AirBox V3
AirBox cost per subsystem
Unit prices at the 10+ price bracket, looked up with KINA (LCSC / TME / Mouser) on 2026-11-02.
Price range across builds: $97.79 (ABI2, plug-in indoor Wi-Fi) to $176.79 (ABO4, outdoor eAQI with NO2 + O3). See section 2 for the build-by-build breakdown.
1. Subsystem unit costs
| ID | Subsystem | Source / MPN | Cost (USD) | Notes |
|---|---|---|---|---|
| B1 | Battery – Li-ion, 1S (min. 1 cell) | TME ACCU-INR18650-HJ2 (LG INR18650-HJ2, 3.0 Ah × 3.6 V = 10.8 Wh), €6.12 |
$0.65 / Wh; minimum pack 10.8 Wh = $7.04 | Cannot be charged below 0 °C. No balancer needed. |
| B2 | Battery – Na-ion, 2S (low-temp option, min. 2 cells) | Jauch NaCR-18650 (1.2 Ah × 3.0 V = 3.6 Wh per cell). Stocked sparingly in chinese stores. | ≈ $1.4–3.3 / Wh; minimum pack 7.2 Wh ≈ $10–24 (+ C2 balancer) | Typically chargeable down to about −20 °C. |
| B3 | Battery – LTO, 2S (low-temp option, min. 2 cells) | Ecolto Energy HC18650 (1.5 Ah × 2.4 V = 3.6 Wh per cell), DigiKey €3.41 | ≈ $1.09 / Wh; minimum pack 7.2 Wh = $7.84 (+ C2 balancer) | Typically chargeable down to about −30 °C, with very long cycle life. |
| E1 | Solar panel | Voltaic Systems P126 (6 V, 2 W ETFE), DigiKey | $19.05 | |
| C1 | Charger | BQ25798RQMR (LCSC C2876593) $2.41, APS0420M1R0A $0.13, 4× AO3400A $0.36, USB-C USB4125-GF-A $0.85, passives/connectors ≈ $0.39 | $4.13 | Solar + USB-C charging. USB-C input only (no battery path) ≈ $0.86. |
| C2 | Balancer | ETA3000D2I (LCSC C7465544) $0.38, APS0420M1R0A $0.13, passives/conn ≈ $0.08 | $0.59 | Required for 2S packs (Na-ion, LTO). |
| D1 | Power converter | TPS630701RNMR (LCSC C2876599) $1.69, APS0420M1R5A + APS0420M1R0A $0.26, 15× 10u + misc $0.26 | $2.21 | |
| M1 | nRF9151 + passives + Kyocera LTE antenna | nRF9151-LACA-R7 (Mouser) $19.14, KYOCERA AVX P822601 (LCSC C5348184) $1.99, TPD3F303DPVR $0.99, Nano-SIM SIM8075 (TME) $0.58, matching + decoupling + ferrite ≈ $0.49, U.FL $0.12 | $23.31 | LTE-M / NB-IoT cellular modem. |
| M2 | Alternative core: nRF7002 DK equivalent | nRF5340-QKAA-R7 (LCSC C3606967) $8.02, nRF7002-QFAA-R7 (TME) $3.58, 32 MHz + 40 MHz + 32.768 kHz crystals $0.42, Wi-Fi dual-band chip antenna WLA.10 (Mouser) $0.95, ≈ 40 passives ≈ $0.60 (est.) | $13.56 | Wi-Fi 6 + BLE, no LTE and no GNSS. |
| G1 | GPS antenna (for nRF9151), commercial | Commercial GNSS antenna with onboard LNA | $12.00 | |
| G2 | GPS antenna, self-made | Self-made QHA (quadrifilar helix) | $4.00 | Used in all builds below. |
| G3 | LoRa hardware and antenna | SX+LNA+Antenna | $6 | |
| S1 | Sensor – PM / T / RH / VOC / NOx / CO2 | Sensirion SEN66-SIN-T (Mouser) $54.92, Sensirion SEN5x jumper cable (GHR-06V → 2.54 mm, Mouser 403-SEN5XJUMPERCABLE) $6.03. | $61.06 | Indoor air quality. |
| S2 | Sensor – PM / T / RH / CO2 (outdoor) | Sensirion SEN63C-SIN-T (LCSC C52088895) $26.40, same SEN5x jumper cable $6.03 | $32.54 | No VOC / NOx (not used by eAQI). |
| S3 | NO2 sensor | SPEC 110-507 bare cell + own LMP91000 board (reproduced from the published 968-045 amplifier schematic) | $30.00 | < 20 ppb (60 s average). |
| S4 | O3 sensor | SPEC 110-406 bare cell + own LMP91000 board | $30.00 | < 20 ppb (60 s average). Needs NO2 subtracted. |
| S5 | SO2 sensor | SPEC SO2 bare cell + own LMP91000 board | $30.00 | 50 ppb (60 s average). |
| S6 | NO2 sensor – Alphasense | Alphasense NO2-B43F | ≈ $230 (≈ €200) | ppb-level, O3-filtered. Needs a low-noise front end (Alphasense ISB or own), not included. |
| S7 | O3 sensor – Alphasense | Alphasense OX-B431 | ≈ $230 (≈ €200) | Reads O3 + NO2; O3 is calculated by subtracting the paired NO2-B43F. Front end not included. |
| V1 | E-ink screen | Waveshare 3.97inch e-Paper HAT+ (G) | $20.00 | Optional display; not used in the builds. |
| L1 | LED matrix | 9×9 LED matrix, 81 LEDs at $0.10 each | $8.10 | Indoor builds, C5349955. |
| X1 | Misc parts, enclosure | $12 | Board, passives | |
| T1 | LTE data plan (2 years) | SIM $2.00 + 24 months × $0.38 (24-month commitment) | $11.12 | All LTE builds. |
2. Builds: cost vs. features
Totals are the sum of the subsystem rows above (IDs from section 1), using the self-made QHA GPS antenna (G2, $4). LTE builds include the SIM and a 24-month data plan (T1, $11.12).
2.1 Build table
| ID | Build | Use case | Subsystems | Total |
|---|---|---|---|---|
| ABI1 | Indoor – Wi-Fi | Homes, offices with Wi-Fi | B1, C1, D1, M2, S1, L1, X1 | $108.10 |
| ABI2 | Indoor – Wi-Fi, no battery (poor man's) | Permanently plugged in, cheapest product | USB-C input only ($0.86, part of C1), D1, M2, S1, L1, X1 | $97.79 |
| ABI3 | Indoor – LTE | Schools and other sites without usable networks | B1, C1, D1, M1, G2, S1, L1, X1, T1 | $132.97 |
| ABO1 | Outdoor – SEN63C | Outdoor PM / CO2; not fully eAQI compliant (no NO2, O3, SO2) | B3, C2, E1, C1, D1, M1, G2, S2, X1, T1 | $116.79 |
| ABO2 | Outdoor – SEN63C, winter shutdown | Same as ABO1, with a Li-ion battery instead of LTO. The device shuts down whenever the temperature is below 0 °C and resumes above it; data for those winter periods is interpolated | B1, E1, C1, D1, M1, G2, S2, X1, T1 | $115.40 |
| ABO3 | Outdoor – eAQI traffic | Roadside stations | ABO1 + S3 (SPEC NO2) | $146.79 |
| ABO4 | Outdoor – eAQI industrial / background | Industrial and background stations; traffic capable with good placement | ABO1 + S3 + S4 (SPEC NO2 + O3) | $176.79 |
Outdoor builds with SPEC sensors must keep them powered all the time (60 min stabilization, 30–45 µW).
2.2 Feature matrix
| Feature | ABI1 | ABI2 | ABI3 | ABO1 | ABO2 | ABO3 | ABO4 |
|---|---|---|---|---|---|---|---|
| Total (USD) | 108.10 | 97.79 | 132.97 | 116.79 | 115.40 | 146.79 | 176.79 |
| Connectivity | Wi-Fi 6 + BLE | Wi-Fi 6 + BLE | LTE-M / NB-IoT | LTE-M / NB-IoT | LTE-M / NB-IoT | LTE-M / NB-IoT | LTE-M / NB-IoT |
| Power | battery + USB-C | USB-C only | battery + USB-C | solar + battery | solar + battery | solar + battery | solar + battery |
| Operates below 0 °C | n/a (indoor) | n/a (indoor) | n/a (indoor) | yes (LTO, to ≈ −30 °C) | no (shuts down < 0 °C) | yes | yes |
| PM2.5 / PM10 / T / RH | yes | yes | yes | yes | yes | yes | yes |
| CO2 | yes | yes | yes | yes | yes | yes | yes |
| VOC / NOx | yes | yes | yes | no | no | no | no |
| NO2 | no | no | no | no | no | yes | yes |
| O3 | no | no | no | no | no | no | yes |
| SO2 | no | no | no | no | no | no | no |
| eAQI complete | no | no | no | no | no | traffic only | traffic + industrial/background |
| Visual indicator | LED matrix | LED matrix | LED matrix | no | no | no | no |
2.3 Cost deltas: what each feature costs
| Change | Delta | Notes |
|---|---|---|
| Drop battery + charger, USB-C only (ABI1 → ABI2) | −$10.31 | Cheapest product. |
| Wi-Fi core → LTE core + GPS antenna (M2 → M1 + G2) | +$13.75 | Independent of the local network. |
| Add 2-year LTE data plan (T1) | +$11.12 | Required for every LTE build. |
| Full indoor sensor → outdoor sensor (S1 → S2) | −$28.52 | Loses VOC / NOx. |
| Add solar panel (E1) | +$19.05 | Outdoor builds. |
| Li-ion → LTO 2S + balancer (B1 → B3 + C2) | +$1.39 | Enables charging down to ≈ −30 °C (ABO1 vs ABO2). |
| Each SPEC gas sensor (S3 NO2, S4 O3, S5 SO2) | +$30.00 each | Bare cell + own LMP91000 board. |
| Alphasense instead of SPEC (S6 / S7) | ≈ +$200 each | ppb-grade; low-noise front end not included. |
| E-ink display (V1, optional) | +$20.00 | Not used in any build. |
| Commercial GNSS antenna instead of self-made QHA (G1 vs G2) | +$8.00 | |
| Net: indoor LTE (ABI3) → outdoor base (ABO1) | −$16.18 | Solar +$19.05, LTO +$1.39, SEN63C −$28.52, no LED −$8.10. |
3. EAQI evaluation
The European Air Quality Index (EEA, bands revised in 2024 to follow the WHO 2021 guidelines)
uses hourly concentrations of PM2.5, PM10, NO2, O3 and SO2. Each pollutant gets a sub-index and the worst one sets the overall level
(Good / Fair / Moderate / Poor / Very poor / Extremely poor). A station's index is shown as complete only if it reports the minimum set for its type:
traffic stations need NO2 + PM, industrial and background stations need NO2 + O3 + PM. SO2 is used when available but is never required.
A pollutant's index is not determined if measurements are missing on three of the four previous days.
EAQI band edges converted to ppb (EEA reference conditions, 20 °C):
| Pollutant | Good / Fair | Fair / Moderate | Moderate / Poor | Poor / Very poor | Very poor / Extremely poor |
|---|---|---|---|---|---|
| NO2 | 5 ppb | 13 ppb | 31 ppb | 52 ppb | 78 ppb |
| O3 | 30 ppb | 50 ppb | 60 ppb | 80 ppb | 90 ppb |
| SO2 | 7.5 ppb | 15 ppb | 47 ppb | 71 ppb | 103 ppb |
A gas sensor is usable for the EAQI when its hourly-average noise and drift are a few ppb.
The noise figures below are 60 s averages. An hourly average divides random noise by about 7.7 (√60), but zero drift and temperature effects don't average out,
so every option needs a zero calibration and co-location with a reference station.
| Gas sensor option | Noise (60 s avg) | Hourly noise (random part only) | EAQI suitability |
|---|---|---|---|
| SPEC ULPSM 968-047 / 968-046 (module) | < 100 ppb (detection limit) | — | Not suitable. The detection limit is above the whole NO2 and O3 scales. |
| SPEC bare cell + LMP91000 (968-045 reference circuit) | < 20 ppb NO2 / O3 | ≈ 3 ppb | Indicative. Separates all NO2 bands except, marginally, Good/Fair (5 ppb), and all O3 bands. Needs per-unit zero and temperature calibration. |
| Alphasense NO2-B43F / OX-B431 | ≈ 7.5 ppb (1σ per reading) | ≈ 1 ppb | Best available. Established in ppb-level sensor networks. |
| SO2 (any SPEC option) | 50 ppb | ≈ 6.5 ppb | Marginal at the lower bands (7.5 / 15 ppb). SO2 is optional for every station type. |
SPEC NO2 filter life: the 110-507 cell's ozone filter is rated for 30–50 ppm·h, about 1–2 months at a typical outdoor 40 ppb O3.
After that the NO2 reading includes O3 (typically 30–60 ppb), which shifts the NO2 sub-index up by one or more bands.
A NO2-only build with a SPEC cell is therefore not suitable long-term. It needs an O3 sensor to subtract, or periodic cell replacement.
The SPEC O3 cell in turn reads NO2 (5 ppm NO2 → 4.2 ppm O3), so the SPEC NO2 and O3 cells are only usable as a pair.
Operating range: the SPEC cells are rated for best accuracy at −20…40 °C and 15–90/95 % RH, non-condensing. Hard frost, fog and condensation are outside that.
| ID | Pollutants | EAQI minimum met | Gas sensors | EAQI suitability |
|---|---|---|---|---|
| ABO1 | PM2.5, PM10 | None | — | PM sub-indices only. Shown as incomplete for every station type. |
| ABO2 | PM2.5, PM10 | None | — | Same as ABO1. No data below 0 °C, and interpolated values aren't measurements, so the index is undetermined after ~3 days of frost. |
| ABO3 | PM, NO2 | Traffic | SPEC NO2 (bare cell + LMP91000) | Indicative for ~1–2 months only, until the O3 filter is spent. |
| ABO4 | PM, NO2, O3 | Traffic, industrial, background | SPEC NO2 + O3 (bare cells + LMP91000) | Indicative. O3 corrects NO2 after filter wear, and NO2 corrects O3. Best-value EAQI build. |
All builds use low-cost sensors, so the result is an indicative EAQI, not a reference-grade one. SEN6x PM10 is derived from the
particle-size measurement rather than measured directly.
Note: Bare SPEC chemical sensors use the LMP91000. Reference implementation: SPEC Digital Gas Sensor Developer Kit
968-045 datasheet. Its schematic, BOM, Gerbers and firmware are published.
-
Front end: TI LMP91000 (potentiostat, I2C-configured), Renesas ISL60002 2.5 V reference, Microchip MCP604 op-amp, SiLabs SI7021 T/RH and a PIC24F16 MCU.
On the AirBox the nRF9151 replaces the PIC, and the SEN6x already provides T/RH. -
Resolution with this front end (page 2 table; noise std. dev. at 0 ppm, 1 Hz sampling, 60 s average):
Gas Range Resolution NO2 5 ppm < 20 ppb O3 5 ppm < 20 ppb SO2 20 ppm 50 ppb Accuracy is 15 % of reading using the sensor's bar-code sensitivity, and repeatability ±3 %.
-
Operation: 60 min warm-up. Keep the sensor always biased: for best accuracy, don't put the front end in standby between measurements.
Do a zero calibration in clean air after warm-up, and use a running average (up to 300 samples) to reduce noise. -
Front-end cost (LCSC/Mouser, 10+): LMP91000SDX $4.31 per sensor, plus a shared ISL60002 reference $5.13 and a shared MCP604 $1.32.
4. School requirements (Romania)
Target build: ABI3 (Indoor – LTE, $132.97 including 2 years of LTE). It already covers every item below.
| Requirement | Why | How the AirBox meets it |
|---|---|---|
| CO2 is the primary measurement | Classroom CO2 rises quickly with 25–30 pupils and closed windows, and it is the best proxy for ventilation. Thresholds: < 1000 ppm good, 1000–1500 ppm ventilate, > 1500 ppm too high (1500 ppm is cited as the Romanian maximum in local studies; confirm against the current Ministry of Health hygiene norms for schools). | SEN66 (NDIR CO2). |
| Visible indicator for the whole class | Teachers and pupils must see at a glance when to open the windows, without an app. | 9×9 LED matrix: traffic-light colour plus CO2 level, readable from the back of the room. |
| PM2.5 / PM10 | Outdoor pollution and heating season (wood/coal stoves) enter through ventilation. Tells whether opening the window helps or hurts. | SEN66. |
| VOC (nice to have) | Cleaning products, paints and markers. | SEN66 VOC index, at no extra cost. |
| Temperature / humidity | Overheated rooms and dry winter air. | SEN66. |
| LTE, independent of the school network | School Wi-Fi is often missing, unreliable, captive-portal or firewalled. Enterprise Wi-Fi (802.1X) onboarding per school doesn't scale. | nRF9151 LTE-M / NB-IoT with its own SIM. No school IT involvement. |
| Cannot be switched off to hide bad air | Readings must stay honest even when they are uncomfortable for staff. | No power switch. The buttons only change the display and can't turn it off. On unplugging, the battery keeps it measuring and it reports a "power removed" event over LTE. Gaps are shown as gaps on the dashboard, never filled in. |
| Tamper evidence | Covering, opening or moving the device must be visible. | Case-detect switch (SW3) reports enclosure opening. The battery-backed event log is sent over LTE. |
| Theft resistance | The device sits unattended in public rooms. | Custom device with no consumer resale value. GPS position reported if moved. SIM locked to the device. |
| Mount above the AC unit, powered from its socket | The AC indoor unit and its socket sit high on the wall (≈ 2–2.5 m), out of pupils' reach. The socket is usually on its own circuit and rarely unplugged, so the device is hard to unplug or steal, and it needs no new wiring. Most wall split units draw room air in through a grille on top, so the air above the unit is well-mixed room air. The LED matrix is visible from the whole room. | Wall-mounted above the AC indoor unit, powered by a USB-C adapter in the AC socket (with a splitter if needed) and a short cable into the bracket. Conditions: (1) keep the clearance above the unit required by the AC manual (typically ≥ 15 cm), or offset sideways so the top intake isn't blocked, and leave room to lift the front panel for filter changes. (2) Never mount in the outlet airflow at the bottom front. Caveats: near the ceiling the temperature reads a few °C high in winter, while CO2 and PM are barely affected. Some AC circuits have an isolator switch or breaker that staff may turn off seasonally; the battery then covers the gap and a "power removed" event is reported. Not every classroom has AC. |
| Privacy (GDPR) | Installed in rooms with minors. | No camera or microphone. Only environmental data is collected. |