Gas sensor

Air blade (rev A)

A tiny plug-in board for the rev E phone-companion board. It stands upright in a notch in rev E's sensor end, between the two gas sensor cans, and pokes 4.5 mm out through a slot in the case face, so its tip sits in the outside air level with the gas sensors' caps. It carries:

• an outside-air temperature/humidity sensor (SHT40-CD1B, I²C address 0x46; an SHT40-BD1B at 0x45 works too, but never an -AD1B, which clashes with rev E's own). Rev E's own SHT40 (0x44) sits inside the case and reads the case air, warmed by the gas sensors' heaters; this one reads the street;
• the heated airflow thermistor (a 1 kΩ NTC). Moving air cools it; the firmware turns that into an airflow index for the gas detector (docs/HARDWARE.md section 1). The SHT40 beside it gives it the air temperature it needs.

Status: designed, not built. The board passes KiCad's checks (no design-rule errors, nothing unconnected, schematic and board in agreement).

Shape

5.7 × 14.9 mm, 2 layers, 1.0 mm thick.

• Tip: two prongs, 1 mm apart. The SHT40 and its capacitor are on one, the heated thermistor on the other, so the thermistor's warmth (2.7 mW) does not reach the SHT40. The gap runs down to just below the case face.
• Foot: five pads on the front face, just above two small shoulders. The shoulders rest on rev E's top surface; below them a tab drops into rev E's notch.
• No copper pours, and thin (0.15 mm) tracks: copper would carry the thermistor's heat away and blunt the airflow signal.

Parts

All three are assembled by JLC (LCSC numbers in production/bom.csv):

Ordering from JLCPCB

The blade is too small for JLC to assemble on its own, and JLC's own panel option ("Panel by JLCPCB") only does V-cut, which can't follow the forked tip or the shoulders. So order the ready-made panel in production/panel/: 6 blades in a 35 × 46 mm frame, with tooling holes and alignment marks for the assembly machine.

1. PCB: upload production/panel/gerbers.zip. 2 layers, 1.0 mm thick. Leave Delivery Format at "Single PCB": the file already is the panel, so the quantity counts panels (minimum 5 = 30 blades of bare board). 0.15 mm track and clearance, 0.45 mm vias with a 0.2 mm drill. Two of the vias sit in C1's pads: choose JLC's filled and capped via-in-pad option, as for rev E.
2. Assembly: top side only, production/panel/bom.csv and production/panel/cpl.csv. Assemble 2 panels (JLC's minimum) = 12 blades. Look over JLC's placement preview, especially each U1's pin-1 corner (positions matched to LCSC's footprints by gen/jlc_align.py).
3. Breaking out: each blade hangs on two small tabs of drilled holes on its long sides, below the fork. Bend each blade out with pliers, then file the two nubs flush so the blade slides through the case slot. The tips, the slot and the foot have no tabs, so they need no filing.

production/gerbers.zip, bom.csv and cpl.csv (one blade) are kept for reference.

Fitting it to rev E

1. Tin rev E's J3 pads (the five pads beside the notch between the cans, labelled BLADE) and the blade's five foot pads.
2. Put the blade's tab into the notch, parts facing the middle of rev E (the foot pads facing J3), with the shoulders flat on the board.
3. Check it stands square, then solder each foot pad to its J3 pad with a fillet of solder across the corner. Pad 1 is the one nearest the CH4 can.
4. Fit the board into the rev E front shell (hardware/enclosure, stl/reve_v2_front.stl) straight down, not tilted: the blade slides into a guide on the inside of the face and out through the slot between the sensor collars. The guide holds it over 5.8 mm, so knocks on the tip go into the case, not the solder joint.

Firmware

Nothing to set: the rev E firmware (pio run -d firmware/phone_board -e phone_board_reve) looks for an SHT40 at 0x45, then 0x46, at start-up. When it finds one, that supplies the temperature and humidity (to the phone and to the gas detector), the on-board SHT40 is logged as case_temp_c / case_rh_pct on USB serial, and the Info JSON says "t_src":2. Then run the still-air calibration (docs/HARDWARE.md section 1.6).

The design is generated

As for rev E: edit gen/design.py, then

Checks: kicad-cli pcb drc --schematic-parity air_blade.kicad_pcb and kicad-cli sch erc air_blade.kicad_sch. Last run (2026-10-10): 0 violations, 0 unconnected, 0 parity issues; ERC 0 errors, 0 warnings.

Not known yet

• How much the thermistor self-heats on the blade (an estimate is about 1 °C: some of its heat goes into the blade, not the air), and how strongly its reading follows airflow. The still-air calibration and the airflow survey in docs/PROTOCOL.md measure both.
• Whether the right-angle solder joint is strong enough for a fleet (the case guide takes side loads; a dab of epoxy along the shoulders would add more).