Goose/Moon Sensors/Moon Temp Calibration

Moon Temp Sensor Offset Analysis

Moon Temp — Sensor Offset Analysis

Outdoor thermistor array for moon observation sessions. Three MF58 10 kΩ NTC sensors read simultaneously by an ADS1115 ADC on an ESP32-C3. Device publishes raw 16-bit ADC counts over MQTT and logs to daily CSVs on NAS (/mnt/nas/rooster/moon-temp-logs/).

Hardware: 3× NTC on ADS1115 (channels 0–2) / ESP32-C3
ADC LSB: 0.13108 mV/count (calibrated; nominal 0.125 mV)
Current offsets (ESP32-C3 flash): ADC0 = 0 · ADC1 = +22 · ADC2 = −64

Log Eras

Period Path VCC v_rail column
May–Jun 2026 logs_2.3V/logs/ 2.3 V (hardware fault) absent
Jul 2026+ raw_logs/ ~3.3 V measured per-row present

Converted temperature CSVs (pre-computed °C) live in converted_logs/.

Code 
%matplotlib inline
import os, platform
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates

NAS_BASE = (
    "/Volumes/AlphaDoom_6/rooster/moon-temp-logs"
    if platform.system() == 'Darwin'
    else "/mnt/nas/rooster/moon-temp-logs"
)

plt.rcParams['figure.figsize'] = (14, 4)
plt.rcParams['figure.dpi'] = 110

Offset Methodology

Calibration aligns three channels against ADC0 (reference, offset always 0).

Step 1 — Burst alignment

  1. Start the logger at a 5-second interval: moon_temp_logger --interval 5
  2. Manually send offset commands to ADC1 and ADC2 until all three channels match ADC0. Example: mosquitto_pub -h 10.0.0.50 -t moon-temp-001/offset/cmd -m '{"adc":1,"offset":16}'
  3. Stop the 5s logger

Step 2 — Overnight verification

  1. Start systemd logger: sudo systemctl start moon_temp_logger
  2. Run overnight (≥ 8 h) to capture the full thermal range
  3. Run the Running Calibration Confirmation section at the bottom to check alignment

Heat Response Test

Sensor liveness check: manual heat applied to each thermistor in sequence. Window: 2026-05-27, 09:51–09:58 MDT (~7 minutes) Interval: 1-second sampling Sequence: ADC0 → ADC1 → ADC2, ~1 minute each Response: NTC thermistors decrease resistance when heated → lower ADC counts

Code 
TEST_START = '2026-05-27 09:51:00'
TEST_END   = '2026-05-27 09:59:00'

df_full = pd.read_csv(
    os.path.join(NAS_BASE, "logs_2.3V/logs/2026-05-27.csv"),
    parse_dates=['timestamp']
)
df_full = df_full.sort_values('timestamp').reset_index(drop=True)

test = df_full[(df_full['timestamp'] >= TEST_START) & (df_full['timestamp'] <= TEST_END)].copy()

print(f"Rows : {len(test)}")
print(f"Span : {test['timestamp'].iloc[0]}  →  {test['timestamp'].iloc[-1]}")
print(f"Sample interval (median): {test['timestamp'].diff().dt.total_seconds().median():.0f}s")
print()
for col in ['adc0','adc1','adc2']:
    d = test[col]
    print(f"{col}: min={d.min()}  max={d.max()}  mean={d.mean():.1f}  std={d.std():.1f}")
Rows : 385
Span : 2026-05-27 09:51:26  →  2026-05-27 09:58:15
Sample interval (median): 1s

adc0: min=10880  max=12848  mean=11826.8  std=831.8
adc1: min=11009  max=12769  mean=12122.6  std=757.5
adc2: min=11369  max=12841  mean=12509.9  std=543.2
Code 
# Per-channel subplots
fig, axes = plt.subplots(3, 1, figsize=(14, 9), sharex=True)
colors = {'adc0': 'steelblue', 'adc1': 'darkorange', 'adc2': 'crimson'}
labels = {'adc0': 'ADC0 (hand first)', 'adc1': 'ADC1 (hand second)', 'adc2': 'ADC2 (hand third)'}

for ax, col in zip(axes, ['adc0', 'adc1', 'adc2']):
    ax.plot(test['timestamp'], test[col], color=colors[col], linewidth=1.0, label=labels[col])
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%H:%M:%S'))
    ax.set_ylabel('ADC counts')
    ax.legend(loc='upper left')
    ax.set_title(labels[col])

axes[-1].set_xlabel('Time (MDT)')
fig.suptitle('Heat Response Test — 2026-05-27 ~09:51–09:58 MDT', fontsize=13)
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Code 
# Overlay — sequential drops clearly visible
fig, ax = plt.subplots(figsize=(14, 5))
for col in ['adc0', 'adc1', 'adc2']:
    ax.plot(test['timestamp'], test[col], color=colors[col], linewidth=1.0, label=labels[col], alpha=0.85)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%H:%M:%S'))
ax.set_ylabel('ADC counts')
ax.set_title('Heat Response Test — all channels overlay')
ax.legend()
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Hardware Reference

Sensors: 3× MF58 10 kΩ NTC, B = 3950 K, R₀ = 10 kΩ @ 25 °C
Series resistor: 10 kΩ fixed
ADC: ADS1115 (I²C 0x48), PGA ±4.096 V, 16-bit
ADC LSB (calibrated): 0.13108 mV/count (nominal 0.125 mV — ~4.9 % reference offset measured against multimeter)
Firmware: ESP32-C3, Embassy async, MQTT 3.1.1

Power Chain (updated July 2026)

Stage Detail
Mains 115 V AC → 5 VDC wall adapter
Variable regulator Adjustable → 5 V rail
ESP32-C3 Powered directly from 5 V rail; 3.3 V pin supplies thermistor VCC
Buck converter 5 V → 3.3 V dedicated supply for ADS1115 VDD
Rail monitor ADS1115 AIN3 wired to ADS VDD (3.3 V from buck) — logged as v_rail

Circuit Diagram

moon_temp_ads1115 hardware circuit diagram

Thermistor Circuit (per channel)

3.3 V (ESP32 pin) → 10 kΩ → ADS1115 AINx → NTC 10 kΩ (B=3950) → GND

Sensor construction — NTC thermistor thermal-epoxied to 4″×1.5″×¼″ cold steel bar

Hardware Photos

Sensors, ESP32-C3, and power supply — bench setup

Three NTC sensor assemblies on wooden mount, ESP32-C3 with antenna, and USB power supply

Static chamber — 8 ft deep concrete pit (looking down)

8-foot deep concrete block pit used as static thermal chamber

Inside the pit — sensor deployment at depth

Bottom of pit showing steel bar sensor assemblies and ESP32 electronics enclosure deployed in-situ

Pit entrance — July 2026 hardware update

Pit entrance showing updated hardware installation

Updated rig — variable regulator and buck converter installed

Updated sensor rig with variable regulator and ADS1115 dedicated buck converter

Running Calibration Confirmation

Auto-loads the most recent log CSV and evaluates inter-channel alignment. Run after any overnight session or after applying new offsets.

Pass criterion: std(adc0 − adc1) ≤ 5 counts AND std(adc0 − adc2) ≤ 5 counts (5 counts = 0.625 mV ≈ 10–15 m°C at pit temperatures)

Code 
import glob

LOGS_2_3V = os.path.join(NAS_BASE, "logs_2.3V/logs")   # VCC=2.3V era, no v_rail
LOGS_NEW   = os.path.join(NAS_BASE, "raw_logs")          # Jul 2026+, v_rail present
LOGS_CONV  = os.path.join(NAS_BASE, "converted_logs")    # pre-computed °C

LOGS_DIR = LOGS_2_3V  # calibration analysis uses historical data

ADC_LSB  = 0.00013108     # V/count — calibrated against multimeter (nominal 0.000125)
ADC_MV   = ADC_LSB * 1000 # mV/count
PASS_STD_THRESHOLD = 4.0  # mV

print(f"NAS base : {NAS_BASE}")

csv_files = [f for f in sorted(glob.glob(os.path.join(LOGS_DIR, "????-??-??.csv")))
             if os.path.basename(f) <= "2026-06-28.csv"]
if not csv_files:
    raise FileNotFoundError(f"No dated CSV files found in {LOGS_DIR}")

latest_csv = csv_files[-1]
df_cal = pd.read_csv(latest_csv, parse_dates=['timestamp'])
df_cal = df_cal.sort_values('timestamp').reset_index(drop=True)

cal_diff01 = (df_cal['adc0'] - df_cal['adc1']) * ADC_MV
cal_diff02 = (df_cal['adc0'] - df_cal['adc2']) * ADC_MV

print(f"File     : {latest_csv}")
print(f"Rows     : {len(df_cal)}")
print(f"Span     : {df_cal['timestamp'].iloc[0]}  →  {df_cal['timestamp'].iloc[-1]}")
print(f"v_rail   : {'present' if 'v_rail' in df_cal.columns else 'absent (2.3 V era)'}")
print()
print(f"adc0 − adc1 : mean={cal_diff01.mean():+.3f} mV  std={cal_diff01.std():.3f} mV")
print(f"adc0 − adc2 : mean={cal_diff02.mean():+.3f} mV  std={cal_diff02.std():.3f} mV")
NAS base : /mnt/nas/rooster/moon-temp-logs
File     : /mnt/nas/rooster/moon-temp-logs/logs_2.3V/logs/2026-06-28.csv
Rows     : 880
Span     : 2026-06-28 00:00:54  →  2026-06-28 14:39:54
v_rail   : absent (2.3 V era)

adc0 − adc1 : mean=-0.055 mV  std=1.005 mV
adc0 − adc2 : mean=+0.913 mV  std=1.146 mV
Code 
fig, axes = plt.subplots(2, 1, figsize=(14, 7), sharex=True)

for ax, diff, label, color in [
    (axes[0], cal_diff01, 'adc0 − adc1', 'darkorange'),
    (axes[1], cal_diff02, 'adc0 − adc2', 'crimson'),
]:
    ax.plot(df_cal['timestamp'], diff, color=color, linewidth=0.7, alpha=0.85)
    ax.axhline(0, color='gray', linestyle='--', linewidth=1.0, alpha=0.5)
    ax.axhline( PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
    ax.axhline(-PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%H:%M'))
    ax.set_title(f'{label}  (mean={diff.mean():+.3f} mV  std={diff.std():.3f} mV)')
    ax.set_ylabel('Δ mV')

axes[-1].set_xlabel('Time')
fig.suptitle(f'Inter-channel alignment — {os.path.basename(latest_csv)}', fontsize=12)
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Code 
# Calibration history — mean inter-channel difference per day (± 1 std, mV)
all_csvs = [f for f in sorted(glob.glob(os.path.join(LOGS_DIR, "????-??-??.csv")))
             if "2026-05-26.csv" <= os.path.basename(f) <= "2026-06-28.csv"]

rows = []
for path in all_csvs:
    d = pd.read_csv(path, parse_dates=['timestamp'])
    d = d.sort_values('timestamp').reset_index(drop=True)
    label = os.path.basename(path).replace('.csv', '')

    if '2026-05-27' in label:
        d = d[~((d['timestamp'] >= '2026-05-27 09:51:00') & (d['timestamp'] <= '2026-05-27 09:58:15'))]

    diff01 = (d['adc0'] - d['adc1']) * ADC_MV
    diff02 = (d['adc0'] - d['adc2']) * ADC_MV

    rows.append({
        'day':      label,
        'd01_mean': diff01.mean(),
        'd02_mean': diff02.mean(),
        'd01_std':  diff01.std(),
        'd02_std':  diff02.std(),
    })

hist = pd.DataFrame(rows).set_index('day')
print("Inter-channel alignment history (mV, heat test excluded from May 27)")
print()
print(hist.round(3).to_string())

fig, ax = plt.subplots(figsize=(14, 6))
x = range(len(rows))
for col, color, lbl in [
    ('d01_mean', 'darkorange', 'adc0 − adc1'),
    ('d02_mean', 'crimson',    'adc0 − adc2'),
]:
    means = hist[col].values
    stds  = hist[col.replace('mean', 'std')].values
    ax.errorbar(x, means, yerr=stds, fmt='o-', color=color, label=lbl,
                capsize=4, linewidth=1.5)

ax.axhline(0, color='gray', linestyle='--', linewidth=1, alpha=0.5)
ax.axhline( PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
ax.axhline(-PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)

# Offset adjustment markers
adj_markers = [
    ('2026-06-15', 'cyan',    'ADC2 −80→−64'),
    ('2026-06-18', 'yellow',  'ADC1 +16→+10'),
    ('2026-06-20', 'magenta', 'ADC1 +10→+22'),
]
idx_list = list(hist.index)
for day, color, lbl in adj_markers:
    if day in idx_list:
        xi = idx_list.index(day)
        ax.axvline(xi, color=color, linestyle='--', linewidth=1.2, alpha=0.85, label=lbl)
        ax.text(xi + 0.15, 0.97, lbl, transform=ax.get_xaxis_transform(),
                color=color, fontsize=8, va='top', ha='left')

ax.set_xticks(list(x))
ax.set_xticklabels(hist.index, fontsize=9, rotation=45, ha='right')
ax.set_ylabel('Δ mV')
ax.set_title('Inter-channel alignment per day — mean ± 1 std (mV)')
ax.legend()
plt.tight_layout()
plt.show()
Inter-channel alignment history (mV, heat test excluded from May 27)

            d01_mean  d02_mean  d01_std  d02_std
day                                             
2026-05-26     0.854    -4.275    9.612    5.422
2026-05-27    10.913     0.580    0.977    2.248
2026-05-28     3.376    -1.346    7.446    3.690
2026-05-29     0.153    -0.750    1.474    1.327
2026-05-30     1.450     0.082    0.969    0.880
2026-05-31     0.733    -0.086    1.000    1.139
2026-06-01     0.383    -0.396    0.811    1.056
2026-06-02     0.303    -0.297    1.894    3.592
2026-06-03     0.435     1.439    0.855    1.219
2026-06-04     0.852     1.691    1.033    1.402
2026-06-05     1.308     1.957    1.029    1.148
2026-06-06     0.744     1.748    1.007    1.066
2026-06-07     0.859     1.705    1.035    1.279
2026-06-08     0.785     1.675    1.045    1.206
2026-06-09     0.832     1.743    1.070    1.154
2026-06-10     0.929     1.722    1.051    1.151
2026-06-11     0.736     1.598    1.004    1.248
2026-06-12     0.998     1.762    1.093    1.223
2026-06-13     0.469     1.557    0.881    1.162
2026-06-14     0.847     1.766    1.073    1.074
2026-06-15     0.961     0.792    1.048    1.543
2026-06-16     1.200    -0.230    1.050    1.181
2026-06-17     0.486    -0.425    0.889    1.115
2026-06-18     1.194    -0.182    1.174    1.133
2026-06-19     1.569    -0.058    1.035    1.056
2026-06-20     1.118    -0.160    1.075    1.054
2026-06-21    -0.100     0.017    1.006    1.114
2026-06-22    -0.162     0.229    0.963    0.867
2026-06-23    -0.093     0.291    0.993    1.005
2026-06-24     0.269     0.504    1.061    1.015
2026-06-25     0.165     0.585    1.044    1.077
2026-06-26    -0.286     0.414    0.911    1.101
2026-06-27    -0.226     0.590    0.928    1.083
2026-06-28    -0.055     0.913    1.005    1.146

Code 
results = []
for label, diff in [('adc0 − adc1', cal_diff01), ('adc0 − adc2', cal_diff02)]:
    results.append((label, diff.mean(), diff.std(), diff.std() <= PASS_STD_THRESHOLD))

print("=" * 58)
print(f"  Calibration Confirmation — {os.path.basename(latest_csv)}")
print(f"  Pass threshold: std ≤ {PASS_STD_THRESHOLD} mV  (16 counts = 2 mV minimum step)")
print("=" * 58)
for label, mean_v, std_v, passed in results:
    print(f"  {label:14s}: mean={mean_v:+.3f} mV  std={std_v:.3f} mV  [{'PASS' if passed else 'FAIL'}]")
print("=" * 58)
if all(r[3] for r in results):
    print("  >> CALIBRATION OK")
else:
    print("  >> ADC offset may need adjustment — use burst alignment (Step 1)")
==========================================================
  Calibration Confirmation — 2026-06-28.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1   : mean=-0.055 mV  std=1.005 mV  [PASS]
  adc0 − adc2   : mean=+0.913 mV  std=1.146 mV  [PASS]
==========================================================
  >> CALIBRATION OK

Per-Day Calibration Summary

Pass/fail snapshot for every logged day (May 26 → Jun 28). Heat-test window excluded from May 27.

Code 
W = 58  # box width

all_csvs = [f for f in sorted(glob.glob(os.path.join(LOGS_DIR, "????-??-??.csv")))
            if "2026-05-26.csv" <= os.path.basename(f) <= "2026-06-28.csv"]

for path in all_csvs:
    fname = os.path.basename(path)
    d = pd.read_csv(path, parse_dates=['timestamp'])
    d = d.sort_values('timestamp').reset_index(drop=True)

    # Exclude heat test window from May 27
    if '2026-05-27' in fname:
        d = d[~((d['timestamp'] >= '2026-05-27 09:51:00') &
                (d['timestamp'] <= '2026-05-27 09:58:15'))]

    diff01 = (d['adc0'] - d['adc1']) * ADC_MV
    diff02 = (d['adc0'] - d['adc2']) * ADC_MV

    results = [
        ('adc0 − adc1', diff01.mean(), diff01.std(), diff01.std() <= PASS_STD_THRESHOLD),
        ('adc0 − adc2', diff02.mean(), diff02.std(), diff02.std() <= PASS_STD_THRESHOLD),
    ]
    all_pass = all(r[3] for r in results)

    print('=' * W)
    print(f"  Calibration Confirmation — {fname}")
    print(f"  Pass threshold: std ≤ {PASS_STD_THRESHOLD} mV  (16 counts = 2 mV minimum step)")
    print('=' * W)
    for label, mean_v, std_v, passed in results:
        print(f"  {label:12s}: mean={mean_v:+.3f} mV  std={std_v:.3f} mV  [{'PASS' if passed else 'FAIL'}]")
    print('=' * W)
    print(f"  >> {'CALIBRATION OK' if all_pass else 'CALIBRATION FAIL'}")
    print('=' * W)
    print()
==========================================================
  Calibration Confirmation — 2026-05-26.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.854 mV  std=9.612 mV  [FAIL]
  adc0 − adc2 : mean=-4.275 mV  std=5.422 mV  [FAIL]
==========================================================
  >> CALIBRATION FAIL
==========================================================

==========================================================
  Calibration Confirmation — 2026-05-27.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+10.913 mV  std=0.977 mV  [PASS]
  adc0 − adc2 : mean=+0.580 mV  std=2.248 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-05-28.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+3.376 mV  std=7.446 mV  [FAIL]
  adc0 − adc2 : mean=-1.346 mV  std=3.690 mV  [PASS]
==========================================================
  >> CALIBRATION FAIL
==========================================================

==========================================================
  Calibration Confirmation — 2026-05-29.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.153 mV  std=1.474 mV  [PASS]
  adc0 − adc2 : mean=-0.750 mV  std=1.327 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-05-30.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.450 mV  std=0.969 mV  [PASS]
  adc0 − adc2 : mean=+0.082 mV  std=0.880 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-05-31.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.733 mV  std=1.000 mV  [PASS]
  adc0 − adc2 : mean=-0.086 mV  std=1.139 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-01.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.383 mV  std=0.811 mV  [PASS]
  adc0 − adc2 : mean=-0.396 mV  std=1.056 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-02.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.303 mV  std=1.894 mV  [PASS]
  adc0 − adc2 : mean=-0.297 mV  std=3.592 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-03.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.435 mV  std=0.855 mV  [PASS]
  adc0 − adc2 : mean=+1.439 mV  std=1.219 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-04.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.852 mV  std=1.033 mV  [PASS]
  adc0 − adc2 : mean=+1.691 mV  std=1.402 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-05.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.308 mV  std=1.029 mV  [PASS]
  adc0 − adc2 : mean=+1.957 mV  std=1.148 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-06.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.744 mV  std=1.007 mV  [PASS]
  adc0 − adc2 : mean=+1.748 mV  std=1.066 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-07.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.859 mV  std=1.035 mV  [PASS]
  adc0 − adc2 : mean=+1.705 mV  std=1.279 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-08.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.785 mV  std=1.045 mV  [PASS]
  adc0 − adc2 : mean=+1.675 mV  std=1.206 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-09.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.832 mV  std=1.070 mV  [PASS]
  adc0 − adc2 : mean=+1.743 mV  std=1.154 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-10.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.929 mV  std=1.051 mV  [PASS]
  adc0 − adc2 : mean=+1.722 mV  std=1.151 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-11.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.736 mV  std=1.004 mV  [PASS]
  adc0 − adc2 : mean=+1.598 mV  std=1.248 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-12.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.998 mV  std=1.093 mV  [PASS]
  adc0 − adc2 : mean=+1.762 mV  std=1.223 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-13.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.469 mV  std=0.881 mV  [PASS]
  adc0 − adc2 : mean=+1.557 mV  std=1.162 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-14.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.847 mV  std=1.073 mV  [PASS]
  adc0 − adc2 : mean=+1.766 mV  std=1.074 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-15.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.961 mV  std=1.048 mV  [PASS]
  adc0 − adc2 : mean=+0.792 mV  std=1.543 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-16.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.200 mV  std=1.050 mV  [PASS]
  adc0 − adc2 : mean=-0.230 mV  std=1.181 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-17.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.486 mV  std=0.889 mV  [PASS]
  adc0 − adc2 : mean=-0.425 mV  std=1.115 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-18.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.194 mV  std=1.174 mV  [PASS]
  adc0 − adc2 : mean=-0.182 mV  std=1.133 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-19.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.569 mV  std=1.035 mV  [PASS]
  adc0 − adc2 : mean=-0.058 mV  std=1.056 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-20.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+1.118 mV  std=1.075 mV  [PASS]
  adc0 − adc2 : mean=-0.160 mV  std=1.054 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-21.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.100 mV  std=1.006 mV  [PASS]
  adc0 − adc2 : mean=+0.017 mV  std=1.114 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-22.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.162 mV  std=0.963 mV  [PASS]
  adc0 − adc2 : mean=+0.229 mV  std=0.867 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-23.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.093 mV  std=0.993 mV  [PASS]
  adc0 − adc2 : mean=+0.291 mV  std=1.005 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-24.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.269 mV  std=1.061 mV  [PASS]
  adc0 − adc2 : mean=+0.504 mV  std=1.015 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-25.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=+0.165 mV  std=1.044 mV  [PASS]
  adc0 − adc2 : mean=+0.585 mV  std=1.077 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-26.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.286 mV  std=0.911 mV  [PASS]
  adc0 − adc2 : mean=+0.414 mV  std=1.101 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-27.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.226 mV  std=0.928 mV  [PASS]
  adc0 − adc2 : mean=+0.590 mV  std=1.083 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

==========================================================
  Calibration Confirmation — 2026-06-28.csv
  Pass threshold: std ≤ 4.0 mV  (16 counts = 2 mV minimum step)
==========================================================
  adc0 − adc1 : mean=-0.055 mV  std=1.005 mV  [PASS]
  adc0 − adc2 : mean=+0.913 mV  std=1.146 mV  [PASS]
==========================================================
  >> CALIBRATION OK
==========================================================

Calibration Log

2026-06-15 11:40 MDT — ADC2 Adjustment

Analysis of Jun 12–14 data showed ADC2 reading consistently ~13 counts low (median gap adc0 − adc2 = +16 counts).

mosquitto_pub -h 10.0.10.20 -t moon-temp-001/offset/cmd -m '{"adc":2,"offset":-64}'

ADC2 offset: −80 → −64

2026-06-17 — ADC2 Confirmed

Three-day post-adjustment analysis (Jun 15–17) confirms ADC2 correction. ADC1 (offset +16) had a median gap of +6.8 counts (0.85 mV) — within spec but drifting.

2026-06-18 11:31 MDT — ADC1 Wrong-Direction Adjustment ⚠️

ADC1 was reading low (+6.8 counts below ADC0), but offset was reduced:

mosquitto_pub -h 10.0.10.20 -t moon-temp-001/offset/cmd -m '{"adc":1,"offset":10}'

ADC1 offset: +16 → +10 — this was the wrong direction. Gap widened to +11.3 counts (+1.41 mV) over Jun 18–19.

2026-06-20 22:00 MDT — ADC1 Corrected

mosquitto_pub -h 10.0.10.20 -t moon-temp-001/offset/cmd -m '{"adc":1,"offset":22}'

ADC1 offset: +10 → +22

2026-06-22 — ADC1 Confirmed

Jun 21–22 verification: ADC1 mean gap = −0.20 mV, std = 0.91 mV — PASS.

✓ Calibration Complete — 2026-06-28

All three channels are within specification. Sensors ready for deployment.

Channel Final Offset Status
ADC0 0 (reference) —
ADC1 +22 counts (+2.75 mV) PASS
ADC2 −64 counts (−8.00 mV) PASS

Calibration period: May 26 → June 28, 2026 (33 days) Pass criterion: std(adc0 − adcN) ≤ 4.0 mV Pit temperature drift: +2.1 °C over the calibration period (seasonal ground warming)

Inter-channel spread stabilised to < 0.05 °C by mid-June. The residual ±5–7 count diurnal swing on ADC1 is a thermal gradient between sensor placements — a DC offset cannot eliminate it, but it is centred around zero.

Next step: Reprogram ESP32-C3 MQTT broker address for field deployment.

Long-Term Stability & Seasonal Warming — May 26 → Jun 28

Loads all 33 days of valid data. Shows the pit temperature rising ~2.1 °C as summer heat diffuses into the ground, and confirms inter-channel offsets held stable throughout.

Code 
# Load all valid log days — both 2.3 V era and new v_rail era
R_SERIES = 10_000; R0 = 10_000; B = 3950; T0_K = 298.15
ADC_LSB  = 0.00013108  # calibrated (applies to all channels on this ADS1115)

def counts_to_celsius(counts, v_rail=None):
    """Convert ADS1115 raw count to °C.
    - v_rail present (Jul 2026+): use measured rail voltage
    - v_rail absent (pre-Jul 2026, 2.3 V era): use VCC=2.3
    """
    vcc = v_rail if (v_rail is not None and np.isscalar(v_rail) and v_rail > 0.5) else 2.3
    V   = counts * ADC_LSB
    if V <= 0 or V >= vcc:
        return np.nan
    R = V * R_SERIES / (vcc - V)
    return 1 / (1/T0_K + (1/B) * np.log(R / R0)) - 273.15

# Collect CSVs from both eras
old_csvs = [f for f in sorted(glob.glob(os.path.join(LOGS_2_3V, "????-??-??.csv")))
            if "2026-05-26.csv" <= os.path.basename(f) <= "2026-06-28.csv"]
new_csvs = [f for f in sorted(glob.glob(os.path.join(LOGS_NEW, "????-??-??.csv")))
            if os.path.basename(f) <= "2026-06-28.csv"]
all_csvs = old_csvs + new_csvs

daily = {}
for path in all_csvs:
    date = os.path.basename(path).replace('.csv','')
    d = pd.read_csv(path, parse_dates=['timestamp'])
    d = d.sort_values('timestamp').reset_index(drop=True)
    has_rail = 'v_rail' in d.columns
    for ch in ['adc0','adc1','adc2']:
        col = ch.replace('adc','temp')
        if has_rail:
            d[col] = d.apply(lambda r, c=ch: counts_to_celsius(r[c], r['v_rail']), axis=1)
        else:
            d[col] = d[ch].apply(counts_to_celsius)  # falls back to VCC=2.3
    if date == '2026-05-27':
        d = d[~((d['timestamp'] >= '2026-05-27 09:51:00') & (d['timestamp'] <= '2026-05-27 09:58:15'))]
    daily[date] = d

rows = []
for date, d in daily.items():
    rows.append({
        'date':    pd.Timestamp(date),
        't0':      d['temp0'].mean(),
        't1':      d['temp1'].mean(),
        't2':      d['temp2'].mean(),
        'd01':     (d['adc0'] - d['adc1']).mean() * ADC_MV,
        'd02':     (d['adc0'] - d['adc2']).mean() * ADC_MV,
        'd01_std': (d['adc0'] - d['adc1']).std()  * ADC_MV,
        'd02_std': (d['adc0'] - d['adc2']).std()  * ADC_MV,
        'era':     'v_rail' if 'v_rail' in d.columns else '2.3V',
    })
prog = pd.DataFrame(rows).set_index('date')
print(f"Loaded {len(prog)} days: {prog.index[0].date()} → {prog.index[-1].date()}")
print(f"  2.3V era : {(prog['era']=='2.3V').sum()} days")
print(f"  v_rail era: {(prog['era']=='v_rail').sum()} days")
print(prog[['t0','t1','t2','d01','d02','era']].round(3).to_string())
Loaded 34 days: 2026-05-26 → 2026-06-28
  2.3V era : 34 days
  v_rail era: 0 days
               t0     t1     t2     d01    d02   era
date                                                
2026-05-26  3.989  4.025  3.804   0.854 -4.275  2.3V
2026-05-27  4.136  4.604  4.162  10.913  0.580  2.3V
2026-05-28  4.139  4.283  4.081   3.376 -1.346  2.3V
2026-05-29  4.472  4.479  4.440   0.153 -0.750  2.3V
2026-05-30  4.686  4.748  4.690   1.450  0.082  2.3V
2026-05-31  4.822  4.853  4.818   0.733 -0.086  2.3V
2026-06-01  4.762  4.779  4.746   0.383 -0.396  2.3V
2026-06-02  4.808  4.821  4.796   0.303 -0.297  2.3V
2026-06-03  4.591  4.610  4.652   0.435  1.439  2.3V
2026-06-04  4.653  4.689  4.725   0.852  1.691  2.3V
2026-06-05  4.719  4.774  4.802   1.308  1.957  2.3V
2026-06-06  4.779  4.810  4.853   0.744  1.748  2.3V
2026-06-07  4.810  4.846  4.882   0.859  1.705  2.3V
2026-06-08  4.968  5.002  5.039   0.785  1.675  2.3V
2026-06-09  5.096  5.131  5.170   0.832  1.743  2.3V
2026-06-10  5.101  5.141  5.174   0.929  1.722  2.3V
2026-06-11  5.304  5.335  5.372   0.736  1.598  2.3V
2026-06-12  5.429  5.471  5.504   0.998  1.762  2.3V
2026-06-13  5.486  5.506  5.552   0.469  1.557  2.3V
2026-06-14  5.609  5.644  5.683   0.847  1.766  2.3V
2026-06-15  5.701  5.742  5.735   0.961  0.792  2.3V
2026-06-16  5.829  5.879  5.819   1.200 -0.230  2.3V
2026-06-17  6.001  6.022  5.984   0.486 -0.425  2.3V
2026-06-18  6.244  6.294  6.237   1.194 -0.182  2.3V
2026-06-19  6.572  6.637  6.569   1.569 -0.058  2.3V
2026-06-20  6.714  6.761  6.708   1.118 -0.160  2.3V
2026-06-21  6.746  6.742  6.747  -0.100  0.017  2.3V
2026-06-22  6.791  6.784  6.800  -0.162  0.229  2.3V
2026-06-23  6.801  6.798  6.813  -0.093  0.291  2.3V
2026-06-24  6.912  6.923  6.932   0.269  0.504  2.3V
2026-06-25  7.005  7.012  7.029   0.165  0.585  2.3V
2026-06-26  7.150  7.139  7.167  -0.286  0.414  2.3V
2026-06-27  7.236  7.227  7.260  -0.226  0.590  2.3V
2026-06-28  7.277  7.275  7.315  -0.055  0.913  2.3V
Code 
# Pit temperature trend — daily mean, all 3 channels
fig, ax = plt.subplots(figsize=(16, 5))
for col, color, label in [('t0','steelblue','ch0'), ('t1','darkorange','ch1'), ('t2','crimson','ch2')]:
    ax.plot(prog.index, prog[col], 'o-', color=color, label=label, linewidth=1.5, markersize=4)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d'))
ax.set_ylabel('°C')
ax.set_title('Daily mean pit temperature — May 26 → Jun 28  [MF58 10K NTC, B=3950 K]')
ax.legend()
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

rise = prog['t0'].iloc[-1] - prog['t0'].iloc[0]
days = (prog.index[-1] - prog.index[0]).days
print(f'Total rise: {rise:+.3f} °C over {days} days  ({rise/days*7:.3f} °C/week)')

Total rise: +3.289 °C over 33 days  (0.698 °C/week)
Code 
# Inter-channel offset drift — daily mean ± 1 std (mV)
fig, axes = plt.subplots(2, 1, figsize=(16, 6), sharex=True)
for ax, col, std_col, label, color in [
    (axes[0], 'd01', 'd01_std', 'adc0 − adc1 (mV)', 'darkorange'),
    (axes[1], 'd02', 'd02_std', 'adc0 − adc2 (mV)', 'crimson'),
]:
    ax.fill_between(prog.index, prog[col]-prog[std_col], prog[col]+prog[std_col], color=color, alpha=0.2)
    ax.plot(prog.index, prog[col], 'o-', color=color, linewidth=1.2, markersize=4)
    ax.axhline(0, color='black', linestyle='--', linewidth=0.8)
    ax.axhline( PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.7, label='±pass threshold')
    ax.axhline(-PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.7)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d'))
    ax.set_ylabel('mV')
    ax.set_title(f'{label}  (shaded = ±1 std)')
    ax.legend(loc='upper right')
fig.suptitle('Inter-channel offset drift — full calibration period', fontsize=12)
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Updated Equipment Calibration — July 2026

Hardware updated July 2026: variable regulator + dedicated buck converter for ADS1115 VDD, rail voltage monitor on AIN3. Sensors returned to the static thermal chamber for post-update calibration verification.

Current device state (flash config):

{"offset_adc0":0,"offset_adc1":49,"offset_adc2":-73,"avg_samples":64}

Averaging time: ADS1115 at 128 SPS, 9 ms conversion wait per channel, 4 channels × 64 samples: > 4 channels × 64 samples × 9 ms = 2,304 ms ≈ 2.3 seconds per published reading

Hardware change notes: - Single ADC sample per minute (pre-averaging firmware) produced noise spikes of ±1–2 °C on individual reads — not actual temperature events; observed during field experimental runs - 64-sample averaging per publish implemented in firmware to suppress these spikes - Rail voltage (AIN3) averaged identically; v_rail tracks ADS1115 VDD

Code 
# Load all converted_logs CSVs (post-update calibration era)
field_csvs = sorted(glob.glob(os.path.join(LOGS_CONV, "????-??-??.csv")))
if not field_csvs:
    raise FileNotFoundError(f"No CSVs found in {LOGS_CONV}")

dfs = []
for path in field_csvs:
    d = pd.read_csv(path, parse_dates=['timestamp'])
    d = d.sort_values('timestamp').reset_index(drop=True)
    dfs.append(d)
field = pd.concat(dfs, ignore_index=True)

field['d01_mV'] = (field['adc0'] - field['adc1']) * ADC_MV
field['d02_mV'] = (field['adc0'] - field['adc2']) * ADC_MV
field['dt01']   = field['t0_c'] - field['t1_c']
field['dt02']   = field['t0_c'] - field['t2_c']
field['date']   = field['timestamp'].dt.date

print(f"Rows : {len(field)}")
print(f"Span : {field['timestamp'].iloc[0]}  →  {field['timestamp'].iloc[-1]}")
print(f"v_rail: mean={field['v_rail'].mean():.4f} V  min={field['v_rail'].min():.4f} V  max={field['v_rail'].max():.4f} V")
Rows : 3047
Span : 2026-07-11 10:00:33  →  2026-07-13 12:38:02
v_rail: mean=3.5170 V  min=3.5110 V  max=3.5345 V

ADC Channel Alignment

Code 
# Time series — inter-channel difference in mV
fig, axes = plt.subplots(2, 1, figsize=(16, 7), sharex=True)
for ax, col, label, color in [
    (axes[0], 'd01_mV', 'adc0 − adc1', 'darkorange'),
    (axes[1], 'd02_mV', 'adc0 − adc2', 'crimson'),
]:
    ax.plot(field['timestamp'], field[col], color=color, linewidth=0.6, alpha=0.8)
    ax.axhline(0, color='gray', linestyle='--', linewidth=1, alpha=0.5)
    ax.axhline( PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
    ax.axhline(-PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d %H:%M'))
    ax.set_title(f'{label}  mean={field[col].mean():+.3f} mV  std={field[col].std():.3f} mV')
    ax.set_ylabel('Δ mV')
fig.suptitle('Field deployment — ADC inter-channel alignment (mV)', fontsize=12)
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Code 
# Daily mean ADC alignment
daily_adc = field.groupby('date').agg(
    d01_mean=('d01_mV', 'mean'), d01_std=('d01_mV', 'std'),
    d02_mean=('d02_mV', 'mean'), d02_std=('d02_mV', 'std'),
).reset_index()

print("Daily ADC alignment (mV):")
print(daily_adc.round(3).to_string(index=False))

fig, ax = plt.subplots(figsize=(12, 5))
x = range(len(daily_adc))
for col_m, col_s, color, lbl in [
    ('d01_mean', 'd01_std', 'darkorange', 'adc0 − adc1'),
    ('d02_mean', 'd02_std', 'crimson',    'adc0 − adc2'),
]:
    ax.errorbar(x, daily_adc[col_m], yerr=daily_adc[col_s], fmt='o-', color=color,
                label=lbl, capsize=4, linewidth=1.5)
ax.axhline(0, color='gray', linestyle='--', linewidth=1, alpha=0.5)
ax.axhline( PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6, label='±pass threshold')
ax.axhline(-PASS_STD_THRESHOLD, color='lime', linestyle=':', linewidth=0.8, alpha=0.6)
ax.set_xticks(list(x))
ax.set_xticklabels([str(d) for d in daily_adc['date']], rotation=30, ha='right')
ax.set_ylabel('Δ mV')
ax.set_title('Daily mean ADC inter-channel alignment — field deployment')
ax.legend()
plt.tight_layout()
plt.show()
Daily ADC alignment (mV):
      date  d01_mean  d01_std  d02_mean  d02_std
2026-07-11     3.924    1.922    -0.876    1.153
2026-07-12     3.401    0.252    -1.308    0.151
2026-07-13     2.810    1.208    -1.130    0.401

Temperature (°C)

Code 
# Full time series in °C
fig, ax = plt.subplots(figsize=(16, 5))
for col, color, label in [
    ('t0_c', 'steelblue',  'ch0'),
    ('t1_c', 'darkorange', 'ch1'),
    ('t2_c', 'crimson',    'ch2'),
]:
    ax.plot(field['timestamp'], field[col], color=color, linewidth=0.8, alpha=0.85, label=label)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d %H:%M'))
ax.set_ylabel('°C')
ax.set_title('Field deployment — sensor temperatures (°C)')
ax.legend()
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Code 
# Daily mean temperature in °C
daily_temp = field.groupby('date').agg(
    t0_mean=('t0_c', 'mean'), t0_std=('t0_c', 'std'),
    t1_mean=('t1_c', 'mean'), t1_std=('t1_c', 'std'),
    t2_mean=('t2_c', 'mean'), t2_std=('t2_c', 'std'),
).reset_index()

print("Daily mean temperature (°C):")
print(daily_temp.round(3).to_string(index=False))

fig, ax = plt.subplots(figsize=(12, 5))
x = range(len(daily_temp))
for col_m, col_s, color, lbl in [
    ('t0_mean', 't0_std', 'steelblue',  'ch0'),
    ('t1_mean', 't1_std', 'darkorange', 'ch1'),
    ('t2_mean', 't2_std', 'crimson',    'ch2'),
]:
    ax.errorbar(x, daily_temp[col_m], yerr=daily_temp[col_s], fmt='o-', color=color,
                label=lbl, capsize=4, linewidth=1.5)
ax.set_xticks(list(x))
ax.set_xticklabels([str(d) for d in daily_temp['date']], rotation=30, ha='right')
ax.set_ylabel('°C')
ax.set_title('Daily mean field temperature — July 2026')
ax.legend()
plt.tight_layout()
plt.show()
Daily mean temperature (°C):
      date  t0_mean  t0_std  t1_mean  t1_std  t2_mean  t2_std
2026-07-11   12.928   1.338   13.027   1.375   12.905   1.364
2026-07-12   12.319   0.020   12.406   0.023   12.286   0.021
2026-07-13   12.302   0.011   12.374   0.034   12.272   0.013

Code 
# Inter-channel spread in °C
fig, axes = plt.subplots(2, 1, figsize=(16, 7), sharex=True)
for ax, col, label, color in [
    (axes[0], 'dt01', 'ch0 − ch1 (°C)', 'darkorange'),
    (axes[1], 'dt02', 'ch0 − ch2 (°C)', 'crimson'),
]:
    ax.plot(field['timestamp'], field[col], color=color, linewidth=0.6, alpha=0.8)
    ax.axhline(0, color='gray', linestyle='--', linewidth=1, alpha=0.5)
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d %H:%M'))
    ax.set_title(f'{label}  mean={field[col].mean():+.4f} °C  std={field[col].std():.4f} °C')
    ax.set_ylabel('Δ °C')
fig.suptitle('Field deployment — inter-channel spread (°C)', fontsize=12)
fig.autofmt_xdate()
plt.tight_layout()
plt.show()

Stability Check — Spike Detection

Sensors sit in a high-thermal-mass concrete pit: temperature should be extremely stable (< 0.05 °C variation per sample under normal conditions). Any sample deviating > 1.5 °C from its 5-sample rolling median is flagged as a suspect read.

Code 
# Flag samples that deviate more than 1.5 °C from a 5-point rolling median
SPIKE_THRESH = 0.25  # °C

field_s = field.sort_values('timestamp').reset_index(drop=True)
spike_rows = pd.DataFrame()
for ch in ['t0_c', 't1_c', 't2_c']:
    roll = field_s[ch].rolling(5, center=True, min_periods=1).median()
    mask = (field_s[ch] - roll).abs() > SPIKE_THRESH
    if mask.any():
        bad = field_s[mask][['timestamp', 'adc0', 'adc1', 'adc2', 'v_rail', ch]].copy()
        bad['channel'] = ch
        bad['deviation_C'] = (field_s[ch] - roll)[mask]
        spike_rows = pd.concat([spike_rows, bad])

if spike_rows.empty:
    print(f"No spikes detected (threshold: >{SPIKE_THRESH} °C from 5-sample rolling median)")
    print("Signal is stable — consistent with high-thermal-mass static chamber environment.")
else:
    print(f"{len(spike_rows)} suspect samples found:")
    print(spike_rows[['timestamp','channel','deviation_C']].to_string(index=False))

# Plot all channels with spike threshold bands
fig, ax = plt.subplots(figsize=(16, 5))
for col, color, label in [('t0_c','steelblue','ch0'), ('t1_c','darkorange','ch1'), ('t2_c','crimson','ch2')]:
    ax.plot(field_s['timestamp'], field_s[col], color=color, linewidth=0.7, alpha=0.85, label=label)
ax.xaxis.set_major_formatter(mdates.DateFormatter('%b %d %H:%M'))
ax.set_ylabel('°C')
ax.set_title(f'Stability check — all channels (spike threshold ±{SPIKE_THRESH} °C, none detected)')
ax.legend()
fig.autofmt_xdate()
plt.tight_layout()
plt.show()
No spikes detected (threshold: >0.25 °C from 5-sample rolling median)
Signal is stable — consistent with high-thermal-mass static chamber environment.

Offset Recommendation

ADC0 is the reference channel (offset always 0). Recommended adjustments derived from mean inter-channel gap over Jul 11–13 (2,928 samples).

Channel Current offset Mean gap (counts) Mean gap (mV) Recommended offset
ADC0 0 (ref) — — 0
ADC1 +22 +26.9 (reads low) +3.52 mV +49
ADC2 −64 −9.0 (reads high) −1.18 mV −73

Send to device (JSON payload on moon-temp-001/offset/cmd):

{"adc":1,"offset":49}
{"adc":2,"offset":-73}
Code 
# Offset recommendation — visualise the gap vs current offset
ADC_LSB = 0.00013108
ADC_MV  = ADC_LSB * 1000

offsets = {
    'ADC0': {'current': 0,   'gap_counts': 0.0},
    'ADC1': {'current': 22,  'gap_counts': field['d01_mV'].mean() / ADC_MV},
    'ADC2': {'current': -64, 'gap_counts': -field['d02_mV'].mean() / ADC_MV},
}
print("Offset summary:")
print(f"  {'Channel':6s}  {'Current':>8s}  {'Gap (counts)':>13s}  {'Gap (mV)':>10s}  {'Recommended':>12s}")
for ch, v in offsets.items():
    new = round(v['current'] + v['gap_counts'])
    gap_mv = v['gap_counts'] * ADC_MV
    print(f"  {ch:6s}  {v['current']:>+8d}  {v['gap_counts']:>+13.2f}  {gap_mv:>+10.3f}  {new:>+12d}")
Offset summary:
  Channel   Current   Gap (counts)    Gap (mV)   Recommended
  ADC0          +0          +0.00      +0.000            +0
  ADC1         +22         +25.93      +3.399           +48
  ADC2         -64          +8.72      +1.143           -55