fieldnode

---
doc_id: FND-CAL-001
title: FieldNode sizing calculations
project: FieldNode
doc_type: Calculation
version: "0.2"
status: Draft
date: '2026-09-25'
author: Amish Chadha
license: CERN-OHL-S-2.0
revisions:
- version: "0.1"
  date: '2026-09-25'
  author: Amish Chadha
  change: First issue for TRL 3 (core consumption, energy budget and autonomy, airtime, link budget, store and forward, enclosure temperature and charging window, panel voltage, wind and mounting, installation and service, mass, cost)
- version: "0.2"
  date: '2026-09-25'
  author: Amish Chadha
  change: Recommendations accepted by Amish (DDR-002)
---

# FieldNode sizing calculations

With the decisions of FND-DDR-002 applied, FieldNode meets fifteen of its eighteen requirements on paper (ten by calculation, five by design), has two at risk, misses none and leaves one that only a timed installation can settle. Version 0.1 found one miss, heat: in full sun at 45 °C ambient the base enclosure reaches 58.7 °C clean and 66.2 °C dusty on a hot design day, and up to 73.3 °C with the sun in its worst position, against the 60 °C of R3, and the cell is then too hot to charge for most of the day. Amish decided on 2026-09-25 to fit a ventilated white sun shield at hot-climate sites only. This note shows that the shield is needed where the site's design maximum exceeds 30 °C; with it the inside peaks at 48.5 to 52.2 °C at 45 °C ambient and the node stores 10.8 to 13.1 Wh on a hot clear day. The shield is BOM line 14, an option at $8.00 and 0.15 kg; the hot-climate node costs $134.00 and weighs 2.55 kg, and R14 now applies to the base node (2.41 kg). The panel is now 9 V class, which keeps 2.42 V of headroom above a typical charger's 5 V minimum input when hot, where the 6 V class panel had none. The published sensor allowance is 100 mW, which gives 5.75 days of autonomy, a 15 % margin; a cold or aged cell still falls short of 5 days (R6 at risk), and the cell will not charge on clear days colder than about -13 °C (R2 at risk). A firmware rule lengthens the reporting interval at SF10 and slower, so airtime stays within 30 s a day on The Things Network. Every number in this note is printed by `docs/04-calcs/sizing.py`; the tag in brackets, for example [A3], is the line of that script's output that carries it.

> **Safety:** These are first-principles estimates for a paper proof of concept. They do not show that the lithium iron phosphate cell, the charger or the pole mounting is safe. Cell temperature, charge lockout and clamp preload must be checked on hardware before any node is left unattended. See FND-PRC-001, Safety.

## Scope and method

The note checks every requirement in FND-REQ-001 v0.4 against the design in FND-PRC-001 v0.4 and the parametric model `cad/src/model.py`. The script imports the model's `PARAMS` and `derived()` and builds the model once for part volumes, so the enclosure, panel position, bracket, back plate and V-blocks used here are the ones in the STEP files and in drawing FND-DWG-001. It also reads `bom/bom.csv` and `budget_usd` in `project.yaml`. Run it from the repo root with `python docs/04-calcs/sizing.py`; it also writes `docs/04-calcs/results.csv`.

The design case is a node on a 48.3 mm pole with the enclosure base 1.75 m above ground, facing the equator, reporting a 20-byte LoRaWAN uplink every 15 min (DDR-001, D6) to a gateway 2 km away in suburban terrain.

## Assumptions

*Table 1. Main assumptions.*

| Area | Assumption | Basis |
| --- | --- | --- |
| Energy chain | 6 W panel; 2 peak sun hours on the tilted panel in the worst month; 20 % panel loss (heat, dust, angle); charger 85 %; cell charge 95 %; rail converters 90 % | As FND-PRC-001 v0.2; screening values |
| Cell | LiFePO4 3.2 V, 6 Ah; 80 % usable; 70 % of capacity at -20 °C; 80 % at end of life; charging allowed 0 to 45 °C | Typical cell data (see FND-PRC-001) |
| Core | SX1262 at +14 dBm 45 mA, receive 4.6 mA; two 0.1 s receive windows; controller awake 0.5 s at 8 mA per report; 33 µA whole-node sleep and quiescent current at 3.3 V | Semtech SX1261/2 datasheet figures; the rest assumed |
| Radio | 20-byte payload plus 13 bytes of LoRaWAN overhead; 125 kHz; coding rate 4/5; 8-symbol preamble; explicit header; CRC on | LoRaWAN defaults |
| Link | +14 dBm; 2 dBi antennas; 0.5 dB node and 2 dB gateway feeder loss; 6 dB receiver noise figure; Okumura-Hata suburban model at 868 MHz; 10 dB fade margin | Screening values; gateway sensitivity assumed equal to the node's |
| Storage | 32 bytes per stored reading; 16 MB SPI flash; 24 bytes per reading when several are batched into one uplink | Assumed record format |
| Heat | Combined convection and radiation 10 W/m²K (still air); absorptance 0.45 clean light grey, 0.70 dusty; clear-sky direct beam from the Meinel air-mass model, diffuse 10 % of direct, ground albedo 0.2; back face against the mounting plate; panel shading traced from the model geometry; the ventilated white shield (BOM line 14, 0.5 mm aluminium standing 15 mm off the box) passes a quarter of the solar gain | Handbook ranges; the shield factor matches WWT-CAL-001 and is not measured |
| Design days | Hot: latitude 28.6° N, late May, clear, 30 to 45 °C. Cold: latitude 50° N, late January, clear, -20 to -10 °C | Chosen to bracket R2 |
| Panel voltage | 9 V class panel (FND-DDR-002), Vmp 9.0 V at 25 °C; the 6 V class panel of v0.1 kept for comparison; -0.35 %/K; panel 30 K above ambient in full sun; charger minimum input about 5.0 V | Typical crystalline silicon and bq24650 or LT3652 class figures; to confirm per part at TRL 4 |
| Wind | 35 m/s gust; air 1.225 kg/m³; drag coefficient 1.2 on the panel (normal to it) and 1.3 on the enclosure | Screening values, not a code check |
| Mounting | Band clamp preload 1,000 N; each clamp presses the V-block with twice its preload; friction 0.2; 6063-T5 class flat bar, E = 69 GPa | Preload assumed, to be measured |
| Mass | Made parts from model volumes (polycarbonate 1.2, aluminium 2.7, ASA 1.07 g/cm³), shield included from its model volume plus 0.02 kg of fixings; bought parts from typical catalogue masses (panel 0.55 kg, cell 0.15 kg); the 9 V class panel assumed the same mass as the 6 V class | Estimates |

## A. Core consumption and energy budget (R5, R6, R7)

- **Core.** Each report costs 16.0 mA·s at SF9, so 96 reports use 0.43 mAh a day and sleep adds 0.79 mAh: the core needs 4.0 mWh a day [A1]. The TRL 2 figure stands.
- **Worst month.** The panel sees 12.0 Wh, delivers 9.60 Wh, the charger passes 8.16 Wh and the cell stores 7.75 Wh a day [A2].
- **Allowance and autonomy.** The cell holds 15.36 Wh usable. A sensor allowance of 115.0 mW is the largest that gives exactly 5.0 days without sun [A3]; at that load the node draws 3.07 Wh a day and the margin on R6 is zero. At -20 °C the same cell lasts 3.50 days and at end of life 4.00 days [A4].
- **At 100 mW,** the published allowance (FND-DDR-002) and the R7 target, the node draws 2.67 Wh a day, stores 2.9 times that in the worst month, and lasts 5.75 days, a 15 % margin on R6, or 4.03 days at -20 °C and 4.60 days at end of life. After five sunless days it refills in 2.6 worst-month days [A5].
- **Energy neutrality.** The worst month supports sensor loads up to 291 mW before the cell runs down day by day [A6]; autonomy, not harvest, limits the allowance.
- **Sibling loads.** AirStreet (45 mW), NoiseMap (21 mW), FloodGauge (7 mW), WellSense, HeatMap Node and SlopeWatch (about 1 mW or less) all fit within the published 100 mW. CurbCount's 300 mW does not [A7]; this is a cross-repo action in `docs/REVIEW.md`.

## B. Airtime, link and store and forward (R8, R9, R10)

*Table 2. Time on air for a 20-byte uplink (33 bytes on air) at 125 kHz [B1].*

| Spreading factor | Per uplink | Per day at 15 min | Shortest interval within 30 s/day | EU868 1 % off-time |
| --- | --- | --- | --- | --- |
| SF7 | 71.9 ms | 6.9 s | 4 min | 7 s |
| SF8 | 133.6 ms | 12.8 s | 7 min | 13 s |
| SF9 | 246.8 ms | 23.7 s | 12 min | 25 s |
| SF10 | 452.6 ms | 43.5 s | 22 min | 45 s |
| SF11 | 987.1 ms | 94.8 s | 48 min | 99 s |
| SF12 | 1,810.4 ms | 173.8 s | 87 min | 181 s |

- **R9 is met on paper by a firmware rule.** The 15 min default fits The Things Network's 30 s a day at SF7 to SF9, which covers the R8 design case, but not at SF10 or slower. Under the rule decided in FND-DDR-002, firmware on The Things Network lengthens the interval to the shortest value in Table 2 when adaptive data rate moves a node to SF10 or slower: 22 min at SF10, 48 min at SF11 and 87 min at SF12, which keeps airtime at 30.0 s a day or less [B1b]. On a private TwinKit gateway only the regional duty cycle applies, and the 15 min interval fits it at every spreading factor. The rule is a firmware requirement; the firmware itself is TRL 4 work.
- **Power limit.** At +14 dBm with a 2 dBi antenna the node radiates 15.5 dBm EIRP, 13.35 dBm ERP, inside the 14 dBm ERP limit of EU868; the whip's middle sits 1.64 m above ground [B2].
- **R8 is met on paper.** At SF9 the link budget is 145.0 dB. The Hata suburban path loss at 2 km to a 30 m gateway is 126.4 dB, a margin of 18.6 dB; with a 10 dB fade margin the range is 3.5 km. A 15 m gateway, below Hata's range of validity, gives about 2.5 km. At SF7 the margin is 13.6 dB and the range with fade margin 2.5 km [B3]. Clutter, foliage and gateway height dominate these figures; only a field survey settles them.
- **R10 is met on paper.** A day of readings takes 3,072 bytes; 30 days take 90 kB of the 16 MB flash, which would hold 5,461 days [B4].
- **Forwarding the backlog.** After a 30-day outage the node holds 2,880 readings. At SF9, batching four per uplink, resending them takes 428 s of airtime: 68 days at the 6.3 s a day left under fair use, but half a day under the EU868 duty cycle alone. At SF7 it takes 5 days under fair use [B5]. The store is ample; on The Things Network the backlog is slow to drain, so old readings should be sent when a private gateway is in range or thinned.

## C. Enclosure temperature and the charging window (R2, R3, R4, R5)

- **Loss coefficient.** The enclosure's exposed faces (all but the back, which sits on the mounting plate) total 0.0930 m² and lose 0.930 W/K. About 852 J/K of enclosure, cell, boards and plate gives a 15 min time constant, so the inside follows the sun within the hour [C1].
- **What the panel hood does.** The panel's front edge overhangs the lid by 56.6 mm and sits 120.7 mm above the enclosure top [C1b]. With the sun 60° high in front it shades 86 % of the top but none of the front face [C2b]. The hood protects the lid seam from rain; it does little for heat when the sun is low or to the side.
- **Worst sun position.** At 45 °C ambient, the worst clear-sky sun position is 35° high and 30° off the front normal: 16.9 W absorbed and 63.3 °C inside when clean, 26.3 W and 73.3 °C when dusty [C2]. This is an upper bound, since the sun is not always in that position at the day's peak ambient. With the shield the same position gives 49.6 °C clean and 52.2 °C dusty [C2c].
- **Where the shield is needed.** Without the shield the dusty box stays at 60 °C or less in the worst sun position up to 31.7 °C ambient, so the shield is fitted where the site's design maximum ambient exceeds 30 °C [C2c]. At that limit, on a clear 15 to 30 °C day, the base node peaks at 44.2 °C clean and 51.2 °C dusty and still stores 19.0 to 25.5 Wh against 2.67 Wh drawn [C3c].

*Table 3. Hot design day, 30 to 45 °C, latitude 28.6° N, clear [C3].*

| Case | Peak inside | Sun hours with charging allowed | Energy stored in the day |
| --- | --- | --- | --- |
| Clean, no shield | 58.7 °C | 2.1 of 13.3 h | 0.8 Wh |
| Dusty, no shield | 66.2 °C | 1.4 of 13.3 h | 0.3 Wh |
| Clean, with shield (hot-climate option) | 48.5 °C | 7.4 of 13.3 h | 13.1 Wh |
| Dusty, with shield (hot-climate option) | 50.4 °C | 6.5 of 13.3 h | 10.8 Wh |

- **R3 is met on paper with the shield at hot-climate sites.** Without the shield the clean enclosure scrapes under 60 °C on the 45 °C design day but not with dust or with the sun in its worst position, which is why R3 now requires the shield where the design maximum exceeds 30 °C (FND-REQ-001 v0.4). The result rests on the assumed shield factor of 0.25, which only a test can confirm.
- **Hot weather stops charging.** Because the 45 °C charge lockout (R4) acts on a cell that sits above 45 °C from mid-morning, a clean node stores 0.8 Wh on a hot clear day against 2.67 Wh drawn. A run of such days loses 1.84 Wh a day, so a full cell lasts 8.3 days (6.4 days dusty). With the shield the day ends 10.4 Wh in credit (8.1 Wh dusty) [C3b]. The lockout is correct and must stay; the fix, now decided, is to keep the cell cooler with the shield at hot sites.
- **Cold weather.** On a clear day at 50° N with -20 to -10 °C ambient, sun on the box lifts the inside to 3.3 °C, which allows 4.2 h of charging and 8.7 Wh stored [C4]. The peak rise is 13.3 K, so on clear days whose maximum is below about -13 °C the cell never reaches 0 °C and cannot charge [C5]. Overcast cold spells longer than the cold autonomy of about 4 days (section A) will stop the node.
- **Panel voltage.** In full sun on the hot day the panel runs at about 75 °C. The 9 V class panel decided in FND-DDR-002 then gives a Vmp of 7.42 V, 2.42 V above a typical 5.0 V charger minimum input; the 6 V class panel of v0.1 fell to 4.95 V, 0.05 V short [C6]. The part choice for the power board (TRL 4 work, on hold) must confirm the minimum input and the charger's maximum input.
- **R2 is at risk at the cold end.** With the shield at hot sites the warm end is covered (52.2 °C dusty worst case at 45 °C, and 58.3 °C for an unshielded node at 30 °C, both under the 70 °C electronics rating), but the cell charges only on clear days warmer than about -13 °C.

## D. Wind and mounting (R12, R13)

- **Loads.** A 35 m/s gust gives a dynamic pressure of 750 Pa: 52.2 N normal to the panel, 29.3 N on the enclosure front-on and 17.6 N side-on [D1].
- **Overturning.** With wind from behind, the panel lifts and pushes forward; together with the node's weight this gives 14.8 N·m about the lower clamp and pulls the upper clamp off the pole with 59 N. Wind from the front presses the node onto the pole [D2]. Against the 2,000 N that a 1,000 N band preload holds, the factor is 33.9 [D3]. With the shield fitted the enclosure presents 36.3 N front-on and the worst clamp pull rises to 63 N, a factor of 31.8 [D3b].
- **Slip.** Weight and wind push the node down the pole with 64 N against 800 N of friction (factor 13); side wind twists it with 1.58 N·m against 19.3 N·m (factor 12) [D4]. These factors depend on the assumed preload, which a torque check at installation must confirm.
- **Bracket.** The rear posts are 170 mm long at 89° and the front struts 171 mm at 50°. Each member carries at most about 41 N, and a 25 x 3 mm flat bar of that length buckles out of plane at 1,317 N, a factor of 32 [D5].
- **Pole range.** On a 40 mm pole the V contacts sit 14.1 mm either side of the center and on a 60 mm pole 21.2 mm, both inside the 50 mm V-block, which seats poles up to 71 mm. Each band needs about 197 to 248 mm of length around the pole and block [D6], [D6b].
- **R13 is met on paper**, subject to the preload assumption. The pole, its footing and any wall anchors are site supplied and outside this note; the node adds about 81 N of wind load about 2 m up the pole, which the site owner should check.

## E. Installation and service (R12, R15)

- **Installation.** Fitting the bracket and panel on the ground, lifting the node, tightening two clamps, aiming the panel, plugging in the sensor and confirming an uplink take an estimated 15 min with a helper [E1], exactly the R12 limit. Only a timed installation can verify it.
- **Cell swap.** Opening the lid, swapping the cell in its fused holder, renewing the desiccant and confirming an uplink take an estimated 7 min with a screwdriver and no soldering [E2]. R15 is met by design.

## F. Mass and cost (R14, R16)

- **Mass.** The enclosure body and lid weigh 0.43 kg, the printed ASA internal plate 0.08 kg, the bracket 0.14 kg and the back plate with V-blocks 0.63 kg; bought parts add 1.14 kg: 2.41 kg in all for the base node [F1]. The TRL 2 figure of about 1.7 kg left out most of the back plate. R14, now stated for the base node, is met on paper with 0.09 kg of margin; the margin is thin and rests on catalogue masses. The shield, 181 x 106 x 206 mm of 0.5 mm aluminium, weighs 0.15 kg with fixings, so a hot-climate node weighs 2.55 kg [F1b].
- **Cost.** The BOM has 14 lines, all priced; line 14, the shield, is an option at quantity 0 in the base node. The base node totals $126.00 against the $150 `budget_usd`, a margin of $24.00 (16 %) [F2]; the 9 V class panel is priced as the 6 V class was. A hot-climate node with the shield costs $134.00 and weighs 2.55 kg [F3]. R16 is met on paper for both.

## L. Results against every requirement

*Table 4. Requirement status from this note [L].*

| ID | Requirement | Value | Target | Status |
| --- | --- | --- | --- | --- |
| R2 | Operating temperature | 52.2 °C dusty worst case with the shield at 45 °C; 58.3 °C unshielded at 30 °C; no charging on clear days below about -13 °C | -20 to +45 °C ambient; electronics -20 to +70 °C inside | **At risk** (cold charging) |
| R6 | Autonomy without sun | 5.75 days at the published 100 mW (15 % margin); 4.03 days at -20 °C; 4.60 days at end of life | 5 days or more | **At risk** (cold or aged cell) |
| R12 | Mounting and install | Seats 40 to 71 mm poles; installation estimated at 15 min | 40 to 60 mm poles and walls; 15 min or less | Not verifiable at TRL 3 (time); fit met by design |
| R3 | Interior temperature in sun | With the shield at 45 °C: 48.5 °C clean, 50.4 °C dusty on the hot day, 52.2 °C worst sun position; unshielded at 30 °C: 58.3 °C worst sun position (at 45 °C: 58.7 to 73.3 °C) | 60 °C or less; shield where the design maximum exceeds 30 °C | Met on paper (shield factor assumed) |
| R5 | Energy neutral in the worst month | 7.75 Wh stored against 2.67 Wh drawn; 10.8 to 13.1 Wh stored on a hot clear day with the shield; 9 V class Vmp headroom 2.42 V when hot | Harvest exceeds demand at 2 peak sun hours | Met on paper |
| R7 | Sensor power allowance | Published allowance 100 mW; 115.0 mW would give exactly 5 days | 100 mW or more | Met on paper |
| R8 | Radio link | 18.6 dB margin at 2 km, SF9, 30 m gateway; 3.5 km with 10 dB fade margin | 2 km suburban at SF9 or faster | Met on paper |
| R9 | Airtime within fair use | 23.7 s a day at SF9 and 15 min; firmware rule gives 22 min at SF10 and 87 min at SF12; 30.0 s a day or less | 30 s a day or less on The Things Network | Met on paper (firmware rule) |
| R10 | Store and forward | 30 days in 90 kB of 16 MB | 30 days or more | Met on paper |
| R11 | Standard sensor interface | Two M12 5-pin ports: switched rail, ground and three signal pins | Two sealed ports with a bus, analog input and one switched rail each | Met on paper (pinout open, DDR-001 O2) |
| R13 | Wind | Clamp pull 59 N (63 N with the shield) against 2,000 N; slip factor 13; bracket factor 32 | 35 m/s without loosening | Met on paper (preload assumed) |
| R14 | Mass | Base node 2.41 kg (0.09 kg margin); 2.55 kg with the shield, which R14 excludes | 2.5 kg or less, base node | Met on paper |
| R16 | Cost | $126.00 base node; $134.00 with the shield | $150 or less for the FieldNode core | Met on paper |
| R1 | Weather protection | IP65 box and ePTFE vent; IP67 class glands, capped ports and bulkhead | IP65; sealed penetrations | Met by design |
| R4 | Safe charging window | NTC gates the charger at 0 and 45 °C | Charging blocked outside 0 to 45 °C | Met by design |
| R15 | Serviceable | Cell swap in about 7 min | 10 min or less, no soldering | Met by design |
| R17 | Open and independent | CERN-OHL-S-2.0 and MIT; standard LoRaWAN | Open files, any network server | Met by design |
| R18 | Privacy pass-through | Core forwards only what the sensor firmware passes | No images or audio leave the node | Met by design |

Counts: none not met, 2 at risk, 10 met on paper, 5 met by design, 1 not verifiable at TRL 3. In v0.1 the counts were 1 not met (R3), 5 at risk (R2, R5, R6, R9, R14) and 6 met on paper; R3, R5, R9 and R14 moved to met on paper because of the decisions in FND-DDR-002. R3 and R5 now rest on an assumed shield factor that only a test can settle.

## Checks against the TRL 2 figures

| TRL 2 claim (FND-PRC-001 v0.2) | This note | Action |
| --- | --- | --- |
| 7.8 Wh/day stored; 3.1 Wh/day drawn | 7.75 and 3.07 Wh/day at 115 mW | Stands |
| Sensor allowance about 115 mW | 115.0 mW gives exactly 5.0 days; 100 mW published (FND-DDR-002) | Precis updated |
| Autonomy about 5.0 days | 5.00 days at 115 mW, 5.75 days at 100 mW; 3.5 to 4.6 days cold or aged | Precis updated |
| Core about 4 mWh/day | 4.0 mWh/day | Stands |
| Airtime 24 s at SF9, 43 s at SF10 | 23.7 s and 43.5 s | Stands |
| Link budget about 145 dB | 145.0 dB; 18.6 dB margin at 2 km | Precis updated |
| Wind about 52 N, about 16 N·m | 52.2 N; 14.8 N·m about the lower clamp | Precis updated |
| Mass about 1.7 kg | 2.41 kg | Precis updated; back plate was under-counted |
| Cost about $126 | $126.00 | Stands |
| Panel hood "helps" with interior heat | Shades 86 % of the top but not the front; R3 not met without a shield; charging blocked in hot weather | Precis updated; shield option for sites above 30 °C (FND-DDR-002) |
| One gland fitted, one blanked | The panel lead needs a gland; both glands are fitted | BOM line 3 updated |