Coverage of typical maker projects

A source

This page is a source of edakit – a concept, a prompt or the rules of a session –, shown as it is in the repository. Read it as such; the guides say what edakit does.

Status 2026-10-30: research and analysis. Built since: maker-coverage-status.toml records each candidate – 92 built, 6 partial, 5 out on 2026-11-01 –, and the catalog page shows it. The candidates were built after the restructuring of the catalog in rules-at-scale.md, as data per domain (its section 10). What becomes a check, a hint, a block or a part fact is the owner's decision (handbook 14.4), and every candidate below still needs its sources opened and read before it gets a catalog entry (skill add-rule, step 2). Basis: edakit at bf7450a, seven research clusters with 249 pitfalls from about 80 maker project types and the guitar's pedals, instrument and amplifiers, and the owner's 21 repositories. Of the 249, 50 are covered, 11 covered once the parts entry carries the datasheet value, 69 partly and 119 not – some of those judgement or out of scope. Every pitfall with its numbers, its source and what covers it: the appendix. edakit is open source and meant for makers' boards in general, not only the owner's: the candidates are a build list in the order of section "Priority", and section "Keeping the catalog stable" says how it stays trustworthy while it grows.

Question

Do the checks, hints and blocks of edakit cover what typical maker projects get wrong – and what is missing?

Short answer

• Covered: the class edakit grew from – boards supplied from USB or cells, with an MCU module, LEDs, buttons, a servo or buzzer, I2C and a radio module for the EU. Supply path, reverse polarity, back-feed between sources, voltage domains, logic levels with known thresholds, decoupling, strapping pins, I2C pull-ups, flyback diodes on the board, USB-C CC, junction and touch temperature, the board and the manufacturing files.
• Covered only with part facts: many frequent faults are found by existing checks once the parts entry carries the datasheet fact – the VIH ratio of a WS2812B (E-3), the strapping pins of an ESP32 (G-12), pin functions (G-15), GPIO limits (E-2). edakit ships 15 checked entries, all of them parts of its blocks. The fault found in the owner's pedal (below) is of this kind.
• Not covered:
• load over time – a servo or motor starting, a Wi-Fi burst, a coin cell's pulse: transients vary only the supplies;
• pin semantics – open-drain and tri-state outputs, unused inputs left nc, bus addresses;
• analog – ADC window and source impedance, op-amps (no model), polarized capacitors;
• switching parts – MOSFETs that are not logic-level, high-side P-MOSFETs driven from 3.3 V;
• inputs above their own supply – back-powering through clamp diodes;
• cables as more than a resistance – length, gauge, capacitance, inductance;
• audio – bridge-tied outputs, speakers, amplifier leads;
• hazardous voltage – spacing, isolation and input topology of mains and high-voltage boards;
• RF placement – the antenna keep-out of modules;
• support circuits a datasheet asks for (EN RC, VCAP, crystal load, USB series resistors), and kinds the graph has no word for: speaker, optocoupler, motor, heater;
• guitar circuits – the input impedance a pickup needs, op-amps at 9 V, DC at switches and pots, the bias spread of discrete stages.
• Stability: today the catalog is quiet – 0–1 waiver per board on the bell, 7 for the reference project's 84 parts. Built one by one, the candidates would about double it. Decided by the owner: domains per project, and rules condensed into mechanisms – seven of them cover 91 of the 128 open maker pitfalls. Proposed besides: a report of what went unchecked, new checks as warnings first, and reference circuits per domain to test them against. For the model – it must run with Sonnet – a task stays at a few ten thousand tokens; the risk is density (33 words per sentence in the skills) and whole files read where a slice would do (edakit rules prints 27.5 KB); six guardrails and an eval with Sonnet on the reference circuits (section "Keeping the catalog stable").
• Recommendation: first domains (FT-20), reference circuits (FT-21) and the report of what went unchecked (FT-19), then load steps in transients (FT-1), pin requirements from the parts entry (CK-13) together with a library of checked entries for the parts makers use most (PF), and the pin semantics (CK-1 to CK-3): they cover the most findings per line of code. Mains and high voltage as the domain mains, with the warning CK-22 as its first step (open decision 3).

Basis and limits

• Seven research agents, one cluster each: LEDs and displays (32 pitfalls), motors and robotics (35), power and batteries (32), MCU boards, buses, radio and keyboards (39), audio, mains, high voltage and measurement (35), guitar effect pedals (37), and the guitar's instrument, amplifiers, DI boxes and switching (39). Each pitfall comes with its numbers, the project types, the usual fix, what would detect it and its source. The guitar clusters read ElectroSmash's circuit analyses in a GitHub archive (mstratman/electrosmash-pdf-backups) and the FV-1, TL07x and CA3080 datasheets in GitHub repositories; most other guitar sources were search extracts (x).
• Sources are tagged r (read in the primary document), x (from a search engine's extract of the cited page) and v (to verify: the host was not reachable, or the value is from memory). The environment's network policy blocked most vendor hosts (ti.com, st.com, espressif.com, microchip.com, analog.com, nxp.com, pololu.com, adafruit.com and others); documents on GitHub were read directly – ESP-IDF, esptool, Espressif's hardware design guidelines, KiCad's tables of IEC 60664-1 and IPC-2221, QMK, ZMK, raspberrypi/hats, Marlin, Klipper, ArduPilot. A v value does not go into a catalog entry before it is read.
• The edakit side was read in the code, not only in the handbook: edakit/checks.py, edakit/rules.py, edakit/sim/run.py, edakit/sim/netlist.py, edakit/release/polarity.py.
• Out of scope by the owner's decision: the pinball machine (FlipperWiki).
• All 77 issues of the repository are closed, every one of them a proposal; none of the candidates below is pending.

The owner's projects

Project	Hardware	edakit	What of this concept applies
pytree	RP2350-Zero, 30 through-hole LEDs on three LED drivers, RGB star (common anode), magnetic buzzer, DIP switch, 3 × AA and USB-C ORed	reference project	most checks come from it; nothing new
automatic_christmas_bell, variants THT, SMD, ESP32	Nano carrier (ESP32 variant: LILYGO T-Display), servo, five LEDs, two buttons, USB	released	FT-1: D-3 simulates the servo as constant 0.3 A and 0.7 A, not its start; CK-11 for the ESP32 variant: copper of the carrier under the module's antenna
tinyjukebox, anTONIEa	Raspberry Pi 3, RC522 (SPI, 3.3 V) in the lid on a ribbon cable, I2S class-D amplifier SHIM with speaker, four buttons with 2.2 kΩ pull-ups, RGB LED with a shunt transistor, power bank behind switch S8	Fritzing drawing	CK-8 (bridge-tied output), FT-2 and HT-1 (SPI on a ribbon cable), E-3 i_min_a (power bank), E-2 contact current of S8; the README's warning about the speaker's magnet near the wiring is a reviewer question
AwesomeStudioPedal	NodeMCU-32S (ESP32-WROOM-32), five buttons or footswitches with 10 kΩ pull-ups, five LEDs, USB or a LiPo with regulator; nRF52840 Feather variant	wiring from CircuitSmith	G-12 with an ESP32 entry (finding below); LEDs and footswitches 30 cm and more from the board: FT-2, HT-1, CK-16; FT-4 and H-6 for the LiPo
asp_pro	concept: BLE-HID gateway for existing footswitches, ESP32 or nRF52840, MIDI out on DIN or TRS at 3.3 V, battery, to be sold	concept	purpose.use = "sale": P-11, G-24, G-25 as warnings; BL MIDI in/out, FT-6 optocoupler, HT-6 (TRS-MIDI type A or B), HT-7 (jacks that carry a supply: expression pedal), FT-4 charger, CK-16 outputs on jacks; battery transport and battery law in M-6
bustaferl	ESP32 DevKit, Waveshare 4.2″ e-paper on SPI, deep sleep, USB	firmware only	nothing new; the e-paper module is a black box (reviewer question exists); CS on GPIO 5 idles high, the strap's default level
system-engineering-fuer-kids	kits: tone generator, active speaker with amplifier, sound-book module	workshop, no PCB	HT-8, HT-13, HT-1 – the active speaker played the sound module's sound with no connection between the two (topic "Audio")
CircuitSmith (archived)	rule catalog E1–E22	taken over	E9, reverse protection at every power input, is opt-in in edakit (checks.reverse_protect): HT-6

Software only: FehlerJagd, agent-chat-ui, signal-backup-extract, AwesomeTaskSystem, test, antoniea-testbed, PartsLedger (concept), and the forks easyeda-mcp-pro, rfid-2-mqtt and rpi-mfrc522 (RC522 on a Raspberry Pi over SPI).

Finding outside edakit. AwesomeStudioPedal's docs/builders/BUILD_GUIDE.md asks for a 10 kΩ pull-up to 3V3 on every button pin, and src/esp32/include/builder_config.h puts button B on GPIO 12. GPIO 12 (MTDI) is a strapping pin of the ESP32: high at reset, it sets the flash supply VDD_SDIO to 1.8 V, and a WROOM-32 with 3.3 V flash does not boot (esptool, boot mode selection, r). Fix: another pin – GPIO 32 or 33 – or no external pull-up there, or the flash-voltage eFuse burnt. G-12 finds it, given an ESP32 entry whose GPIO 12 has strap: "low".

What edakit covers today

Area	Checks and hints
supply path	G-10 reverse polarity (topology), G-16 supply and ground pins, G-17 fuse against what it protects, H-1 bulk capacitor, H-8 two supplies on one net, E-3 regulators, minimum load, back-feed into sources and cells, E-5 plug, unplug and switch-over
levels and states	G-3 vmax, E-3 VIH/VIL and supply range where the entry gives them, G-4 floating inputs and gate/base resistors, G-5 one driver per net, G-12 strapping pins
stress and heat	G-8 voltage ratings, E-2 currents, power, LED current, GPIO current per pin and in total, junction and touch temperature
buses	G-11 I2C pull-ups, G-15 pin functions and peripheral instances, H-9 read-back edge
inductive loads	G-13 relay coils and magnetic buzzers on the board, H-4 a load at a connector
USB, radio	G-14 USB-C CC, P-11, G-24, G-25 radio statement, power and bands
board and files	L-1 to L-20, F-1 to F-6: copper from the Gerber files, clearances between pads, track width from the simulated current (L-5), hot loops (G-18, L-11), holes, silkscreen, manufacturing files

Gaps found in the code

Gap	Where	Consequence
G-4 looks at nets only; P-1 catches only supply pins in nc	checks.py:353, checks.py:1158	an unused CMOS input, a WS2815's BI pin or an A4988's RESET left nc passes
no open-drain or tri-state notion among the pin types	handbook 4.1	INT, ALERT, 1-Wire or a fan's tach without pull-up passes; two SPI devices sharing MISO fail G-5 unless typed bidirectional
transients vary the supplies only; loads are constant	sim/run.py:762	a brown-out from a servo starting or a Wi-Fi burst is not simulated
cells have a charge limit only	sim/run.py:545	a CR2032 at 20 mA passes
G-10 counts every capacitor as harmless on a reversed source	checks.py:137 (REVERSE_OK)	an electrolytic in front of the reverse protection is reverse-biased; params.polarized exists, only the polarity sheet reads it (release/polarity.py:105)
no check of a polarized capacitor's DC polarity in the simulation	sim/run.py	a coupling capacitor between stages or on ±12 V the wrong way round passes
G-11's upper bound is a fixed 10 kΩ	checks.py:140	400 kHz and the bus capacitance of a cable are not considered
fn knows pwm, adc, spi, i2c, uart	checks.py:144	no I2S, CAN, 1-Wire or USB
behaviour models: ldo, buck, boost, mcu, cc_sink, logic, current, none	handbook 5.1	op-amps, chargers, motor drivers and audio amplifiers are not simulated
part kinds without speaker, optocoupler, motor, heater	handbook 4.1	a speaker, a MIDI input's optocoupler, a motor on the board have no model and no check
no I2C address, ADC window, antenna area, or spacing for hazardous voltage	–	see CK-3, CK-6, CK-11, CK-9
G-8's catalog text names capacitors, diodes, transistors, MOSFETs, fuses, switches and connectors; the code also holds resistors, potentiometers, buttons, DIP switches and buzzers to v_max	checks.py:443, rules.json G-8	the catalog understates the check – a text fix

Topics in depth

The owner's questions during the analysis, each answered against the code.

Communication between ICs

edakit checks the electrical side of a link between ICs, not its protocol or its timing.

Link	Checked today	Missing
every link	G-15: a net's fn (spi.sck, spi.tx, spi.rx, spi.cs, i2c.sda, i2c.scl, uart.tx, uart.rx, pwm, adc) on a pin that has the function, the nets of one fn_bus on one peripheral instance; G-3 and E-3 levels between the parts; G-4 a select or enable input that only the MCU drives has a pull; G-5 two push-pull outputs on one net	fn for I2S, CAN, 1-Wire and USB (FT-15)
SPI	H-9 the read-back edge; the firmware contract: chain order (first_bit_to), read-back	MISO shared by several devices fails G-5 unless typed bidirectional, and a device that never releases MISO is not found (CK-2); SPI mode and maximum clock per device (FT-16); SPI on a ribbon cable (HT-1 – tinyjukebox's RC522 in the lid)
I2C	G-11 the pull-up window, the pull-ups of modules summed	addresses (CK-3); bus capacitance and speed (CK-3 with FT-2); the speed of level shifters (BL-9)
UART	G-5 TX on TX; G-4 RX on RX without a pull	RX on RX with a pull-up passes (CK-26); a 5 V TX into a 3.3 V RX is G-3
open-drain lines	I2C only	INT, ALERT, 1-Wire, a fan's tach, a wired-OR reset without pull-up (CK-2)
differential	–	RS-485 and CAN termination and bias (CK-17); USB-C D± in both rows (CK-12); USB series resistors (CK-13)

Cables and off-board parts

A cable is a connection between two boards (cables in the graph): r_ohm_per_wire gives the simulation its voltage drop, each board with a ground of its own, and i_max_a per wire holds G-17. A length, a capacitance, an inductance, coupling and ESD on the cable are not known; H-7 (ESD at connectors), H-4 (inductive load at a connector) and H-3 (debouncing) touch the subject.

The pedal's LEDs at 30 cm and more from the board show what matters:

1. DC is no issue: 30 cm of AWG26 is about 0.04 Ω – 0.4 mV at 10 mA.
2. Where the series resistor sits: on the board. Soldered to the LED's leg, a wire pinched against the enclosure shorts the GPIO to ground; nothing checks that today (CK-16, FT-8).
3. ESD through the LED's bezel and the footswitch reaches the GPIO over the wire; H-7 reminds, it checks nothing.
4. Emission, for a board that is sold (asp_pro): PWM edges on 30 cm wires radiate – slow edges at the source, the return wire next to the signal.
5. Inputs on long wires – the footswitches – catch interference: the build guide's 10 kΩ pull-ups are there for it, RC with a Schmitt-trigger input and debouncing (H-3) go with them.

Modelled today: the panel parts – LEDs, footswitches, jacks – on a board of their own with bought = true, joined by cables; the simulation then checks every LED's polarity and current, and H-7 sees the lines. bought means "no PCB" here, which the word does not say (FT-2). Candidates: FT-2, CK-16, FT-8, HT-1, CK-3 for I2C on a cable.

Audio: hum, transmitter, receiver

The simulation knows DC operating points and supply transients; frequency response, noise and coupling it does not know. There is no part kind for a speaker – the graph type audio maps to the kind buzzer – and no op-amp model, so an op-amp preamp is not simulated at all. G-6, G-8, E-2 and the hot-loop check (G-18, L-11, where a class-D amplifier's entry gives hot_loop) apply as anywhere.

Class	Mechanism	Decidable	Candidate
hum	ground loop between devices – amplifier, PC on USB, earthed equipment	no: it forms outside the board	HT-8
hum	returns of the output stage, LED PWM, a servo or a Wi-Fi burst through the input's reference ground	as a layout requirement	HT-8, reviewer
hum	ripple of switching loads on the amplifier's supply (PSRR)	partly, from the topology	HT-8
hum	an analog input without a DC path floats and catches 50 Hz	yes	CK-25
transmitter	filterless class-D (PAM8403, MAX98357A, PAM8302) drives a square wave of several hundred kHz onto the speaker leads	with the lead length and the datasheet's limit	HT-13, FT-2
transmitter	PWM and clock lines on cables	–	HT-1
receiver	RF rectified at the first junction of an audio input – the board's own Wi-Fi or BLE bursts, a mobile phone, the PWM speaker leads of a device nearby	partly: a board with a radio and an audio input without RC low-pass	HT-8
receiver	high-impedance inputs on long leads – guitar 1 MΩ, piezo, electret	–	HT-1
receiver	magnetic coupling – a speaker's magnet, coils next to leads and sensors (tinyjukebox's README)	no	reviewer

Beyond these: bridge-tied outputs (CK-8), DC on a speaker (FT-6), the pop at power-up (strap with G-12, CK-20), an op-amp's operating point (FT-3), outputs on jacks shorted while a plug goes in (CK-16, HT-7); level, impedance and frequency response only with AC analysis (FT-12).

The owner's case – the steering wheel of system-engineering-fuer-kids. The active speaker (a DIY kit from Conrad) played the sound module's donkey with no connection between the two: each ran from a battery of its own, no wire, no common ground. Shorter and shielded cables – and luck – made it work. Without a conductive path only fields couple:

• electric (capacitive): sound-book modules drive their speaker directly, often with PWM or a bridge output, so the module's speaker leads swing with steep edges; an input cable lying next to them picks up a few pF of it, and the amplifier's first junction rectifies the carrier back into audio;
• magnetic (inductive): the audio current in the module's speaker coil and leads induces a voltage in the loop of the amplifier's input cable – the larger the loop and the closer, the more.

A shielded cable stops the electric field, a shorter and twisted one has less loop for the magnetic one; distance helps both. Separate batteries do not help at all. What edakit can decide sits on each side alone: on the receiver an input without RC low-pass or ferrite (HT-8), on the transmitter a PWM speaker output on leads (HT-13), and cables as such (HT-1, FT-2). The coupling between two devices that share nothing is no check: it stays judgement.

Guitar electronics

Two more research clusters: effect pedals (37 pitfalls) and the instrument, amplifiers, DI boxes and switching (39).

• The audio band at its operating point only. Most of it is analog in the audio band, which edakit's DC-first simulation sees at its operating point. The bias of discrete stages – fuzz, boosts, phasers – rests on the spread of the transistors (AC128 gain 45–165, r; J201 IDSS 0.2–1.0 mA, r); the simulation has corners for the LEDs' forward voltage only (FT-17).
• Impedance is the guitar's main topic. A passive pickup wants about 1 MΩ, a piezo 10 MΩ and more; LM386 inputs, codecs and passive DI boxes load it to 10–300 kΩ (CK-27, FT-18).
• Op-amps at 9 V run out of swing and common-mode range (TL07x: 1.5–3 V short of each rail, r), and the Klon's charge pump puts 27 V across one – 36 V from a 12 V outlet (r): FT-3, and E-3 with the op-amp's vdd_op_max and a 12 V supply corner.
• Pops and scratches are DC where it should not be: across switches, relay contacts and pots, and polarized capacitors around the 4.5 V reference in every state (CK-30, CK-7).
• The pedalboard is a system: outlets with a current each, positive-ground and AC pedals, daisy chains as ground loops (HT-18), charge pumps and PWM in the audio band (HT-16).
• Layout: input and output of a high-gain stage apart, a charge pump or a converter's switch node away from inputs, the PT2399's two grounds joined short (CK-31).
• Covered already: pinouts of regulators and transistors (P-1, L-10: 78L05 against 7805, BS170 against 2N7000), a regulator's temperature (E-3), capacitor ratings at 18 V (G-8 with the supply corner), flyback diodes and pull-in of relays (G-13, E-3), reverse polarity (G-10).
• Out of scope: tube amplifiers – B+ of 250–500 V, heaters, output transformers, and tube models that do not hold at low anode voltage – unless edakit takes hazardous voltage in (open decision 3).

Candidates

The kinds are those of add-rule: a check where a program decides without guessing, a hint where it takes judgement, a block for a circuit worth reusing, a part fact for one part's datasheet, and a feature where edakit lacks the capability (add-feature). "Seen" names the clusters of the pitfalls list: LED, MOT (motors), PWR (power), MCU, AUD (audio, mains, high voltage, measurement), GTP (guitar pedals), GTI (guitar instrument, amplifiers, DI, switching), OWN (the owner's projects).

Features

ID	What	Seen
FT-1	Load steps in transients. An external load, an MCU's i_core_a and the outputs of a current model as [t, A] points in a run: a servo starting or stalling, a Wi-Fi burst (20 → 350 mA within 1 µs for 1–5 ms), a motor's start current V/R, a radio pulse from a coin cell. E-5's rule that an MCU's supply stays above vin_min then finds the brown-out.	MOT 16, 35; PWR 6, 18; MCU 5, 31, 32; AUD 3; LED 8; OWN bell
FT-2	Cables with length and gauge. length_m and awg or mm2 give r_ohm_per_wire (copper at temperature) and i_max_a; optional c_pf_per_m and l_nh_per_m feed CK-3's bus capacitance, the LC spike at hot-plug (FT-1, E-5) and HT-1. Hand-wired panel parts get a board without a PCB – today that is bought = true, a misleading word for it.	LED 4, 9, 10; MOT 9, 31; AUD 28–30; MCU 16; OWN pedal
FT-3	Op-amp model. Output limited to the rails minus the swing of the datasheet, input common-mode range; E-3 holds both.	AUD 6; MCU 23, 38
FT-4	Charger model. Linear Li-ion charger: constant current from R_PROG (TP4056: 1200 V / R_PROG, x), 4.2 V, termination current; E-3 already holds every cell to its i_charge_max_a, E-2 the charger's temperature.	PWR 1–4; OWN asp_pro
FT-5	Adjustable regulators and enable. Output from the feedback divider and vref_v instead of a declared vout_v; a regulator's minimum load (LM317 10 mA, x); a boost converter's output with EN low (VIN − VD).	PWR 13–15
FT-6	Part kinds speaker, optocoupler, motor, heater, and coil_ohm on any inductive part. Speaker: impedance, power, no DC (CK-8). Optocoupler: its LED's current at the driver's level (a PZEM's 1 kΩ at 3.3 V, x; MIDI input). Motor: winding resistance and start current. Heater: resistance, so V²/R.	AUD 1, 2, 23; MCU 30; MOT 1, 4, 21, 25–28
FT-7	Voltage domains. Nets above the ES1 limits of IEC 62368-1 (60 V DC, 30 V RMS AC, v) are hazardous; nets with a galvanic path from L or N are mains. The basis of CK-9, CK-10, CK-22, CK-24.	AUD 14–18, 26, 35; LED 30, 31
FT-8	Generated fault scenarios. A reversed supply as a DC scenario (beyond G-10's topology: shunt-diode designs, a swapped Eurorack ribbon), each wire of a cable shorted to ground at its far end, one supply off while the others are on (CK-5).	AUD 10, 11; PWR 20; MOT 15
FT-9	DC bias of ceramic capacitors. Capacitance against voltage in the entry; the effective value counts in CK-14 and in transients (4.7 µF, 6.3 V, 0603 X5R at 5 V: about 0.33 µF, x).	PWR 16
FT-10	Analog constants in the firmware contract. Divider ratios, pull-up values (a thermistor's 4.7 kΩ that Marlin and Klipper assume, r), sense resistors and ADC channels into edakit firmware, with drift reported.	MOT 29; MCU 22; PWR 17
FT-11	Expectations for DC loads. expect per load as transients have it: a source's current in the load "sleep" against a budget – evidence for a battery-life requirement.	PWR 10, 17
FT-12	AC analysis (ngspice .ac, .noise) with expected gain and corner frequency. Open decision 5: it goes against edakit's DC first.	AUD 7
FT-13	LED strip model with resistance per segment and injection points.	LED 9–12
FT-14	Constant-current LED buck (PT4115, AL8860): output current V_sense / R_S (PT4115: 0.1 V, x), headroom against the LEDs' forward voltage.	LED 19
FT-15	fn for I2S, CAN, 1-Wire and USB (i2s.bclk, i2s.ws, i2s.sd, can.tx, can.rx, onewire, usb.dp, usb.dn): G-15 then holds them to the pin functions as it does SPI, I2C and UART.	topic "Communication between ICs"
FT-16	Bus parameters in the firmware contract: per SPI device its mode and maximum clock, per I2C device its address and the bus speed its pull-ups allow, from the parts entries; drift reported as for the pins.	topic "Communication between ICs"; H-9
FT-17	Corners of transistor parameters: hFE, leakage (doubling per 10 °C), IDSS and VGS(off), VGS(th) – min and max from the entry as corners of the simulation, as the LEDs' forward voltage is today; with expected windows on nets (FT-11) the bias of fuzz, boost and phaser stages is checked.	GTP 1, 5, 6, 12, 13, 18
FT-18	Source models for the audio band: a passive pickup (inductance, resistance, capacitance), a piezo (voltage source behind its capacitance), a cable's capacitance per metre – for FT-12 and the loading of CK-27.	GTI 1–5, 10, 12; GTP 3
FT-19	Coverage report: per check what it could not check for lack of data – "E-3: U4.3, U4.5 without thresholds" – in edakit check and the gate protocol. The first step before any check that depends on part facts (section "Keeping the catalog stable").	stability
FT-20	Domains: [project] domains – audio, motors, radio, mains – switch domain checks and hints on; edakit rules and edakit hints show what applies to the project.	stability
FT-21	Reference circuits per domain: well-known open designs as edakit graphs in the test fixtures, beside tests/fixtures/led-bar – an ESP32 DevKitC, a WLED controller, a QMK keyboard, a Tube Screamer and a Fuzz Face, a relay board with an HLK-PM01, a LiPo charger. add-rule's run against real projects (step 5) uses them for every domain none of the owner's projects has.	stability
FT-22	Short rule listing: edakit rules prints ID and title, --long the rule sentences; today the listing is 27.5 KB because it prints every rule sentence.	model stability
FT-23	The graph by query: edakit graph show <designator or net> gives a part with its nets or a net with its pins, so that a session reads a slice instead of the whole graph (65 KB in the reference project); with the circuit language of circuit-language.md the whole is half the size.	model stability
FT-24	Eval with Sonnet: typical tasks on the reference circuits (FT-21) as an eval suite, run with Sonnet and with Opus on every change to skills, prompts and the catalog.	model stability

Checks

ID	Stage	What it tests	Needs	Seen
CK-1	G	Unused inputs are not left open: an input pin in nc is an error unless its entry gives an internal pull or allows it to float.	float_ok on a pin	MCU 38; LED 13; MOT 14; gap above
CK-2	G	Open-drain and tri-state outputs: a net whose drivers are all open-drain has a pull-up (G-11 does it for I2C only); open-drain outputs and tri-state outputs on a shared line do not clash in G-5; an output that never releases the line on a shared line does.	open_drain, tristate on pins	MCU 20, 21, 34, 39; MOT 23, 32
CK-3	G	Unique addresses on a bus: each device's address from its entry and the levels of its address pins; two equal ones on one bus are an error. Covers I2C (DS3231 and MPU-6050 at 0x68; PCF8574 0x20–0x27 against PCF8574A 0x38–0x3F; PCA9685's All-Call 0x70) and addressed UARTs (TMC2209: MS1/MS2). With FT-2, G-11's upper bound from the bus capacitance: Rp ≤ tr / (0.8473 · Cb) – 2.95 kΩ at 100 kHz and 885 Ω at 400 kHz for 400 pF (UM10204, r).	address with pins and extra addresses	MCU 15–17; LED 27, 28; MOT 14
CK-4	E	MOSFET switch states: on, the gate drive the circuit gives reaches a VGS at which the entry specifies RDS(on) (IRF520: only at 10 V, v); off, a P-MOSFET's gate reaches its source within |VGS(th)| min (a 3.3 V GPIO cannot switch off a P-MOSFET on 5 V or 12 V).	rds_on: [[vgs, ohm], …], vgs_th	LED 17; MOT 25, 26; PWR 30
CK-5	E	No input above its own supply: every input stays within its part's supply plus the margin of the absolute maximum, also while that part's supply is off, and the current into its clamp within the injection limit (WS2812B DIN ≤ VDD + 0.5 V, x; ±2 mA in TI SLAA689, x).	vin_rel_max_v, i_inject_max_a; FT-8	LED 5; PWR 20; MOT 15; MCU 26
CK-6	E	ADC inputs: the voltage at a pin of a net with fn: adc lies within the ADC's window in every scenario (ESP32 at 11/12 dB: 150–2450 mV, r; ESP8266 0–1 V, r), and the source resistance the pin sees is within the ADC's limit unless a hold capacitor sits at the pin (ATmega328P 10 kΩ, x; ESP32 0.1 µF, r).	adc: {range_v, r_source_max_ohm, c_hold_min_f}	PWR 17; MOT 19, 29; MCU 22, 23; AUD 6, 34
CK-7	G, E	Polarized capacitors: G-10 no longer counts a capacitor with params.polarized as harmless; E-2 holds V(+) − V(−) in every scenario to the reverse voltage the series allows.	pin roles plus, minus on polarized capacitors	AUD 7, 10, 11; gap above
CK-8	G	Bridge-tied outputs: neither pin of a BTL pair on ground, on a net with another driver or another channel's pin, or on another part's input; the load between them is a speaker.	roles out_p, out_n, btl; FT-6	AUD 1; OWN tinyjukebox
CK-9	L	Spacing of hazardous voltage: between domains, clearance and creepage of IEC 60664-1 – at 230 V, overvoltage category II, pollution degree 2: clearance 1.5 mm basic and 3.0 mm reinforced, creepage 2.5 mm basic and 5.0 mm reinforced, a slot counting from 1 mm width (KiCad's tables, r); within a domain IPC-2221 by peak voltage – 230 V RMS is 325 V peak, 2.5 mm between uncoated outer conductors (r); a Nixie's 170 V: 1.25 mm (r).	FT-7	AUD 14, 15, 26, 35; LED 31
CK-10	G	Isolation and mains input: a part with pins in two domains states its isolation (basic or reinforced, V RMS); an off-board connector in a non-isolated mains domain is an error; the first series part from L is a fuse, MOV and X capacitor sit behind it, an X capacitor above 0.1 µF has a bleeder, a capacitor across the barrier is Y-class.	FT-7; isolation, safety_class	AUD 16–18
CK-11	L	Antenna keep-out: no copper, via or part within the antenna area a module's entry gives, on any layer, and the antenna at or over the board edge (Espressif: over the edge, or ≥ 15 mm clear around it, r); for a module plugged over a carrier, the carrier's copper counts.	antenna: {keepout, edge}	MCU 6; OWN bell ESP32 variant
CK-12	G	USB-C data in both orientations: a receptacle's D+ (A6, B6) and D− (A7, B7) are joined in pairs, or none is used; extends G-14.	roles dp, dn per contact	MCU 9
CK-13	G	What a pin needs from the circuit, from the parts entry: a capacitor to ground in a range, a pull-up or pull-down in a range, a series resistor, "to ground" or "never to ground". One mechanism for ESP32 EN 10 kΩ and 1 µF (r), STM32 VCAP, BOOT0, NRST, VDDA, VBAT (AN4488, r), ATmega AVCC and AREF, USB series resistors (RP2040 27 Ω, r) and the D+ pull-up, RP2040 QSPI_SS, DS18B20 DQ 4.7 kΩ (r), MAX6675 T− to ground against MAX31855 T− never to ground (x), a regulator's output capacitor.	needs on a pin	MCU 1, 2, 8, 10–14, 21; AUD 33; PWR 9
CK-14	G	Capacitors a regulator needs for stability: input and output capacitance at least the datasheet's minimum, after DC bias, and the ESR within its window (AMS1117: 22 µF tantalum, x). Can be part of CK-13.	c_out_min_f, esr_ohm; FT-9	PWR 9, 16; MCU 5
CK-15	G	Inductive loads beyond relays and buzzers: G-13 for every part with coil_ohm and the kinds motor and solenoid; an IC's bridge output without internal clamp diodes (L298, v) at a connector has diodes to VS and GND.	FT-6; clamp: none	MOT 2, 21
CK-16	G	Pins that leave the board reach the connector through a series resistance on their own board, or the part is rated for it – a short on the wire or an ESD strike then does not hit the pin directly (emonTx: 1 kΩ and TVS, r). Warning; waive for USB, I2C and the like.	short_proof	AUD 28; MOT 31; OWN pedal
CK-17	G, E	Termination of RS-485 and CAN: exactly two 120 Ω on the bus across all boards, RS-485 fail-safe bias where the receivers need it (idle ≥ 200 mV).	roles a, b, canh, canl	MCU 28, 29; AUD 30
CK-18	E	Cell current: every cell within its continuous discharge current in every scenario and within its pulse current in transients (CR2032: 3 mA continuous, x).	i_max_a per cell state	PWR 6; gap above
CK-19	G	Keyboard matrix: every switch has a diode, all pairs point the same way to rows and columns, and as the firmware contract says (QMK COL2ROW; ZMK defaults to row2col, r).	roles row, column	MCU 33
CK-20	E	Configuration pins with voltage windows: strap as a table of windows, checked in the operating point with the internal pull (MAX98357A SD_MODE: < 0.16 V off, up to 0.77 V mix, up to 1.4 V right, above left, r; CH224K asks for 20 V with CFG1 open, x); extends G-12.	strap as windows	AUD 5; PWR 27
CK-21	E	Supply-pin current of logic ICs: the sum of the output currents within the supply pin's limit, as E-2 does it for MCUs (74HC595: 70 mA through VCC or GND, x).	i_supply_pin_max_a	LED 23
CK-22	G	Hazardous voltage without a spacing check: a net above the ES1 limits is a warning that edakit does not check creepage and clearance – the first step before FT-7 and CK-9.	net v	AUD 14–27; LED 30, 31
CK-23	E	MCU clock against its supply: the maximum clock at the lowest supply from the entry's curve (ATmega328P at 16 MHz needs about 3.78 V, from 10 MHz at 2.7 V and 20 MHz at 4.5 V, v).	f_max_vs_v, the crystal's frequency	PWR 19; MOT 35
CK-24	G	Capacitor banks above the hazardous limit discharge: a resistor across them brings them below 60 V within 60 s.	FT-7	AUD 27
CK-25	G	Analog nodes have a DC path: a net with an analog input that reaches no resistor or driver to a defined level – only capacitors – floats: it picks up hum, and a switch on it pops. The analog sibling of G-4.	–	AUD 13
CK-26	G	UART lines cross: a net with fn uart.tx or uart.rx has exactly one output (a TX) and at least one input (an RX); a net with inputs only is an error, also with a pull-up – G-4 lets it pass then.	–	MCU 14; topic "Communication between ICs"
CK-27	G	Instrument inputs load their source within its limit: a connector input declared for a passive pickup (guideline 1 MΩ, x) or a piezo (10 MΩ and more, x) sees the shunt resistors and the first active input's impedance in parallel above that; tuners and other branches on the node count.	z_in_ohm on input pins; source on the connector pin	GTI 1, 10, 15, 33, 34; GTP 3, 14, 19, 26
CK-28	G	Amplifiers inside their stable gain: the closed-loop gain from the feedback network is at least the part's minimum stable gain (LM1875: 10, x), and a part that needs external compensation has its capacitor (LM308: 30 pF across pins 1 and 8, r).	gain_min, requires_ext_comp	GTI 19; GTP 11
CK-29	G	Series capacitors share their voltage: two or more capacitors in series across a supply have balancing resistors (2 × 450 V in a tube supply, supercapacitor stacks).	–	GTI 21; PWR 23
CK-30	E	No DC where signals are switched: both sides of a switch, relay contact or pot in a signal path sit at the same DC level within a few mV in every state (the pop, the scratchy pot), and a contact that carries signal currents is rated for them (gold, minimum load 10 µA / 1 mV, x).	nets marked as signal path; min_load of contacts	GTP 31, 37; GTI 31; AUD 13
CK-31	L	Nets kept apart: pairs the hand-over names (layout.apart: two nets, a distance, a parallel length) keep them on the board – input and output of a high-gain stage (2.5 mm, x), a charge pump or switch node against high-impedance inputs, a crystal against noisy lines. The same measurement serves CK-9 and CK-11.	layout.apart	GTP 10, 22, 23, 25, 33; GTI 16, 26, 29; MCU 6, 7

Hints

ID	When	About	Seen
HT-1	a line on a cable (with FT-2: from a length on)	inputs: pull-up of 10 kΩ or less, RC with a Schmitt-trigger input, TVS; outputs: series resistor at the source, slow edges; data and clocks: series termination, return wire beside the signal, twisted pair	MOT 31; AUD 28–30; LED 4; MCU 16; OWN pedal
HT-2	a connector with a vbus pin	USB current budget: 500 mA (USB 2.0), 900 mA (USB 3), 1.5 A or 3 A only when the CC advertisement is read (x)	PWR 25; LED 7, 8; MCU 35
HT-3	more than 10 µF on VBUS ahead of a soft start	USB 2.0 §7.2.4.1 limits a device to 10 µF at attach (x) – against H-1's bulk capacitor; the hint names both	PWR 28
HT-4	a boost converter with EN on a GPIO	without load disconnect its output stays at VIN − VD with EN low (MT3608, x)	PWR 15
HT-5	a charger whose BAT net feeds the load	without a power path the charge never terminates (Microchip AN1149, x)	PWR 2
HT-6	a power input that is not keyed and not in checks.reverse_protect	no polarity standard: JST-PH batteries differ between vendors (x), barrel jacks are centre-positive or, after Boss, centre-negative; also TRS-MIDI type A or B	PWR 7, 32; AUD 11; CircuitSmith E9
HT-7	a jack whose pin carries a supply	plugging shorts tip, ring and sleeve: series resistor or PTC, no hot-plug (QMK on TRRS links, r)	MCU 34; OWN asp_pro
HT-8	an audio input or output	ground loops through USB or mains earth, returns of large currents under the input's reference, the amplifier's supply shared with switching loads or a radio (filter), RF rectified at inputs on a board with a radio (RC low-pass, ferrite)	AUD 6, 8; OWN horn
HT-9	pwm_hz on a net that switches LEDs	flicker after IEEE 1789: 100 % modulation needs ≥ 1250 Hz for low risk and ≥ 3 kHz for no observable effect; Arduino's 490 Hz allows 39 % (x)	LED 18
HT-10	a heater (FT-6) switched by one MOSFET or SSR	a second, independent interruption: thermal fuse, relay, PSU enable (Marlin, Klipper, Voron, r)	MOT 30
HT-11	a part with rth_ja_k_per_w	Rth(j-a) holds for the datasheet's copper; the layout must give it (hand-over)	MOT 11; PWR 3, 12; LED 20
HT-12	a motor driver	bulk capacitor at VM per driver (Pololu: ≥ 47 µF, x); a series diode in front of VM blocks the motor's returned energy	MOT 5, 9
HT-13	a filterless class-D output at a connector	ferrite or LC filter above the datasheet's lead length – a check with FT-2 and the length in the entry	AUD 1–3; OWN horn
HT-14	a relay whose contacts switch an off-board load	the rating is resistive; LED drivers and switch-mode supplies draw tens of amps at start, inductive loads need a snubber	AUD 20
HT-15	a battery whose minus reaches ground through a jack's ring	a cable left plugged drains it, a TRS cable leaves the circuit off; the DC jack's switch lug must break battery plus (EMG: about 80 µA per pickup, x)	GTI 8; GTP 32
HT-16	a charge pump, a converter with a light-load mode, or PWM on a board with an audio path	switching in the audio band: ICL7660 and TC1044 at about 10 kHz with a 5 kHz subharmonic (x), boost converters bursting at light load, LED and tap-tempo PWM – above 20 kHz or filtered	GTI 29, 38; GTP 17, 34; AUD 12
HT-17	a class-2 ceramic (X7R, Y5V) in a signal path	it distorts; C0G or film (r)	GTI 36; GTP 24
HT-18	a pedal's supply input	positive-ground and AC pedals need an isolated outlet; a series part in the DC jack's ground leg is bypassed by the audio cables; daisy chains form ground loops	GTP 2, 35; GTI 32, 38
HT-19	an instrument's string ground, a capacitor from mains to chassis	the bridge ground ties the player to the amplifier's ground: a line-to-chassis capacitor (death cap) is Y-class or gone – the amplifier's side is CK-10	GTI 7, 27

Blocks

ID	Content	Seen
BL-1	addressable LED input: 74AHCT1G125 at 5 V, 330 Ω series resistor, 1000 µF	LED 1–6
BL-2	logic-level N-MOSFET low-side switch (AO3400 class, specified at 2.5 V) with gate resistor, pull-down and an optional flyback diode – npn-switch stops at about 100 mA	LED 17; MOT 25
BL-3	high-side P-MOSFET switch with an NPN or N-MOSFET level shift	PWR 30
BL-4	Li-ion charger with protection (MCP73831 or TP4056, DW01A, FS8205) and load sharing	PWR 1–5
BL-5	3.3 V LDO with low quiescent current and its capacitors, rated ≥ 500 mA for an ESP32	PWR 8–11; MCU 5
BL-6	battery voltage measurement: switched divider with hold capacitor	PWR 17
BL-7	ESP32 module support: EN 10 kΩ and 1 µF, BOOT button, auto-program with two NPN, 10 µF and 0.1 µF	MCU 1, 2, 5
BL-8	USB-C with data: D± of both rows joined, USBLC6-2, 5.1 kΩ per CC pin, polyfuse	MCU 8, 9
BL-9	I2C level shift (BSS138 at 100 kHz, or PCA9306)	MCU 18, 19
BL-10	RS-485 at 3.3 V (MAX3485 class) with switchable termination, bias, DE pull-down	MCU 28
BL-11	MIDI in (6N138, 220 Ω, 1N4148, pin 2 open) and MIDI out (220 Ω/220 Ω at 5 V, 33 Ω/10 Ω at 3.3 V, v)	MCU 30; OWN asp_pro
BL-12	cable input: pull-up, RC, TVS	AUD 28, 29; MOT 31
BL-13	servo connector with bulk capacitor	MOT 16
BL-14	DC motor driver (DRV8833 or TB6612) with VM capacitors	MOT 1–5
BL-15	Eurorack power entry: keyed 10- or 16-pin header, series Schottky per rail, 10 µF	AUD 10
BL-16	instrument buffer: 1 MΩ JFET or op-amp input, 100 Ω–1 kΩ output resistor, RF filter at the jack (series resistor, then 47–100 pF)	GTI 1, 2, 39; GTP 14
BL-17	pedal power entry: 9 V centre-negative jack, series Schottky or P-MOSFET, 100 Ω and 100 µF, 4.5 V reference with ≥ 10 µF	GTP 9, 34, 36; AUD 11
BL-18	relay bypass with mute: signal relay with gold contacts, transistor coil drive with flyback diode, photo-MOS mute window (35 ms, r)	GTP 31; GTI 30, 31

Part facts

The checks of the catalog fire only where entries carry the datasheet: strap, functions, vih_ratio, vdd_op_min, rth_ja_k_per_w, the GPIO limits. A library of checked entries for the parts makers use most turns on G-12, G-15, E-2 and E-3 without a line of code – and then CK-3, CK-6, CK-13 and CK-20 where they are built:

• modules: ESP32-WROOM-32, ESP32-S3 and ESP32-C3 modules (straps, input-only and flash pins, ADC2, peak current, radio, antenna area), NodeMCU-32S, Wemos D1 mini, Arduino Nano, RP2040 and RP2350 boards (the Pico 2 is in the example);
• LEDs and displays: WS2812B per revision (VIH 0.7, 0.63 or 0.55 × VDD, x), SK6812, APA102, SSD1306 module, PCF8574 and PCF8574A backpacks, MAX7219;
• power: AMS1117-3.3, a low-quiescent LDO, TP4056, MCP73831, DW01A with FS8205, MT3608, CH224K;
• switching and drivers: AO3400, IRLZ44N, IRF520 (as the counter-example), DRV8833, TB6612, L298N, A4988, DRV8825, TMC2209;
• sensors and buses: DS3231 (with the module's AT24C32), MPU-6050, BME280, INA219, INA226, DS18B20, HX711, ACS712, HC-SR04, MAX31855, MAX6675, CH340, CP2102, MAX485, MAX3485;
• audio: MAX98357A, PAM8403, INMP441.

Reviewer questions

What stays judgement goes into the questions of the reviewers, as "Not covered by a check yet" does today:

• electrical-reviewer – audio and analog: ground loops, shared returns under references, supply filtering for amplifiers and microphones, RF rectification at inputs; cables: length, what is on them, what a short or an ESD strike at the far end reaches;
• pcb-reviewer – RF: antenna areas of modules and of modules plugged over the board; hazardous voltage: spacing between domains, slots; thermal pads: copper and vias under exposed pads.

Keeping the catalog stable

The owner's questions: with all these rules, does edakit stay stable, or does it confuse more than it helps – and does the context of a Claude session become the limit?

Stability analysis

Dimension	Today (measured 2026-10-30)	Risk if the candidates were built one by one	Guardrail
false alarms	the bell: 0–1 waived finding and 0–1 accepted warning per variant (19 parts); the reference project: 7 waived findings for 84 parts and 77 nets, one accepted warning	most candidates come from research: add-rule's run against real projects (step 5) cannot show their false alarms while no project of the owner has audio, mains or a guitar circuit – a check that guesses produces waivers	4, 5
silent passes	checks that need data pass without it, unreported: E-3 checks only inputs with thresholds, G-12 only pins with strap	more such checks are more coverage on paper	3
conflicts and overlaps	one known conflict: H-1's bulk capacitor at a power input against USB's 10 µF at attach (HT-3); overlaps G-4 and CK-1, G-11 with CK-2 and CK-3, G-12 and CK-20, G-14 and CK-12, H-7 and CK-16	two messages that contradict each other, one fault found twice	2
tests	983 tests in 16,700 lines; every check seen red once; the example project stays clean	none in itself – but a negative test cannot show a false alarm	5
maintenance	23,900 lines of Python; a catalog of 101 KB, 1.25 KB per entry (median); the handbook repeats every check in a table	catalog and tables about twice as large; more fields in every parts entry, each one to sign (G-9)	2
run time	edakit check under a second, edakit sim seconds	load steps and fault scenarios add runs	they run only where sim.json asks for them
the model (also Sonnet)	per task a few ten thousand tokens; always loaded 11.4 KB (skill and agent descriptions 9.9 KB, SessionStart hook 1.5 KB); edakit rules prints 27.5 KB, the reference project's graph is 65 KB; skills average 33 words per sentence – section "Stability of the model"	five places grow with the catalog: the rules listing, the handbook's tables, the reviewers' questions, the hints, the fields of parts entries; dense text and whole files read where a slice would do weigh more for a smaller model than the size itself	1; L1–L6
process	77 proposals, all accepted – the triage has never rejected one	research-based proposals are the first real test of "rejected" and "question"	5

Verdict: stable today, and the context is no immediate limit – what every session loads does not grow with the catalog. Built one by one, the candidates would make it worse in four places at once: false alarms without a project to find them, silent passes, overlapping messages, and files read whole. The guardrails keep it stable.

Guardrails

edakit is open source: its rules are meant for the boards of everybody who uses it, not only for the owner's. A guardrail therefore never waits for one of the owner's projects; it keeps a growing catalog trustworthy.

How the catalog itself is restructured so that it scales – engines, data, domain profiles, engine-level tests – is the concept rules-at-scale.md; the guardrails below are its ground rules.

Decided by the owner:

1. Domains (FT-20). A project states its domains – audio, motors, radio, mains, guitar; domain checks, hints and the reviewers' questions run only there, and edakit rules and edakit hints show what applies. A board without audio never sees an audio hint, and a session never reads rules it does not need.
2. Condensed rules. Seven mechanisms cover 91 of the 128 maker pitfalls of the first five clusters that are not fully covered: requirements from parts entries (38: CK-13, CK-14, CK-20, CK-12, CK-21, CK-23, CK-3, PF), loads over time (16: FT-1, FT-8, FT-11, CK-18), analog windows (13: CK-5, CK-6, CK-25, FT-3), pin semantics (10: CK-1, CK-2, CK-26), cables (10: FT-2, CK-16, CK-17, HT-1), hazardous voltage and RF geometry (9: FT-7, CK-9, CK-10, CK-11, CK-22, CK-24), switching parts and polarity (7: CK-4, CK-7, CK-8, CK-15). The guitar clusters fall into the same mechanisms, plus CK-27 and CK-30. The knowledge goes into parts entries as data, as strap and pull do today – not into new check IDs.

Proposed, for the owner's decision:

1. Say what was not checked (FT-19). A check that needs a datasheet value skips a part without it and stays silent: G-12 finds the GPIO 12 fault of the owner's pedal only if the ESP32's entry has strap; without it the check passes, and nobody sees that it never looked. The report names it – "G-12: U1 without strap data, not checked" – in edakit check and the gate protocol. The more checks rest on part facts, the more this matters.
2. New checks as warnings first. A new check has been seen red once and has run on the test projects; on a board nobody tested it against it can still be wrong. As a warning it shows its finding without stopping a release – the user fixes the board or accepts the finding with a reason –; once it has run on several boards without a false alarm it becomes an error. P-11, G-24 and G-25 are graded by the board's purpose this way already.
3. Reference circuits per domain (FT-21). add-rule runs a new check against real projects before its merge (step 5), so that false alarms show before anybody else meets them. For the domains none of the owner's projects has, well-known open designs become edakit graphs in the test fixtures, and each domain's checks run against them.

Stability of the model

The owner's requirement: edakit runs reliably with Sonnet, not only with the largest model. What lands in a session's context, measured on the bell at bf7450a:

What	Size	When
descriptions of the 12 skills and 7 agents	9.9 KB, about 2,500 tokens	always, in every session with the plugin
SessionStart hook: the working agreements	1.5 KB, about 400 tokens	always, in an edakit project
one skill	2.7–15 KB (new-project 15, pcb-round 13.7, circuit-design 10.8)	when it runs; a circuit change runs start, hardware-change and part-search: about 21 KB
edakit check	1.5 KB	per run
edakit hints	1.9 KB	per run
edakit sim	about 6 KB: a row per scenario and the failures	per run
edakit rules, the listing	27.5 KB, about 7,000 tokens: ID, title and the whole rule sentence of every check, lines up to 1,300 characters	the electrical reviewer and propose-rule call it
edakit rules G-11	1.1 KB	per ID
the graph	the bell 9.3 KB; the reference project 65 KB, about 16,000 tokens	every circuit change and every review reads it
a parts entry	median 1.3 KB, up to 8.9 KB	per part
a handbook chapter	34–80 KB	when a skill sends the model there
datasheets	PDFs, often the largest item	the reviewers and the part selector, each in a context of its own

The density. The skills average 33 words per sentence and a quarter of their sentences are longer than 40 words, with 97 references into the handbook; the agents 38 words and 30 percent; the handbook 36 words, 30 percent and 340 references.

What it means. A task stays at a few ten thousand tokens – no problem of size for Sonnet. The risk for a smaller model lies elsewhere: dense sentences that carry a rule, its exceptions and a reference at once; whole files read where a slice would do – the rules listing, a 65 KB graph, a handbook chapter –; and knowledge the model applies itself where no code checks it, above all the fields of a parts entry taken from a datasheet.

What more rules do to it. A check in code costs the model nothing until it fails – then its entry, about 1 KB. Growth reaches the model in five places: the edakit rules listing (about 360 bytes per check, 55 KB at twice the catalog), the handbook's check tables, the reviewers' question lists, the hints, and the fields of parts entries – condensing rules moves knowledge into parts entries, which an agent fills from datasheets.

Guardrails for the model:

• L1 Code decides, the model orchestrates. As today – findings come only from code; no prompt carries rule knowledge for the model to apply by itself.
• L2 Bounded output. edakit rules lists ID and title (FT-22: about 2 KB instead of 27.5); the text comes by ID. Every command prints failures, open points and a summary; details go into files.
• L3 Domains filter what the model sees: rules, hints, the reviewers' questions, the fields of parts entries (guardrail 1).
• L4 Parts entries from templates: per kind and domain only its fields; edakit part checks units and ranges; checked library entries (PF) for common parts, so that less is taken from datasheets by the model.
• L5 Short texts: one task per skill, numbered steps, short sentences (under 25 words as a target, h); the handbook read by section, not by chapter; the circuit in the compact form of circuit-language.md – half the JSON's bytes for the bell – and slices of the graph by query (FT-23) instead of 65 KB read whole.
• L6 Measured with Sonnet (FT-24). The reference circuits (FT-21) are also the test bench: typical tasks – add an LED, swap a part, find a seeded fault, review a board – run with Sonnet and with Opus. A run passes when the checks end green, the right commands were used and no finding was invented; every change to skills, prompts and the catalog runs the suite (Claude Code's claude plugin eval runs such suites).

Priority

Ordered by how often the pitfall appeared across the clusters, what it costs when it happens, and how much it takes – within the guardrails above:

#	Candidate	Why first
0a	FT-20 domains, FT-19 report of what went unchecked – the restructuring of rules-at-scale.md	the frame every check runs in: domains keep rules off boards that do not need them, the report makes silent passes visible; the reference circuits (FT-21) come after the integration of the candidates, by the owner's decision
0b	FT-22 short rule listing, FT-24 eval with Sonnet	edakit must run with Sonnet: the listing drops from 27.5 KB to about 2 KB, and every change is measured with Sonnet on the reference circuits
1	FT-1 load steps	in all five maker clusters and in both guitar clusters; the most frequent "it resets when the motor starts"; the bell's servo
2	CK-13 with PF	one mechanism for about 15 pitfalls; the knowledge stays in parts entries, as strap and pull do
3	CK-1, CK-2	small changes to G-4 and G-5; floating inputs and missing pull-ups recur in four clusters
4	CK-3	fully decidable, frequent in sensor stacks
5	CK-4	IRF520 modules and high-side P-MOSFETs: a hot part or a load that never switches off
6	CK-6	in four clusters; every battery, thermistor and current measurement
7	CK-7	a gap in G-10 itself; audio and ±12 V
8	CK-5 with FT-8	dead first pixels, half-powered drivers
9	CK-11	a board respin when it is missed
10	FT-2 with HT-1, CK-16	the owner's pedal; cable length is a question that comes back

For the owner's projects in particular: CK-8 (tinyjukebox), BL-11 and FT-6's optocoupler (asp_pro's MIDI), HT-7 (jacks with a supply), FT-4 (asp_pro's battery), CK-11 (ESP32 boards). For a guitar circuit: CK-27 (input impedance), FT-3 (op-amps at 9 V), CK-30 (DC at switches), FT-17 (bias of discrete stages), BL-16 to BL-18.

Mains and high voltage (FT-7, CK-9, CK-10, CK-24) prevent the most expensive faults – shock and fire. They become the domain mains, with CK-22 as its first step: a net above the ES1 limits says that edakit does not check its spacing. Whether edakit then measures safety distances – a statement other people may rely on – is open decision 3.

Out of scope

• Radiated emission and immunity as field strengths, transmission lines – the hints and the reviewers cover them.
• Settings that live in firmware only: SPI mode, motor PWM frequency, debounce times, brown-out levels – hints, the firmware contract, and H-9 where a part of the circuit decides.
• Certification itself: the regulatory review (M-6) is support, not a test lab.
• Tube amplifiers: B+ of 250–500 V, heaters, output transformers, tube models that do not hold at low anode voltage – unless edakit takes hazardous voltage in (open decision 3).
• The pinball machine (owner's decision).

Open decisions

1. The proposed guardrails 3 to 5 (section "Keeping the catalog stable"); domains and condensed rules are decided.
2. Which candidates become proposals – one issue each through /edakit:propose-rule and /edakit:propose-feature, or built directly with add-rule and add-feature.
3. Whether edakit takes boards with mains or high voltage (FT-7, CK-9, CK-10), or only says that it does not check them (CK-22). Decided 2026-10-31: recognised, not checked – a domain mains, CK-22 as an error a project may accept; FT-7, CK-9, CK-10, CK-24 stay out (rules-at-scale.md, decision 7).
4. Whether a library of checked entries for popular parts (PF) belongs into edakit – who signs them, and how a project takes one over (today: edakit/blocks/parts/, copied by edakit block add). Decided 2026-10-31: in edakit (edakit/parts/), taken over as a copy that names its library version; the owner signs every entry (rules-at-scale.md, decision 8).
5. Whether AC analysis (FT-12) fits edakit's DC-first approach at all. Decided 2026-10-31: yes, opt-in where sim.json names an expectation, domains audio and guitar, warnings first (rules-at-scale.md, decision 9).
6. The network policy: the vendor hosts the sources need (ti.com, st.com, espressif.com, microchip.com, analog.com, nxp.com, pololu.com, adafruit.com, onsemi.com, diodes.com, vishay.com, infineon.com, allegromicro.com, midi.org, usb.org) are allowed in the environment, or the sources are read on a machine that reaches them.

Sources

Per pitfall with its tag in the appendix. The documents read in full (r):

• Espressif: ESP32 Hardware Design Guidelines (schematic checklist, PCB layout), ESP32-S3 Hardware Design Guidelines v1.2, esptool docs/en/advanced-topics/boot-mode-selection.rst and troubleshooting.rst, ESP-IDF GPIO and ADC documentation (v4.4 and current), ESP-IDF Kconfig.system (brown-out levels)
• Raspberry Pi: "Hardware design with RP2040", raspberrypi/hats design guide
• ST AN2867 Rev 5 (oscillator design), ST AN4488 Rev 7 (STM32F4 hardware), NXP UM10204 Rev 7.0 (I2C)
• KiCad pcb_calculator: IEC 60664-1 tables (F.1, F.2, F.5) and IPC-2221 spacing and track width
• QMK docs/config_options.md, how_a_matrix_works.md, features/split_keyboard.md, drivers/serial.md, platformdev_rp2040.md; ZMK docs/config/kscan.md
• Marlin Configuration.h and Configuration_adv.h; Klipper docs/TMC_Drivers.md, Config_Reference.md; ArduPilot "Powering the Pixhawk"; Voron documentation, electrical
• OpenEnergyMonitor CT and AC-AC documentation; Tasmota "Getting Started" and PZEM-0XX; ESPHome I2C, 1-Wire, ADC and AC dimmer documentation
• SparkFun Weather Meter Kit library, SparkFun BOB-12009 schematic; RF24 COMMON_ISSUES.md; RadioLib troubleshooting; RobTillaart PCF8574, INA219, INA226 and ACS712 READMEs; bogde/HX711 and olkal/HX711_ADC; autowp/arduino-mcp2515; sandeepmistry arduino-CAN and arduino-LoRa; Arduino analogWrite reference; ArduinoCore-avr boards.txt; MiniCore
• Guitar: ElectroSmash's circuit analyses (Fuzz Face, germanium fuzz, Rangemaster, Big Muff, Tube Screamer, Klon, ProCo RAT, 1Wamp, Dyna Comp, Phase 90, CE-2, Crybaby, PT2399, MN3007, Back Talk, JRC4558) in mstratman/electrosmash-pdf-backups; Spin FV-1 datasheet SPN1001-DS-170829, ST TL074 datasheet DocID2297 Rev 5 and the Intersil CA3080 datasheet, each in a GitHub repository; PJRC Audio library (control_sgtl5000.cpp, gui/index.html); mstratman/relay-bypass; tardate/LittleArduinoProjects (Little Gem, Ruby); jonnor/piezopro; wagiminator/ADuM3160-USB-Isolator; DIYSynthMNL/Eurorack-PT2399-Delay; flpvsk/pt-workshop; ElectricCanary/Bontempo; Eight4aWish/eurorack_electronics (PCM3060 notes); hed0rah/partinfo

Appendix: the pitfalls

One row per pitfall: the pitfall with its numbers, what edakit does about it today, the candidate that would cover it, and the source.

edakit today: covered (the check or hint named), with part facts (an existing check finds it once the parts entry carries the datasheet value), partly, no. Source tags: r read in the primary document, x from a search engine's extract of the cited page, v to verify – not reachable from this environment, or from memory. A value tagged x or v is read in its source before it goes into a catalog entry.

LEDs, lighting and displays (32)

#	Pitfall	edakit today	Candidate	Source
1	3.3 V data into a 5 V pixel: VIH 0.7 × VDD = 3.5 V (WS2812B V1.0), 0.63 (V5), 0.55 (V6), SK6812 and APA102 0.7, WS2815B 2.7 V at 12 V	with part facts: E-3	PF, BL-1	WS2812B V1.0, V5, V6, SK6812mini, APA102C, WS2815B datasheets (x)
2	level shifter of the HC family (CMOS VIH ≈ 0.7 × VCC) instead of HCT/AHCT (VIH 2.0 V); shifter fed from 3.3 V; OE and unused inputs floating	with part facts: E-3; floating nc inputs: no	CK-1, BL-1	TI SN74AHCT125 (x)
3	BSS138 shifter with 10 kΩ too slow for 800 kbit/s: 2.2·R·C ≈ 1.1–2.2 µs against T0H 0.4 µs	no – the simulation knows no edges	BL-1, HT-1	forum threads (x)
4	no series resistor on the data line (Adafruit: 300–500 Ω); unbuffered runs beyond 1–2 m	no	BL-1, HT-1	Adafruit NeoPixel best practices (x)
5	data high while the strip is unpowered: DIN abs. max VDD + 0.5 V, current through the clamp, first pixel dies	no	CK-5, FT-8	WS2812B V1.0 (x); Adafruit (x)
6	no bulk capacitor at the strip (500–1000 µF), no 100 nF per LED	covered: H-1, G-6	BL-1	Adafruit (x); WS2812B application circuit (x)
7	current budget: up to 60 mA per pixel at white, 300 pixels = 18 A	covered: a strip (kind led_strip) draws i_pixel_a per pixel at the load's share of full white – against the source's i_max_a, each feed against its connector and cable (LIM.i, LIM.i/strip)	FT-13	Adafruit powering guide (x)
8	LEDs fed through the MCU board's regulator or USB (NCP1117 dissipation, 500 mA USB, ESP32 brown-outs)	covered: E-3 module LDO, source limit; the burst: no	FT-1	Arduino and ESP32 forums (x)
9	voltage drop along the strip: 5 m at 60 LEDs/m, 2.5 V at the far end; blue fails below about 3.3 V	covered: LIM.range/strip on the ladder of the strip's rails, fed at its ends and along it (cable at_m); the rail resistance from the strip's seller or the project (PARTS.fact)	FT-13	QuinLED injection guide (x); Hackaday (x); Pololu WS2812B strips, measured drops (r)
10	wire gauge of feeds: 22 AWG ≈ 53 mΩ/m, 2 m at 3.6 A drops 0.76 V	partly: cables r_ohm_per_wire	FT-2	QuinLED (x); WLED wiring (x)
11	no fuse between a 300 W supply and thin wiring	covered: G-17 for a modelled cable; a missing fuse: no	FT-2	QuinLED fuses (x); Hackaday (x)
12	data reference: shared ground missing or thin (5 A through 1 m 22 AWG lifts 0.26 V)	covered: per-board ground with the cable's resistance, E-3 levels to the part's own ground	FT-2	Adafruit (x)
13	WS2815 backup input BI of the first pixel floating	G-4 in a net; nc: no	CK-1	WLED forum (x)
14	pin states at boot: ESP32 GPIO 1, 3, 5, 14, 15 toggle; GPIO 12 strap; gate without pull-down	covered: G-12, G-4; toggling pins as a pin hint	PF	Espressif (x); WLED forum (x)
15	footprint pinout and orientation (SK6812 MINI-E datasheets disagree, reverse mount)	covered: P-1, L-10, L-8, P-2, G-9	–	Fritzing forum, KiCad MR (x)
16	LED supply tracks too narrow (IPC-2221, 10 °C: 3 A ≈ 1.37 mm, 5 A ≈ 2.8 mm)	covered: L-5	–	IPC-2221 formula
17	strip MOSFET not logic-level: IRF520 and IRFZ44N specified at VGS 10 V; TIP120 drops about 2 V	partly: E-2 with the SPICE model, no threshold corner	CK-4, BL-2	Adafruit RGB strips (x); datasheets (x)
18	PWM flicker: IEEE 1789 – 100 % modulation needs ≥ 1250 Hz (low risk), ≥ 3 kHz (no effect); Arduino 490 Hz	no	HT-9	US DOE summary; Finelite white paper (x)
19	constant-current buck sized wrong (PT4115: I = 0.1 V / RS, RS ≥ 0.082 Ω; inductor table; VIN above ΣVf)	covered: behaviour cc_led_driver – LIM.i/cc (current, inductor saturation), REQ.value/cc (RS), LIM.range/cc (VIN above the string); the inductance table: no	FT-14	PT4115 PT4115_DS Rev EN_2.9, AL8860 DS39014 Rev. 8 - 2 (r)
20	power LED without heat sink: θ ≤ (Tj − Ta) / P	partly: E-2 rth_ja_k_per_w of the part alone	HT-11	star LED datasheets (x)
21	MAX7219 at 3.3 V logic: VIH 3.5 V; common-cathode displays only	with part facts: E-3	PF	MAX7219 datasheet (x)
22	MAX7219 RSET, about 1 W dissipation, 10 µF + 0.1 µF at the chip	partly: G-6	PF	MAX7219 table 11 (x)
23	74HC595 driving LEDs: ±35 mA per output, 70 mA through VCC or GND	no	CK-21	TI CD74HC595 (x)
24	multiplexed commons on MCU pins (8 × 15 mA); charlieplexing reverses LEDs	covered: E-2 GPIO current per scenario, LED vr_max	–	Arduino pin limits (x)
25	TLC5940 headroom heat; PCA9685 25 mA sink	covered: cc_sink, E-3 driver temperature; PCA9685 limits: no	CK-21	TI TLC5940, NXP PCA9685 (x)
26	SPI TFT: 5 V logic into 3.3 V inputs, backlight 40–70 mA from a GPIO	covered: G-3, E-2 GPIO current	–	display vendor pages (x)
27	SSD1306 module: pull-ups to 5 V, addresses 0x3C/0x3D only, pull-ups stacked	covered: G-3, G-11; the address: no	CK-3	module datasheets (x)
28	PCF8574 backpack: VIH 0.7 × VCC; 0x27 (PCF8574) or 0x3F (PCF8574A); contrast bias at 3.3 V	with part facts: E-3; address: no	CK-3	NXP PCF8574 (x)
29	e-paper booster: sense resistor per panel, booster part ratings	partly: G-8	PF	Waveshare wiki (x)
30	Nixie 170–180 V boost: switch, diode, capacitor ratings; anode resistor; 74141 off-state about 55–60 V	covered: G-8, E-2; the spacing: no	CK-22, CK-9	builder reports (x)
31	high-voltage clearance: IPC-2221 151–250 V, outer uncoated 1.25 mm	no	CK-9	IPC-2221 via KiCad (r)
32	VFD: filament, grid voltage, driver range, filament bias	partly: G-8, E-3	–	MAX6921 (x)

Motors, motion, robotics, 3D printers and CNC (35)

#	Pitfall	edakit today	Candidate	Source
1	L298N drop 1.8 V typ, 3.2 V max at 1 A; 1.8–3.2 W per bridge	no model	FT-6, BL-14	ST L298 (x)
2	bipolar bridge (L298, L293 non-D) without clamp diodes	no	CK-15	ST L298 application circuits (v)
3	VM rating against a full LiPo: DRV8833 10.8 V, TB6612 15 V abs. max, 3S = 12.6 V	covered with part facts: E-3 vdd_op_max with full cells, G-3	PF	Toshiba TB6612 (x); Pololu forum (x)
4	stall and start current above the driver's rating (I = V / R_winding)	partly: a constant fault case	FT-1, FT-6, BL-14	L9110, TB6612 datasheets (x)
5	no bulk capacitor at VM; a series diode lets the motor pump VM up (12 V, 1 mH, 2 A into 10 µF → 23 V)	no	HT-12, BL-14	TI DRV8833 (v)
6	brush noise; capacitors on PWM outputs trip over-current	no – judgement	–	Pololu (v)
7	PWM at 490 Hz: whine and ripple	no – firmware	–	Arduino analogWrite reference (r)
8	shunt sized for value, not power (0.5 Ω at 2 A = 2 W); sense through shared ground	covered: E-2 resistor power; Kelvin: no	reviewer question	DRV8833 (v)
9	VMOT LC spike at hot-plug: up to about 2× the supply, above the A4988's 35 V; ≥ 47 µF electrolytic	partly: H-1; E-5 has no inductance	FT-2, HT-12	Pololu A4988 product page (x)
10	Vref formula by board: A4988 I = VREF / (8 · R_CS), R_CS 0.068, 0.050 or 0.1 Ω	no	PF, FT-10	Pololu (x); BTT TMC2209 wiki (x)
11	driver heat through the exposed pad: A4988 about 1 A per coil without cooling	partly: E-2 rth_ja_k_per_w	HT-11	Klipper TMC drivers (r)
12	motor connected or disconnected under power; coil pairs mixed up	no – judgement	–	Pololu (x); Klipper (r)
13	A4988 VIH 0.7 · VDD at VDD 5 V against a 3.3 V MCU	with part facts: E-3	PF	Allegro A4988 (x)
14	A4988 RESET floating; TB6612 STBY and DRV8833 nSLEEP pulled down; TMC UART wiring and addresses	G-4 in a net; nc, addresses: no	CK-1, CK-3	Marlin Configuration_adv.h (r); Pololu forum (x)
15	TMC driver loses its settings when VM drops; logic inputs feed the unpowered driver	no	CK-5, FT-8	Klipper (r)
16	servo stall on the MCU's rail: MG996R 2.5 A, Uno polyfuse 500 mA, AVR BOD 2.7 V, ESP32 2.43 V	partly: constant currents	FT-1, BL-13	MG996R (x); ArduinoCore boards.txt (r); ESP-IDF Kconfig.system (r)
17	back-driven servos push the rail above 5.7 V	no	FT-1	ArduPilot "Powering the Pixhawk" (r)
18	BECs in parallel; linear BEC 7.6 W at 3S, 1 A; ESC signal without ground	covered: G-5, H-8, E-3	–	ArduPilot (r)
19	LiPo range; battery divider beyond the ADC window (ESP32 150–2450 mV at 11/12 dB)	partly: E-3 with cell states; the window: no	CK-6	ArduPilot (r); ESP-IDF ADC (r)
20	fuse, switch, connectors, copper in the high-current path (MicroFit 5 A per contact)	covered: G-17, E-2 contact current, L-5	–	Voron documentation (r)
21	coils without clamp; coils on a GPIO	covered: G-13 (relays, buzzers), H-4, E-2; solenoids and motors on the board: no	CK-15, FT-6	v
22	5 V relay module (LED anode at 5 V) on a 3.3 V MCU: does not release, pin pulled to 5 V	covered where modelled: G-3, E-3 relay release	–	judgement
23	4-pin PWM fan: open-drain PWM input pulled to 5 V, open-collector tach	partly: G-3; open-drain: no	CK-2	Linux pwm-fan binding (r); Intel fan specification (v)
24	12 V fans and loads on 24 V; supplies without a common 0 V	covered: G-3, E-3, G-16	–	Marlin (r); Voron (r)
25	heater MOSFET IRF520 (RDS(on) 0.27 Ω only at 10 V) from 5 or 3.3 V	partly: E-2 with the SPICE model	CK-4, BL-2	IRF520, IRLZ44N (v)
26	gates without pull-down; ESP32 pins with PWM at boot, GPIO 12 strap	covered: G-4, G-12; boot activity as a pin hint	PF	Espressif hardware design guidelines (r)
27	bed current through terminals and polyfuses (RAMPS 11 A)	covered: E-2, G-17, L-5 with an external load	FT-6	Voron (r); RepRap (v)
28	12 V parts on 24 V: a heater gives 4× the power, NCP1117 20 V abs. max	covered: G-3, G-8, E-3; the heater as a load: no	FT-6	v
29	thermistor input: firmware assumes 4.7 kΩ, 5 V pull-up on a 3.3 V ADC, no filter	partly: G-3; window and firmware constant: no	CK-6, FT-10	Marlin Configuration.h (r); Klipper (r)
30	heater switch fails closed, firmware cannot stop it	no	HT-10	Marlin, Klipper, Voron (r)
31	endstops on long leads, probes at 12–24 V into 3.3 V pins, NO switches fail silently	covered: G-3, G-4; the cable: no	HT-1, BL-12	Marlin (r); Voron (r)
32	encoders: open-collector outputs without pull-ups, RC slower than the edges	partly; open-drain: no	CK-2	Bourns PEC11R (v)
33	HC-SR04 echo at 5 V into a 3.3 V pin	covered: G-3	–	v
34	TCRT5000 LED current (60 mA abs. max)	covered: E-2 LED current where modelled	–	Vishay (v)
35	motor or servo start pulls the logic rail below brown-out (4 × AA 0.6–1 Ω; AMS1117 1.1 V dropout; 16 MHz AVR needs about 3.8 V)	no	FT-1, CK-23	ESP-IDF (r); MiniCore (r)

Power supplies, batteries and low-power nodes (32)

#	Pitfall	edakit today	Candidate	Source
1	TP4056 with R_PROG 1.2 kΩ charges at 1 A – 2 to 6.7 C on a 150–500 mAh cell; MCP73831 I = 1000 / R_PROG(kΩ) mA	no	FT-4, BL-4	TP4056, MCP73831, NCR18650B datasheets (x)
2	load on the cell while charging: termination at C/10 never reached	no	HT-5, BL-4	Microchip AN1149 (x)
3	linear charger dissipation: TP4056 at 1 A from 5 V into 3.0 V = 2 W	no model	FT-4	MCP73831 thermal example (x)
4	charger's TEMP/NTC pin grounded: charging below 0 °C	no	PF	TP4056 (x); Battery University (x)
5	protection IC thresholds against the cell (DW01A 2.4 V); no undervoltage cut-off	partly: H-6	BL-4	DW01A (x)
6	coin cell: CR2032 3 mA continuous, pulses 15–30 mA, internal resistance tens of Ω	partly: cell r; a limit: no	CK-18, FT-1	Energizer CR2032 (x); TI SWRA349 (x)
7	JST-PH battery polarity differs between vendors	partly: G-10 for connectors in reverse_protect	HT-6	Adafruit (x)
8	AMS1117 dropout 1.1–1.3 V on a LiPo or behind a Schottky	covered: ldo model, E-3	–	AMS1117 (x)
9	LDO output capacitor outside its window (AMS1117: 22 µF tantalum, ESR)	no	CK-14, CK-13	AMS1117 (x); LM2596 (x)
10	sleep current of the board: AMS1117 5 mA, power LED, CP2102 (DevKitC 3.8 mA in deep sleep)	partly: the simulation's currents, no budget	FT-11	ESP32 datasheet (x); Spiess #193 (x)
11	low-Iq LDOs: MCP1700 250 mA, 6 V input; ESP32 needs ≥ 500 mA	covered: E-3, G-3	BL-5	MCP1700, HT7333, XC6206 (x)
12	linear regulator headroom and heat: 7805 2 V dropout, 12 → 5 V at 0.5 A = 3.5 W	covered: ldo model, E-3	–	L7805 (x)
13	buck module at an unknown trimmer setting, no reverse protection	no	FT-5	MP1584 (x); LM2596 (x)
14	feedback divider (VREF 1.25, 1.23, 0.8, 0.6 V); LM317 needs 10 mA load	no	FT-5	onsemi LM317 (x)
15	non-synchronous boost passes VIN − about 0.3 V with EN low (MT3608)	covered: HINT.boost_enable; behaviour boost shows VIN − VD with EN low, the load's expect holds it	HT-4, FT-5	TI SLVA191A (r); EEVblog (x)
16	ceramic DC bias: 4.7 µF, 6.3 V, 0603 X5R at 5 V ≈ 0.33 µF	no	FT-9	ADI tutorial 5527 (x); Nordic (x)
17	battery divider: 100 k + 100 k draws 21 µA; MΩ dividers against the ATmega's 10 kΩ	no	CK-6, BL-6, FT-11	ATmega328P §23.6.1 (x); ESP-IDF ADC (x)
18	Wi-Fi bursts 240–335 mA against the supply's impedance: brown-out	partly: E-3 at i_core_a	FT-1, BL-7	Espressif schematic checklist (x)
19	brown-out and clock against the supply: ATmega328P at 16 MHz needs about 3.78 V	no	CK-23	ESP-IDF Kconfig.system (x); Microchip (x)
20	back-powering through I/O pins into a switched-off rail (TI: ±2 mA)	no	CK-5, FT-8	TI SLAA689 (x)
21	6 V panel straight into a TP4056 (VIN 4–8 V, panel Voc about 7.5 V)	partly: G-3, E-3	–	TP4056 (x); CN3791 (x)
22	reverse current into the panel at night	covered: E-3 back-feed into sources	–	x
23	supercapacitors: 2.7 V per cell, balancing	partly: G-8	–	Würth ANP090 (x)
24	supercapacitor ESR (about 50 Ω for 0.47 F coin parts) and inrush	partly: E-5 plug-in	–	Eaton KR (x)
25	USB budget: 500 mA (USB 2.0), 900 mA (USB 3), Type-C 1.5 or 3 A only when advertised	partly: E-3 against the i_max_a set by hand	HT-2	Microchip; Infineon (x)
26	USB-C CC resistors missing or shared	covered: G-14	–	Hackaday; kernel.org (x)
27	PD trigger asks for 20 V (CH224K with CFG1 open)	partly: G-3 when the net's range is right	CK-20	CH224K (x)
28	inrush into capacitance: USB 2.0 allows 10 µF at attach	partly: E-5 peak current	HT-3	USB 2.0 §7.2.4.1 via Maxim (x)
29	several supplies on one rail; a LiPo on the 3V3 pin (4.2 V > 3.6 V)	covered: H-8, E-3, G-3	–	Espressif DevKitC (x)
30	high-side P-MOSFET driven from a 3.3 V GPIO on 5 or 12 V	no	CK-4, BL-3	x
31	soft-power latches that latch, drop or leak	no	FT-1 (partly)	Circuit Cellar (x)
32	reverse polarity at a barrel jack or battery input; Boss centre-negative	covered: G-10 for listed inputs	HT-6, CK-7	TI SLVA139 (x)

MCU boards, buses, radio, keyboards and HATs (39)

#	Pitfall	edakit today	Candidate	Source
1	ESP32 EN without RC: 10 kΩ and 1 µF; tSTBL 50 µs after VDD	no	CK-13, BL-7	ESP32 HDG §3.3 (r)
2	auto-program circuit: RTS straight to EN, no EN capacitor (1–10 µF), large capacitor on GPIO0	no	BL-7, CK-13	esptool boot mode selection (r)
3	loads on strapping pins (ESP32 GPIO 0, 2, 12, 15; S3 GPIO 0, 45, 46; C3 GPIO 8, 9)	with part facts: G-12	PF	esptool (r); ESP32 and ESP32-S3 HDG (r)
4	flash and PSRAM pins, input-only GPIO 34–39 without pulls, ADC2 with Wi-Fi, USB pins	with part facts: G-15, E-3, pin hint	PF	ESP-IDF GPIO and ADC (r)
5	3.3 V supply too weak: Espressif asks for ≥ 500 mA and 10 µF	partly: E-3 at DC	FT-1, BL-5	ESP32 HDG §3.2 (r)
6	module antenna over copper: over the edge or ≥ 15 mm clear	no	CK-11	ESP32 HDG §4.2 (r)
7	crystal load capacitors: C = 2 · (CL − Cstray), Cstray 3–5 pF; gain margin ≥ 5	no	CK-13	ST AN2867 (r); RP2040 HDG (r)
8	USB full speed: series resistors (RP2040 27 Ω), D+ pull-up 1.5 kΩ (Blue Pill 10 kΩ), 90 Ω pair, ESD at the connector	no	CK-13, BL-8	RP2040 HDG §2.4 (r); USB 2.0 (v)
9	USB-C: D+ and D− of both rows joined; CC per pin	CC covered: G-14; D±: no	CK-12	USB Type-C specification (v)
10	STM32 VCAP 2 × 2.2 µF, BOOT0, NRST 100 nF, VDDA, VBAT	no	CK-13	ST AN4488 Rev 7 (r)
11	no way back after bad firmware: SWD without NRST, no BOOTSEL, UPDI loaded	no	CK-13	AN4488 (r); RP2040 HDG (r); AVR-Guidance (r)
12	ATmega AVCC unconnected; AREF tied to VCC	partly: G-16 for AVCC	CK-13	ATmega328P datasheet (v)
13	Arduino auto-reset (100 nF, 10 kΩ); modules on D0/D1 behind the 16U2's 1 kΩ	partly: G-5	CK-13	Uno Rev3 schematic (v)
14	UART TX to TX, RX to RX; CH340 at 5 V into a 3.3 V RX; CH340 V3 pin	covered: G-5, G-4, G-3; RX to RX with a pull-up passes	CK-26, CK-13	QMK serial driver (r)
15	I2C address clashes: DS3231 and MPU-6050 at 0x68, PCA9685 and INA219 at 0x40, PCF8574 against PCF8574A, All-Call 0x70	no	CK-3	RobTillaart READMEs (r); datasheets (v)
16	pull-ups too weak for the bus: tr = 0.8473 · Rp · Cb, 400 pF; Rp ≤ 2.95 kΩ (100 kHz), 885 Ω (400 kHz)	partly: G-11 with a fixed 10 kΩ	CK-3, FT-2	NXP UM10204 §7.1 (r)
17	too many pull-ups in parallel: Rp ≥ 967 Ω at 3.3 V	covered: G-11	–	UM10204 (r)
18	5 V I2C modules on a 3.3 V bus	covered: G-3, E-3	BL-9	UM10204 (r)
19	wrong level shifter: BSS138 with 10 kΩ too slow for 400 kHz, TXB0108 with pull-ups	no	BL-9	SparkFun BOB-12009 (r); TI TXB0108 (v)
20	open-drain INT and ALERT without pull-up; DS3231 module charges a CR2032	no; charging a primary cell: E-3 where modelled	CK-2	RobTillaart PCF8574 (r); ESP-IDF GPIO (r)
21	DS18B20 without 4.7 kΩ; parasite power wired wrong	no	CK-2, CK-13	Arduino-Temperature-Control-Library (r)
22	sensor output beyond the ADC (ACS712 0.5–4.5 V around 2.5 V; 10 kΩ NTC at 80 °C)	no	CK-6	ESP32 HDG §3.11 (r); arduino-esp32 ADC (r); RobTillaart ACS712 (r)
23	ADC source impedance (ATmega 10 kΩ; ESP32 0.1 µF at the pin)	no	CK-6, FT-3	ESP32 HDG (r); ATmega328P (v)
24	resistive soil probes corrode; NE555 (4.5 V min) at 3.3 V	partly: E-3 with part facts	PF	v
25	HX711: AVDD headroom, ±20 mV full scale, RATE strap	partly: E-3 with part facts	CK-13, CK-20	bogde/HX711 (r)
26	5 V outputs into pins that are not 5 V tolerant	covered: G-3	–	QMK RP2040 (r); RadioLib (r)
27	RP2350 erratum E9: inputs latch at about 2.1 V with the internal pull-down	with part facts: a pin hint	PF	Hackster (x)
28	RS-485: more than two 120 Ω, no fail-safe bias, DE floating at reset, 5 V RO into 3.3 V	partly: G-4 (DE), G-3 (RO)	CK-17, BL-10	TI SLLA272 (v)
29	CAN: termination, TJA1050 at 5 V, 8 MHz crystal against a 16 MHz library default	partly: G-3	CK-17	autowp/arduino-mcp2515 (r); arduino-CAN (r)
30	MIDI: input without opto or with a slow one, no reverse diode, pin 2 grounded; output resistors not for 3.3 V	no	BL-11, FT-6	MIDI CA-033 (v)
31	nRF24L01 bursts; bulk capacitor at the module	partly: G-6	FT-1	RF24 COMMON_ISSUES.md (r)
32	LoRa SX127x: 3.3 V only, 120 mA bursts, antenna before the first TX	covered: G-3; the burst: no	FT-1	arduino-LoRa (r); RadioLib (r)
33	keyboard matrix: diode direction against the firmware (QMK COL2ROW, ZMK default row2col)	no	CK-19	QMK (r); ZMK (r)
34	split keyboards: TRRS with VCC not hot-pluggable; pull-ups of the link	partly: G-11	HT-7, CK-2	QMK split keyboard (r)
35	per-key RGB: VIH 3.5 V, 70 LEDs × 60 mA against 500 mA USB	with part facts: E-3	BL-1, HT-2	QMK RGB matrix (r)
36	Raspberry Pi HAT: back-powering without ideal diode, ID EEPROM on GPIO 0/1 with 3.9 kΩ	partly: H-8, E-3 back-feed	CK-13	raspberrypi/hats design guide (r)
37	Arduino shield logic on 5 V instead of IOREF	covered: G-3 where modelled	–	v
38	floating CMOS inputs, unused gates, unused op-amp halves	partly: G-4 for nets; nc: no	CK-1, FT-3	TI SCBA004 (v); ESP32 HDG §3.10 (r)
39	SPI: CS floating at reset; modules whose buffer never releases MISO	partly: G-4 for CS; tri-state: no	CK-2	judgement

Audio, mains, high voltage, outdoor and measurement (35)

#	Pitfall	edakit today	Candidate	Source
1	bridge-tied class-D outputs grounded, shared between channels or fed onward (PAM8403, PAM8302, MAX98357A, TPA3116D2)	no	CK-8, HT-13	PAM8403 (v); Adafruit MAX98357A (v)
2	supply or speaker impedance outside the window (PAM8403 2.5–5.5 V; TPA3116D2 RL ≥ 3.2 Ω BTL)	partly: E-3 with part facts; the speaker: no	FT-6	TPA3116D2 operating conditions (r)
3	amplifier supply budget (2 × 3 W into 4 Ω ≈ 1.4 A at 5 V) and local decoupling	partly: G-6, E-3 with a current model	FT-1	TPA3116D2 layout (r)
4	pop at power-up: shutdown pin tied or floating	partly: G-4 floating; the level at start: strap with G-12	CK-20, PF	PAM8403 (v)
5	configuration pins by voltage window: MAX98357A SD_MODE < 0.16 / 0.77 / 1.4 V; INMP441 L/R floating	partly: G-12 knows high and low only	CK-20	Adafruit MAX98357A pinouts (r); ADMP441 (r)
6	microphone bias against the ADC window (ESP32 0 dB: 100–950 mV); Wi-Fi noise on the microphone's rail	no	CK-6, HT-8, FT-3	ESP-IDF ADC (r)
7	coupling capacitor too small (32 Ω with 10 µF: 497 Hz) or reverse-biased	no	CK-7, FT-12	first-order high-pass; TI headphone amplifiers (v)
8	ground loops, shared return copper, Y-capacitor noise of switch-mode supplies	no – judgement	HT-8	OpenEnergyMonitor (r)
9	linear supply at high mains (25 V AC → about 42 V unloaded) against capacitors and the LM3886	covered: G-8, G-3 when the net's range holds the high line	–	OpenEnergyMonitor (r); LM3886 (v)
10	Eurorack: reversed 10-pin ribbon swaps ±12 V, 16-pin offset, shunt diodes short the supply	partly: G-10 for a 2-pin reversal	FT-8, CK-7, BL-15	elby-designs (r); EuroBusboard (r)
11	pedals: 9 V centre-negative (Boss), series Schottky	covered: G-10 for listed inputs	HT-6	Boss PSA (v)
12	charge pumps: ICL7660 10 V max, 10 kHz whine, doubled rail against 16 V capacitors	covered: E-3, G-8; the whine: judgement	–	TC1044S, LT1054 (v)
13	true-bypass pop: switched audio nodes without a DC path	no	CK-25	R.G. Keen, The Technology of Pops (v)
14	mains to SELV: IEC 60664-1 at 230 V, OVC II, PD 2 – clearance 1.5 / 3.0 mm, creepage 2.5 / 5.0 mm (basic / reinforced), slots from 1 mm	no	CK-9, FT-7	KiCad iec60664.cpp (r)
15	spacing between mains nets from RMS instead of peak: 325 V peak → IPC-2221 2.5 mm outer uncoated	no	CK-9	KiCad IPC-2221 (r)
16	parts on the barrier rated basic only (relay 1500 VAC coil to contact; ACS712 184 V peak reinforced)	no	CK-10	HLK datasheet (r); Allegro ACS712 (v)
17	non-isolated supplies (capacitive dropper, Sonoff) with programming or sensor ports	no	CK-10	Tasmota (r)
18	mains input: fuse first, MOV and X capacitor after it, bleeder, Y-class across the barrier	no	CK-10	HLK (r); IEC 62368-1 §5.5.2.2 (v)
19	copper for 10 A: 4.7 mm at 20 °C rise on 1 oz	covered: L-5 with a declared current; AC loads are not simulated	–	KiCad track width (r)
20	relay contacts: resistive rating against inrush of LED drivers and inductive loads	no – judgement	HT-14	Songle, Mean Well (v)
21	TRIAC dimmer: zero-cross optotriac for phase control, VDRM 400 V, gate resistor, snubber	no	PF	ESPHome ac_dimmer (r); onsemi MOC302x (v)
22	resistor voltage in mains and high-voltage chains (1206: 200 V)	partly: G-8 against the highest net voltage	–	Yageo RC (v)
23	optocoupler LED sized for 5 V driven from 3.3 V (PZEM's 1 kΩ)	no	FT-6	Tasmota PZEM (r)
24	current transformer without burden (SCT-013-000: 22 Ω at 3.3 V); bias network	no	CK-13	OpenEnergyMonitor (r)
25	protective earth and equipment class	no – judgement, M-6 for a sale	–	IEC 61140 (v)
26	Nixie 170–200 V and Geiger 400 V: switch, diode, capacitor ratings; 1.25 / 2.5 mm spacing	covered: G-8, E-2; spacing: no	CK-9, CK-22	KiCad IPC-2221 (r)
27	capacitor banks (flash 5–9 J, coilgun 320 J) without bleeder	no	CK-24	KiCad IEC 60664 F.2 (r)
28	ESD and surge on pins to a cable: IEC 61000-4-2 8 kV contact; emonTx 1 kΩ and TVS	partly: H-7	CK-16, BL-12	OpenEnergyMonitor (r); IEC 61000-4-2 (v)
29	reed switches on cables: pull-ups, 100 ms debounce, vane voltages 30 mV apart	partly: H-3; the ADC: no	HT-1, CK-6	SparkFun Weather Meter Kit library (r)
30	buses on long wires: I2C bus capacitance, 1-Wire, RS-485 terminations and bias	partly: G-11	CK-3, CK-17, FT-2	ESPHome I2C and 1-Wire (r); UM10204 (v)
31	INA219 26 V, INA226 36 V common mode; shunt full scale and power	covered with part facts: G-3, E-3, E-2	PF	Tasmota (r); TI INA219, INA226 (v)
32	Kelvin connection of shunts: 1 mΩ of shared copper is 10 % on 10 mΩ	no	reviewer question	v
33	thermocouple: MAX6675 T− to ground, MAX31855 T− never to ground, 10 nF across	no	CK-13	Analog Devices MAX31855, MAX6675 (x)
34	ADC window and source impedance (ESP32 per attenuation; ESP8266 0–1 V; ACS712-05B 3.43 V at +5 A)	no	CK-6	ESP-IDF (r); ESPHome ADC (r)
35	measurement front ends on mains: CAT II 300 V → 2.5 kV impulse, clearance 3.0 / 5.5 / 8.0 mm for 4 / 6 / 8 kV	no	CK-9, CK-10	KiCad IEC 60664 F.1, F.2 (r)

Guitar effect pedals (37)

#	Pitfall	edakit today	Candidate	Source
1	germanium Fuzz Face bias: the fixed parts need Q1 gain 70–80 and Q2 110–130; AC128 spans 45–165; leakage doubles every 10 °C	with part facts: sim.json spread – hFE and leakage (icbo_a) as corners, the bias against a load's expect (FT-17, FT-11); no germanium datasheet could be read – a project enters its measured spread	FT-17, FT-11	ElectroSmash Fuzz Face, germanium fuzz (r)
2	PNP positive ground: the jack sleeves sit on +9 V and short a shared supply	no	HT-18	ElectroSmash germanium fuzz, Rangemaster (r)
3	Fuzz Face input 5.2–8.4 kΩ, output 15 kΩ (about 125 kΩ at mid volume: 2.1 kHz with 600 pF of cable)	no	CK-27, FT-18, FT-12	ElectroSmash Fuzz Face (r)
4	AM radio in high-gain transistor stages: the base–emitter junction rectifies; 100 pF base–collector, base stopper	no	HT-8	ElectroSmash germanium fuzz (r); forums (x)
5	Big Muff: 48.6 dB in two clipping stages, 470 pF Miller capacitors for stability, bias windows	the bias windows with part facts: spread corners against expect (FT-17, FT-11); stability: no	FT-17, FT-11, CK-31	ElectroSmash Big Muff (r)
6	Tone Bender MKII self-biases through leakage; the bias target is disputed	with part facts: the leakage in the spread corner max (icbo_a, FT-17); the target is the project's expect	FT-17	forums (x)
7	op-amps at 9 V: TL07x swing 1.5–3 V short of each rail, common mode 3–4 V above V−; phase reversal	no model	FT-3	ST TL07x (r); ElectroSmash Klon (r)
8	clipping diodes against the swing: an LED (1.5–1.7 V) in a 9 V feedback loop; octave rectifiers need matched diodes	no	FT-3, FT-12	ElectroSmash Tube Screamer, Klon, Big Muff (r)
9	4.5 V reference: loads ≥ 10 × the divider, 47–100 µF, 0.5 s settling thump	partly: G-6	FT-11, BL-17	ElectroSmash Klon, Tube Screamer (r)
10	Klon charge pump +18/−9 V: 27 V across an op-amp; from 12 V about 36 V, the TL07x maximum	with part facts: E-3 vdd_op_max with a 12 V supply corner	PF, CK-31	ElectroSmash Klon (r); ST TL07x (r)
11	LM308 needs 30 pF across pins 1 and 8; slew rate 0.3 V/µs	no	CK-28	ElectroSmash ProCo RAT (r)
12	JFET spread (J201 IDSS 0.2–1.0 mA, VGS(off) −0.3…−1.5 V): drain between about 5.2 and 8.2 V with fixed resistors	covered: IDSS with VGS(off) as spread corners, the drain against expect – MMBFJ201 (C891687), the J201 in SOT-23 (FT-17)	FT-17	ElectroSmash 1Wamp (r)
13	BS170 is D-G-S, 2N7000 S-G-D; VGS(th) 0.8–3.0 V; gate unprotected	covered: P-1, L-10 for the pinout; VGS(th) as spread corners, BS170 (C111691) and 2N7000 (C232838) in the library (FT-17)	FT-17	forums (x); partinfo (r)
14	input impedance against 1 MΩ (Tube Screamer 446k, RAT 494k, V847 wah 69.5k); input high-pass; series resistor against ESD	no	CK-27, BL-16, HT-8	ElectroSmash analyses (r)
15	output impedance and minimum load: PT2399 output wants > 100 kΩ, FV-1 ≥ 10 kΩ, PCM3060 ≥ 10 kΩ	no	CK-13	ElectroSmash PT2399 (r); FV-1 datasheet (r)
16	OTA compressors: CA3080 bias current ≤ 2 mA, differential input ≤ 5 V, about 100 mVpp linear range	no model	CK-13, FT-3	CA3080 datasheet (r); ElectroSmash Dyna Comp (r)
17	LFO and tap-tempo ticks through supply and ground (vactrol LED 20–40 mA; 1 kHz PWM into a PT2399)	no	HT-16, HT-8	ElectroSmash CE-2 (r); Bontempo (r)
18	phaser: 2N5952 VGS(off) −1.3…−3.5 V, matched FETs, trimmer range	with part facts: VGS(off) corners, matched FETs at one corner (FT-17); no 2N5952 datasheet reachable (onsemi: not found); the trimmer range: no	FT-17	ElectroSmash Phase 90 (r)
19	wah: V847 input 69.5 kΩ; the 500 mH inductor picks up the amplifier's transformer field	no	CK-27; reviewer question	ElectroSmash Crybaby (r)
20	5 V regulators for PT2399 (30 mA) and Belton (60–100 mA): 78L05 at about 200 °C/W; 78L05 OUT-GND-IN against 7805 IN-GND-OUT	covered: E-3 regulator temperature with rth_ja_k_per_w, P-1 and L-10 for the pinout	–	ElectroSmash PT2399 (r); forums (x)
21	PT2399 pin 6 below 2 kΩ during the first 400 ms latches the chip	no	CK-13	ElectroSmash PT2399 (r); DIYSynthMNL (r)
22	PT2399 noise: pins 3 and 4 joined short; filter capacitors per delay; input 0.5–1 Vrms, clips above 1.3 Vrms	no	CK-13, CK-31	ElectroSmash PT2399 (r)
23	BBDs: MN3007 −10…−15 V, MN3207 +5…+10 V (a 12 V outlet exceeds it); clock pins about 700 pF; clock filter	with part facts: E-3 supply range	PF, CK-31	ElectroSmash MN3007, CE-2 (r)
24	FV-1: 3.3 V, abs. max 3.5 V; every pin −0.5 V … VDD + 0.5 V, so pots go to 3.3 V; T0 and T1 tied; input RC; output load ≥ 10 kΩ	partly: G-3, E-3	CK-5, CK-13	FV-1 datasheet (r)
25	FV-1 crystal 32.768 kHz: short leads, 15 pF on X2; EEPROM solder joints	no	CK-13, CK-31	FV-1 datasheet (r)
26	converter full scale and input impedance against the guitar: SGTL5000 0.24–3.12 Vpp, PCM3060 10 kΩ, Daisy Seed 3.6 Vpp at 13.6–30 kΩ	no	CK-6, CK-27, BL-16	PJRC Audio library (r); PCM3060 via a GitHub transcription (r)
27	anti-aliasing: pedalSHIELD first order near 5 kHz; RP2040 ADC about 8.7 effective bits	no	FT-12	x
28	3.3 V and 9 V domains: a 4.5 V bias into 3.3 V parts; 9 V live while the 3.3 V rail is still 0 V	partly: G-3	CK-5, FT-8	FV-1 datasheet (r); ST TL07x (r)
29	digital pedals against the outlet: FV-1 40–70 mA, an ESP32 asks for ≥ 500 mA, outlets 100–500 mA	covered: E-3 source i_max_a; the bursts: no	FT-1	FV-1 datasheet (r); Espressif HDG (r)
30	power-up and brown-out: reset held until VDD is valid; the DC step at the output; relay state	partly: E-5 supply ramps	FT-1, FT-11	ElectroSmash Back Talk (r); relay-bypass (r)
31	relay bypass: a latching relay keeps its state when power is lost – silence; coil 20–34 mA; mute 35 ms	partly: G-13, E-3 pull-in	BL-18, CK-30	relay-bypass (r)
32	battery switched through the input jack's ring; the DC jack's lug must break battery plus	no	HT-15	ElectroSmash (r); forums (x)
33	layout oscillation: input and output closer than 2.5 mm along a run; shared return paths	no	CK-31	forums (x); ElectroSmash 1Wamp (r)
34	supply filtering against current (47 Ω at 30 mA drops 1.4 V); switch-mode adapter noise	partly: E-3 voltage at the part	HT-16, BL-17	ElectroSmash (r)
35	daisy chains, AC and centre-positive pedals; a series part in the DC jack's ground leg bypassed by the cable shields	no	HT-18	ElectroSmash CE-2 (r); x
36	12 and 18 V outlets: 16 V electrolytics marginal; a 9.1 V shunt Zener on 12 or 18 V; a shunt reverse diode	covered: G-8 with the supply corner, G-10, E-2 for the Zener	–	ElectroSmash Crybaby, Dyna Comp (r)
37	electrolytic polarity around the 4.5 V reference in every state; DC across pots	no	CK-7, CK-30	forums (x)

Guitar instrument, amplifiers, DI and switching (39)

#	Pitfall	edakit today	Candidate	Source
1	pickup loaded by a low input impedance: LM386 50 kΩ, Daisy Seed 13.6–30 kΩ, SGTL5000 about 29 kΩ, passive DI 140–290 kΩ	covered: REQ.value/instrument holds the DC load to the net's load_min_ohm; an AC run with a pickup – its maker's DC resistance by row of catalog/tables/audio_sources.toml, the project's inductance – shows the resonance the load leaves (SIM.response)	CK-27, BL-16, FT-18	x; Seymour Duncan, DiMarzio, Mojotone product pages (r)
2	cable 66–164 pF/m; an op-amp buffer oscillates into the cable without a series resistor (about 100 Ω)	partly: a cable's c_pf_per_m × length_m loads the source in an AC run; the oscillation into the cable: no (no gain-bandwidth in the op-amp model)	CK-16, FT-2, FT-18	x
3	pot load and taper: 500k ∥ 500k ∥ 1 MΩ = 200 kΩ; "A" is log in the US and linear in Europe; 25k for EMG	no	PF, FT-12	x
4	treble loss with the volume turned down (about R/4 = 125 kΩ: 2.1 kHz with 600 pF); treble-bleed values	partly: an AC run with the pickup, the pot at its position in the load and the cable's capacitance holds the treble against an expectation; treble-bleed values: no rule	FT-12, FT-18	x
5	humbucker colour codes, phase and coil split	no – judgement	FT-18	x
6	cavity shielding not grounded; copper tape shorting the jack	partly: L-15 finds floating copper on a board	–	x
7	the string ground ties the player to the amplifier's ground; an RC in it needs a Y-class capacitor	no	HT-19	x
8	battery switched through the output jack's ring; EMG about 80 µA per pickup	no	HT-15, FT-11	x
9	headroom at 9 V against 18 V: TL072 about 1.5 V from each rail	no model	FT-3	x
10	piezo needs 10–20 MΩ and more; a charge amplifier needs its feedback resistor	partly: a piezo disc behind its capacitance (Murata 7BB, by row) in an AC run shows the bass the input's resistance leaves (SIM.response, corner)	CK-27, CK-25, FT-18	x; piezopro (r); Murata 7BB data sheets (r)
11	op-amp bias current through a bias resistor of megohms: NE5532 200–800 nA × 10 MΩ = 2–8 V offset	no model	FT-3	x
12	passive DI: the input impedance is the preamp's input × n² (12:1 gives 216–288 kΩ)	no	FT-12, FT-18	x
13	48 V phantom power reaching the circuit through 6.81 kΩ, 10 mA; hot-plug surges to −25…−30 V	partly: G-8 when the net's range holds 48 V	FT-8, CK-13	x
14	XLR pin 1 on signal ground (the pin-1 problem)	no	CK-13	x
15	LM386: inputs 50 kΩ to ground and ground-referenced; DC coupling from a biased stage	no	CK-27, FT-3	x; tardate Ruby (r)
16	LM386 oscillation: Zobel 10 Ω + 47 nF at pin 5, bypass at pin 7, decoupling at pin 6	partly: G-6	CK-13, CK-31	x; tardate Little Gem (r)
17	LM386 supply (N-1 and N-3 4–12 V, N-4 5–18 V), output capacitor 220–250 µF, a 9 V battery's internal resistance	partly: E-3 supply range, cell r	FT-1, FT-12	x
18	TDA2003 18 V maximum; TDA7052 bridge-tied with a fixed 40 dB	partly: E-3, G-3	CK-8	x
19	LM1875 and TDA2050: tab at V− on a grounded heat sink; stable only at a gain of 10 or more	partly: E-2 temperature with rth_ja_k_per_w	CK-28, HT-11	x
20	headphone amplifiers: 1 V into 32 Ω is 31 mA; TL072 wants ≥ 2 kΩ; a divider as virtual ground	no model	FT-3	x
21	tube filter capacitors below the unloaded B+ (330 VAC: 467 V, 513 V at +10 % mains); series capacitors without balancing	partly: G-8 when the net's range holds it	CK-29, FT-7	x
22	bleeders and standby: the charge stays with the tubes pulled; standby ahead of the reservoir	no	CK-24	x
23	heater supply: reference, elevation (12AX7 Vhk 180 V), 6.3 VAC too low for an LM317	no	out of scope	x
24	output transformer without a load: plate spikes arc	no	out of scope	x
25	fuses: primary at most V / R_primary, slow-blow; an HT fuse	partly: G-17	CK-10	x
26	tube amplifier ground returns; grid stoppers at the pin	no	CK-31; out of scope	x
27	death cap and ground switch: 0.047 µF from line to chassis	no	CK-10, HT-19	x
28	starved-plate 12AX7 at 9–24 V: models do not hold; heater 300 mA, cold inrush 2.5–7 ×	partly: E-3 source current, with a heater (FT-6)	FT-1, FT-6	x
29	high-voltage boost in pedals (12 → 200–300 V): MOSFET ≥ 400 V, light-load bursts in the audio band	partly: G-8	CK-22, HT-16, CK-31	x
30	relay coils on the pedalboard supply (TQ2-9V 15.5 mA, eight relays 124 mA); relays on the MCU's rail cause brown-outs	covered: E-3 pull-in, G-13; the dip: no	FT-1, BL-18	x; relay-bypass (r)
31	relay contacts at signal level need gold (minimum load 10 µA / 1 mV); DC across the contact pops	no	CK-30, BL-18	x; relay-bypass (r)
32	ground loops between loops, amplifiers and supplies	no – judgement	HT-18	x
33	a tuner or analyser loads the pickup	no	CK-27	x
34	codec line inputs against guitar level (120–850 mV peak); a TL072 does not work on 3.3 or 5 V	no	CK-6, CK-27, FT-3	PJRC Audio library (r); x
35	the MCU's ADC as guitar input: ESP32 windows, bias at the window's centre, anti-aliasing	no	CK-6, FT-12	ESP-IDF ADC (r); PJRC (r)
36	codec outputs: SGTL5000 HP_VGND never to ground; out-of-band noise; X7R and Y5V distort	no	CK-13, HT-17	x; PJRC (r)
37	USB audio: ground loop with the PC; ADuM3160 full speed only; its isolated converter needs 10 % load	partly: E-3 minimum load of a source	FT-5, HT-8	ADuM3160 USB isolator repository (r)
38	pedalboard power: shared returns; charge pumps at about 10 kHz	no	HT-16, HT-18	x
39	RF at high-impedance inputs: a series resistor first, then 47–100 pF at the jack	partly: H-7	HT-8, BL-16	x