EncoderHeads 2026

R tooling for the thermometry encoder heads deployed at Kohnen Station (Antarctica) and on the Greenland ice sheet.

Repository structure

EncoderHeads2026/
├── EncodingHeads.qmd          # Encoding documentation: binary → hex → physical units
├── rechner.qmd                # Message size / transmission budget calculator
├── meta/
│   └── HeadsUndKetten.xlsx    # Head and chain inventory
│
├── KohnenRecords_Analyse/     # Kohnen Station (Antarctica) — heads 03 & 04
│   ├── fetch_cloudloop.R      # Fetch MO messages from Cloudloop API
│   ├── decode_results3.R      # Decode from Cloudloop CSV export
│   ├── decode_combined.R      # Merge old + new CSV exports and decode
│   ├── head03_profile.qmd     # Analysis profile — head03 (35 nodes, modem 231709)
│   ├── head04_profile.qmd     # Analysis profile — head04 (25 nodes, modem 231710)
│   ├── explore_head04.R
│   ├── make_plots.R / plot_*.R
│   └── data/                  # Raw CSV exports and decoded results
│
└── Greenland2026/             # Greenland field campaign — heads 26-02…26-05
    ├── decode_core.R          # Shared decode + calibration functions (sourced by both docs)
    ├── decode_snowmelt.qmd    # ← SnowMelt head (IMEI 300434065508020): vertical melt profile
    ├── decode_chain.qmd       # ← Doppelkette (IMEI 301434062008130): borehole depth profile
    ├── tape_experiment.qmd    # Black-tape solar-absorption experiment (SnowMelt)
    ├── derive_mean_calibration.R  # Universal mean lab-S4 fallback coefficients
    └── data/                  # Field decode outputs (tracked; filenames keyed by message time)

Instruments

Head	Modem	Nodes	Fields per node	Deployment
26-02	228680	24	NTC1, NTC2, TestSB	Greenland 2026
26-03	234047	24	NTC1, NTC2, TestSB	Greenland 2026
26-04	234048	20	NTC1, NTC2, GND	Greenland 2026
26-05	228679	20	NTC1, NTC2, GND	Greenland 2026
head03	231709	35	NTC1, NTC2, GND, P	Kohnen (Ant.)
head04	231710	25	NTC1, NTC2, TestSB, GND	Kohnen (Ant.)

Heads 26-02 and 26-03 are SnowMelt sensors with nodes at 4 cm vertical spacing (node 1 = deepest).

Greenland field use

The decoder is field-driven and keyed solely by IMEI (the .bin filename prefix, e.g. 300434065508020-57.bin). Two devices send in parallel, each with its own report document sharing decode_core.R:

• SnowMelt head — IMEI 300434065508020 → decode_snowmelt.qmd
• Doppelkette — IMEI 301434062008130 → decode_chain.qmd

Workflow:

1. Save the .bin attachment(s) from the Cloudloop email into testdata/.
2. Open the matching document (decode_snowmelt.qmd or decode_chain.qmd) in RStudio. BIN_FILE already globs all .bin for that IMEI; a vector of files renders a time series. (A hex payload can still be pasted into MSG1 / MSG2 in the SnowMelt doc instead.)
3. Render (Ctrl+Shift+K) → profile, diagnostics, and CSV(s) saved to Greenland2026/data/.

The head's n_nodes / fields / type / spacing / calibration come from head_config[[IMEI]] inside each document — never derived from the message body. Counts→°C use the per-sensor lab Steinhart–Hart fit where a calibration table exists, else the universal mean lab-S4 curve (derive_mean_calibration.R); the legacy Beta curve is kept only for diagnostics.

Expected .bin message sizes (two firmware variants — every field is 0x7C-separated or concatenated; a defective sensor can add a few bytes):

Device (IMEI)	Nodes × fields	sep (B)	nosep (B)
SnowMelt (300434065508020)	24 × NTC1/NTC2/TestSB	299	225
Doppelkette (301434062008130)	20 × NTC1/NTC2/GND/Pressure	311	229

Dependencies

install.packages(c("dplyr", "tidyr", "ggplot2", "lubridate", "patchwork",
                   "httr2", "base64enc", "readr", "purrr"))