5 carrier; chip-down DDR | Impedance, length match, stackup, return paths | Geometry yes; signal integrity no | Partial (carrier); Poor (high-speed) |
196
- | 2.8 RF chip-down | Sub-GHz radio with matching | 50 Ω lines, matching, keep-outs, certification | Weakly (EM sim estimates) | Poor |
197
- | 2.9 Precision analog, audio | ADC front end, microphone board | Grounding, noise, drift, guarding | Circuit yes; layout no | Partial |
198
- | 2.10 LED and lighting | PWM LED board; mains bulb driver | Thermal path, current accuracy, mains spacing | Partly | Good (LV); Poor (mains) |
199
- | 2.11 Drone avionics | Flight controller, GPS, ESC | IMU power and bus rules, vibration, high current | Netlist rules yes; flight no | Good-Partial (FC); Poor (ESC) |
200
- | 2.12 Automotive, EV | CAN node; VCU; HV BMS | AEC-Q parts, transients, CISPR 25, ISO 26262 | Parts lookup yes; safety no | Partial (prototype); Poor (production) |
201
- | 2.13 Medical | Patient monitor | 2×MOPP isolation, leakage, IEC 60601-1-2 | Spacing yes; process no | Poor |
202
- | 2.14 High-voltage, mains | Offline SMPS, inverter, OBC | Creepage, magnetics, power loops, certification | Spacing yes; rest no | Poor |
203
-
204
- ## 4. Cross-cutting success factors and implications
205
-
206
- **What industry does on every bucket.**
207
- - **Design reviews against datasheets and reference designs.** Betaflight's first review step is to check schematics and layouts against the application notes of each datasheet [63]. Copperhead's v0.12 `review` command covers track widths against current, differential pairs, length matching and clearances [17].
208
- - **DFM with the fab house.** Stackup, impedance, minimum track and space and copper weight are set by the fab's process, not by the designer alone [37][38]. Fabs and assemblers expect panel rails, fiducials and tooling holes [42].
209
- - **BOM availability and lifecycle.** Obsolescence is managed as a BOM-level risk; not-recommended-for-new-design parts should be avoided in new boards [43].
210
- - **Test coverage.** Flying probe for prototypes, bed-of-nails in-circuit test (ICT) for stable volume, and functional test as the final powered gate [39]; boundary scan where BGAs block probe access [40].
211
- - **Certification.** EMC and safety testing is a gate in most buckets; about half of devices reportedly fail EMC on the first try [45], and planning ranges run from about $1,500 for a simple FCC test to $8,000-$18,000 for medical EMC at one lab [44].
212
-
213
- **Implications for Copperhead (assessment).**
214
- - Its strongest buckets (2.1, 2.2, simple 2.3 and 2.10) are those where the hard parts are schematic correctness and part choice, which it automates, and where layout risk is low.
215
- - The single largest gap across 2.4, 2.5, 2.6, 2.12, 2.13 and 2.14 is that it does not translate voltage and current requirements into net classes: creepage and clearance by voltage class [24][31] and IPC-2152 widths by current [28]. KiCad already provides the checker; Copperhead would need to write the rules. This would move several Partial ratings up without solving routing.
216
- - Controlled impedance (2.7, 2.8) is also mostly rule-writing against a published fab stackup [37][38].
217
- - Production gates (Class 3, AEC-Q, ISO 26262, IEC 60601) are process and documentation work that no PCB tool replaces.
218
-
219
- **Implications for training data and verifiable environments (assessment).**
220
- - **Cleanest checkable rewards:** ERC/DRC pass with no collateral edits on real boards; pin-map and footprint grounding against KiCad libraries; datasheet-limit checks [19]; netlist-level rule sets such as Betaflight's flight-controller rules [63]; IPC-2221 spacing and IPC-2152 width lookups [29][30]; SPICE claims for analog and power circuits [8]. These map to buckets 2.1, 2.2, 2.3, 2.9, 2.11 (flight controllers) and the schematic side of 2.6.
221
- - **Partly checkable:** isolation and creepage (rules are deterministic once voltage classes are labelled correctly, but the labelling is a judgment); impedance geometry; thermal estimates.
222
- - **Hard to reward automatically:** EMC outcomes, RF performance, signal integrity on real hardware, vibration, and certification results. These need lab data. Copperbench, the company's benchmark, explicitly does not claim fabrication readiness or design quality [18].
223
-
224
- ## 5. Sources
225
-
226
- All sources were public web pages accessed 2026-10-19 to 2026-10-21. Standards are cited through publicly readable summaries where the standard itself is paywalled.
227
-
228
- **Copperhead and Fang**
229
- 1. Copperhead README: https://github.com/copperheadhq/copperhead/blob/main/README.md
230
- 2. `create` stage prompts (stage 5 layout: route power and short critical nets): https://github.com/copperheadhq/copperhead/blob/main/src/commands/create.ts
231
- 3. Scaffolded design rules (single default net class, 0.2 mm clearance): https://github.com/copperheadhq/copperhead/blob/main/src/kicad/bootstrap.ts
232
- 4. Docs, design from a brief: https://docs.copperhead.sh/workflows/create-from-brief/
233
- 5. Example briefs and difficulty tiers: https://github.com/copperheadhq/copperhead/blob/main/examples/README.md
234
- 6. Hard brief, 4S LiFePO4 BMS: https://github.com/copperheadhq/copperhead/blob/main/examples/hard/lifepo4-bms.md
235
- 7. Medium brief, dual brushed-DC motor driver: https://github.com/copperheadhq/copperhead/blob/main/examples/medium/dual-brushed-dc-motor-driver.md
236
- 8. Fang repository and docs: https://github.com/copperheadhq/fang · https://fang.copperhead.sh
237
- 9. Fang `servo_drive` example: https://github.com/copperheadhq/fang/tree/main/examples/servo_drive
238
- 10. Copperhead on X, 400 V/10 A gate-drive simulation: https://x.com/copperheadhq/status/2111021836310118742
239
- 11. Copperhead on X, Fang thermal and MC34063 checks: https://x.com/copperheadhq/status/2108111703141699785 · https://x.com/copperheadhq/status/2107169309659435409
240
- 12. Copperhead on X, openEMS antenna simulation: https://x.com/copperheadhq/status/2107451867701858779
241
- 13. Copperhead on X, firmware simulation: https://x.com/copperheadhq/status/2106386087635812737
242
- 14. Careers, Forward Deployed Hardware Engineer (CM5 carrier project): https://copperhead.sh/careers/forward-deployed-hardware-engineer/
243
- 15. Careers, Senior AI PCB Design Engineer (industry nice-to-haves): https://copperhead.sh/careers/senior-ai-pcb-design-engineer/
244
- 16. STM32 data logger example: https://github.com/copperheadhq/copperhead-stm32-data-logger
245
- 17. Review prompts (v0.12.0): https://github.com/copperheadhq/copperhead/blob/main/src/review/prompts.ts
246
- 18. copperbench: https://copperbench.org · standard: https://copperbench.org/standard
247
- 19. Copperhead on X, datasheet-limit checks ("33 of 35"): https://x.com/copperheadhq/status/2106006961255002223
248
- 20. Copperhead homepage: https://copperhead.sh
249
- 21. Blog, "Hardware as fast as software" (Kyra demo): https://copperhead.sh/blog/hardware-as-fast-as-software/
250
- 22. Blog, LED bulb teardown: https://copperhead.sh/blog/crompton-led-bulb-teardown/
251
- 23. Quiet Orbit LED board: https://github.com/copperheadhq/copperhead-quiet-orbit
252
-
253
- **KiCad**
254
-
255
- 24. KiCad 9.0 release notes (creepage DRC): https://www.kicad.org/blog/2025/02/Version-9.0.0-Released/
256
- 25. KiCad forum, using creepage custom rules: https://forum.kicad.info/t/how-do-i-use-the-new-creepage-custom-rules-v9-0-0/59328
257
- 26. KiCad custom rule constraints (diff_pair_gap, length, skew): https://github.com/KiCad/kicad-source-mirror/blob/c81ed537/pcbnew/dialogs/panel_setup_rules_help_2constraints.h
258
- 27. KiCad command-line interface docs: https://docs.kicad.org/9.0/en/cli/cli.html
259
-
260
- **Standards, fabs and industry practice**
261
-
262
- 28. IPC-2152 (ANSI webstore summary): https://webstore.ansi.org/standards/ipc/IPC21522009
263
- 29. Sierra Circuits, trace current capacity: https://www.protoexpress.com/blog/trace-current-capacity-pcb-design/
264
- 30. IPC-2221 Table 6-1 conductor spacing (reproduced by Altair): https://help.altair.com/pollex/topics/pollex/verification/dfe_ipc_material_r.htm
265
- 31. TI SLUP419, creepage and clearance standards (IEC 60664-1, IEC 61800-5-1): https://www.ti.com/lit/pdf/slup419
266
- 32. Creepage.com, IEC 62368-1 calculator (third-party tool): https://www.creepage.com/
267
- 33. Analog Devices, IEC 61010-1 Edition 3: https://www.analog.com/media/en/technical-documentation/technical-articles/iec-61010-1-edition-3.pdf
268
- 34. Analog Devices, protection strategies for medical device safety (MOOP/MOPP table): https://www.analog.com/en/resources/technical-articles/dont-get-zapped.html
269
- 35. PCB International, IPC-6012 class differences: https://www.pcbinternational.com/tech/ipcrequirementdifferences/
270
- 36. ACDI, IPC Class 2 vs Class 3: https://www.acdi.com/ipc-class-2-vs-class-3-assembly-processes-what-is-the-difference-part-1/
271
- 37. JLCPCB capabilities: https://jlcpcb.com/capabilities/
272
- 38. JLCPCB, standard laminated structures and impedance testing: https://jlcpcb.com/help/article/multi-layer-pcb-standard-laminated-structures
273
- 39. Electronics Brief, ICT vs flying probe vs functional test: https://www.electronicsbrief.com/articles/ict-vs-flying-probe-vs-functional-test
274
- 40. XJTAG, boundary scan FAQ: https://www.xjtag.com/support/faq/faq-jtag-boundary-scan/
275
- 41. Circuitnet, general design rules for conformal coating (white paper): https://www.circuitnet.com/news/uploads/2/General_Design_Rules_for_Conformal_coating_WHITE_PAPER.pdf
276
- 42. PCBWay, panelization, rails, fiducials, tooling holes: https://www.pcbway.com/helpcenter/design_instruction/PCB_Panelization__Breakaway_Rails__Fiducial_Marks__Tooling_Holes.html
277
- 43. CALCE (University of Maryland), part obsolescence: https://calce.umd.edu/electronic-part-obsolescence-forecasting-mitigation-and-management
278
- 44. Compatible Electronics, EMC testing cost guide: https://www.celectronics.com/Resources/EMC-Testing-Cost-Guide
279
- 45. EMI Software, first-time EMC failure rates (citing Intertek): https://www.emisoftware.com/blogs/percentage-products-fail-emc-tests-first-time/
280
- 46. TI SLYT614, five steps to a great step-down converter layout: https://www.ti.com/lit/an/slyt614/slyt614.pdf
281
- 47. TI SLUA736, lightning surge considerations for PoE powered devices: https://www.ti.com/lit/an/slua736/slua736.pdf
282
- 48. Espressif, ESP32 PCB layout design: https://docs.espressif.com/projects/esp-hardware-design-guidelines/en/latest/esp32/pcb-layout-design.html
283
- 49. FCC KDB 996369 D04, modular transmitter integration guide: https://apps.fcc.gov/kdb/GetAttachment.html?desc=996369+D04+Module+Integration+Guide+V02&id=bNCiEdkFEKnHsZF9GHCNdg%3D%3D&tracking_number=44637
284
- 50. TI SPRAAR7, high-speed interface layout guidelines: https://www.ti.com/lit/an/spraar7/spraar7.pdf
285
- 51. Raspberry Pi Compute Module 5 datasheet: https://datasheets.raspberrypi.com/cm5/cm5-datasheet.pdf
286
- 52. Analog Devices MT-031, grounding data converters: https://www.analog.com/mt-031
287
- 53. TI SLVA959, best practices for board layout of motor drivers: https://www.ti.com/lit/an/slva959b/slva959b.pdf
288
- 54. TI E2E FAQ, connecting cells to BQ769x2 monitors · TI fuel-gauge PCB layout guidance: https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1176908/faq-bq76952-connecting-battery-cells-to-the-bq769x2-family-of-monitors · https://www.ti.com/content/dam/videos/external-videos/en-us/1/3816841626001/3877771975001.mp4/subassets/pcb-layout-guidelines-fuel-gauge-design-presentation.pdf
289
- 55. Analog Devices, industrial digital inputs with the MAX22191 (IEC 61131-2 types): https://www.analog.com/en/resources/design-notes/industrial-digital-inputs-with-the-max22191.html
290
- 56. TI SLLA272, the RS-485 design guide · TI SLLA416, isolating RS-485 signal and power: https://www.ti.com/lit/an/slla272d/slla272d.pdf · https://www.ti.com/lit/ab/slla416c/slla416c.pdf
291
- 57. TI SNVA238, thermal performance of LM3404 (metal-core PCB): https://www.ti.com/lit/an/snva238a/snva238a.pdf
292
- 58. Automotive Electronics Council documents (AEC-Q100, AEC-Q200): http://www.aecouncil.com/Documents/AECDocuments.html
293
- 59. Analog Devices, load-dump protection for automotive applications: https://www.analog.com/en/resources/technical-articles/loaddump-protection-for-24v-automotive-applications.html
294
- 60. ISO 26262-5:2018 (hardware level): https://www.iso.org/standard/68387.html
295
- 61. exida, IEC 61508 overview: https://www.exida.com/articles/IEC-61508-Overview.pdf
296
- 62. FAA AC 20-152A, development assurance for airborne electronic hardware: https://www.faa.gov/documentLibrary/media/Advisory_Circular/AC_20-152A.pdf
297
- 63. Betaflight flight controller manufacturer design guidelines (v3.0, 2026-09-28): https://betaflight.com/docs/development/manufacturer/manufacturer-design-guidelines
298
- 64. PX4 reference flight controller design and vibration isolation: https://docs.px4.io/main/en/hardware/reference_design · https://docs.px4.io/main/en/assembly/vibration_isolation
299
- 65. Siemens, PCB high-voltage spacing: https://blogs.sw.siemens.com/electronic-systems-design/2025/04/29/pcb-high-voltage-spacing-what-every-engineer-should-know/
300
- 66. TI SNLA426, high-speed PCB layout for PCIe Gen 5: https://www.ti.com/lit/an/snla426/snla426.pdf
301
- 67. TI ISO1212 isolated 24 V digital input receiver datasheet: https://www.ti.com/lit/gpn/ISO1212
302
- 68. TI TIDA-00548, isolated dual-channel 4-20 mA analog input reference design: https://www.ti.com/tool/TIDA-00548
303
- 69. Founder on X, "we are building claude code for robot electronics": https://x.com/animeshsingh38/status/2082702268990931032
304
- 70. Copperhead on X, reply guessing at third-party drone boards: https://x.com/copperheadhq/status/2106969953509540263
305
- 71. Copperhead pricing: https://copperhead.sh/pricing/
 
4
  tags: [pcb, kicad, electronics]
5
  ---
6
 
7
+ # Copperhead circuit buckets: one-page list
8
+
9
+ ## Good fit now
10
+ - **MCU control and dev boards**: RP2040, ESP32 or STM32 boards. Matters: power tree and decoupling; 90 Ω USB pair; boot pins. Why this fit: core of its demo record.
11
+ - **Sensor and IoT nodes (module-based)**: ESP32 soil sensor, BLE beacon. Matters: µA sleep budget; antenna keep-out; pre-certified module. Why this fit: low layout risk.
12
+ - **Simple low-voltage regulators**: 12 V to 5 V buck, LDO rails. Matters: tight input-capacitor loop; IPC-2152 widths; thermal. Why this fit: Fang buck demo.
13
+ - **Single-cell chargers**: Li-ion charger with fuel gauge. Matters: charger datasheet compliance; thermal. Why this fit: used in its logger demo.
14
+ - **Low-voltage LED boards**: PWM LED board. Matters: thermal path; current accuracy. Why this fit: LED demo board.
15
+
16
+ ## Partial: needs a human for layout
17
+ - **Motor drives and robotics**: brushed-DC driver, 24 V servo drive. Matters: short power loop; gate loop; heavy copper; thermal. Why this fit: servo drive is netlist only.
18
+ - **Multi-cell low-voltage BMS**: 4S–16S BMS. Matters: Kelvin sensing; FET current path; protection timing. Why this fit: own example rated "hard".
19
+ - **Industrial I/O**: 24 V PLC inputs, RS-485, 4–20 mA. Matters: IEC 61131-2 behaviour; isolation; surge. Why this fit: no creepage rules.
20
+ - **USB-PD supplies and PoE**: 100 W bench supply, PoE device. Matters: switching layout; 1500 V PoE isolation; EMC. Why this fit: no isolation net class.
21
+ - **Precision analog and audio**: ADC front end, microphone board. Matters: grounding; noise; drift. Why this fit: layout decides.
22
+ - **Drone flight controllers**: Betaflight-style FC boards. Matters: clean IMU supply; dedicated gyro SPI bus; vibration. Why this fit: dense layout.
23
+ - **Automotive controller prototypes**: CAN node, vehicle control unit. Matters: load-dump protection; AEC-Q parts; CISPR 25. Why this fit: no AEC-Q filtering.
24
+ - **SBC/SoM carrier boards**: CM5 or Jetson carrier. Matters: controlled impedance; length matching; stackup. Why this fit: human-led; no impedance defaults.
25
+
26
+ ## Out of reach
27
+ - **Chip-down high-speed digital**: DDR or PCIe processor boards. Matters: impedance; return paths; signal integrity. Why this fit: no impedance rules.
28
+ - **Chip-down RF**: discrete-matched radio, 5.8 GHz transmitter. Matters: 50 Ω lines; matching; certification. Why this fit: rated "hard".
29
+ - **High-current ESCs**: 4-in-1 drone ESC. Matters: power loop; heavy copper; thermal. Why this fit: routing left to humans.
30
+ - **High-voltage and mains power**: offline SMPS, inverter, on-board charger, high-voltage BMS. Matters: creepage; magnetics; certification. Why this fit: no creepage rules.
31
+ - **Medical and safety-certified production**: patient monitor, ISO 26262 or DO-254 hardware. Matters: 2×MOPP isolation; safety analysis; certification. Why this fit: process work, not PCB tooling.
32
+
33
+ ## Easiest to check automatically
34
+ - MCU control and dev boards
35
+ - Module-based sensor and IoT nodes
36
+ - Simple low-voltage regulators
37
+ - Precision analog (circuit level)
38
+ - Drone flight controllers (netlist rules)
39
+ - Industrial I/O (schematic side)
40
+
41
+ ## Open questions
42
+ - Does the Copperhead CLI call Fang simulation? [UNVERIFIED]
43
+ - Do IEC 62619 / UL 2580 apply to the BMS boards in scope? [UNVERIFIED]
44
+ - Can Copperhead or Fang run full signal-integrity simulation? [UNVERIFIED]
45
+ - Are AEC-Q filtering and ISO 26262 work truly absent? [UNVERIFIED]