Acoustic Pressure Sensor Calibration Dashboard
Field test 2026-06-23 & 2026-06-24 |
7 pressure tests, 14 sensor IDs, NUI reference at 1 Hz (local time)
Fourteen acoustic underwater pressure transmitters were lowered to increasing depths
in seven incremental tests (8.6 bar → 201.6 bar target).
This dashboard compares each sensor's reported pressure (sensor_bar) against
the NUI reference instrument (nui_bar, ground truth).
Calibration statistics are computed only in the valid range where sensors are not
saturated (0–29 bar, raw 100–3000). Saturated points are shown de-emphasised
in Section 2 so you can see when sensors hit the ceiling.
All timestamps are Europe/Oslo local time.
1 — Full test run overview
NUI pressure across both test days. Shaded bands are the auto-detected test windows;
the dashed red line is the 29 bar sensor saturation ceiling (~290 m water depth).
Use the Plotly toolbar to zoom and pan.
Above 29 bar every sensor maxes out (flat line at the top), so the meaningful
accuracy comparison happens during the pressure ramps and in tests 1–2.
2 — Per-test detail
Select a test from the dropdown. The blue line is the NUI ground truth; coloured dots
are individual sensor transmissions. Hollow/faded markers indicate saturated or below-min
readings. Click an ID in the legend to hide/show it. Hover any point for full detail.
3 — How accurately do the sensors read the true pressure?
Each dot is one sensor reading. The horizontal position is the real pressure measured
by NUI’s reference gauge; the vertical position is what our sensor reported for
the same moment. The dashed line is where a dot would sit if the sensor were perfectly
correct. The closer the dots hug that line, the more accurate the sensor. This only
covers the 0–29 bar range (down to about 290 m), which is the deepest
our sensors can physically measure before they max out.
4 — The accuracy in numbers
This table puts numbers on the chart above, per sensor and per product type.
“Bias” tells you whether a sensor tends to read slightly high (positive)
or slightly low (negative) on average. “MAE” (mean absolute error) is
the typical size of the miss, in bar — lower is better. As a rule of thumb,
0.1 bar is about 1 metre of water depth. “Std” shows how
consistent the sensor is. The takeaway: in the validated 0–29 bar range,
the sensors read the true pressure within roughly 0.1 bar (about 1 metre),
which gives us empirical backing for the depth ratings.
| ID | Type | N | Bias (bar) | Std (bar) | MAE (bar) | MaxAbs (bar) |
| Per sensor ID (sorted by MAE, valid range 0–29 bar) |
| 119 | 9mm Press epoxy sleeve | 41 | +0.034 | 0.071 | 0.054 | 0.330 |
| 115 | 9mm Press | 44 | +0.026 | 0.152 | 0.075 | 0.635 |
| 102 | 6mm Press | 55 | +0.062 | 0.088 | 0.087 | 0.405 |
| 107 | 7mm Press | 58 | +0.066 | 0.083 | 0.090 | 0.370 |
| 123 | 13mm Press | 39 | +0.084 | 0.159 | 0.091 | 0.789 |
| 114 | 9mm Press | 52 | +0.053 | 0.122 | 0.093 | 0.569 |
| 128 | 13mm Press v2 lithium | 51 | +0.067 | 0.109 | 0.093 | 0.542 |
| 129 | 13mm Press v2 lithium | 29 | +0.082 | 0.105 | 0.094 | 0.405 |
| 103 | 6mm Press | 23 | +0.081 | 0.056 | 0.095 | 0.140 |
| 118 | 9mm Press epoxy sleeve | 40 | +0.049 | 0.192 | 0.102 | 0.770 |
| 106 | 7mm Press | 55 | +0.097 | 0.103 | 0.105 | 0.712 |
| 122 | 13mm Press | 55 | +0.063 | 0.153 | 0.111 | 0.590 |
| 134 | 16mm Press | 30 | +0.089 | 0.223 | 0.121 | 1.186 |
| 135 | 16mm Press | 27 | -0.214 | 1.346 | 0.345 | 6.905 |
| Per sensor type |
| 9mm Press epoxy sleeve | 81 | +0.041 | 0.143 | 0.077 | — |
| 9mm Press | 96 | +0.041 | 0.136 | 0.085 | — |
| 6mm Press | 78 | +0.067 | 0.080 | 0.089 | — |
| 13mm Press v2 lithium | 80 | +0.073 | 0.107 | 0.093 | — |
| 7mm Press | 113 | +0.081 | 0.094 | 0.097 | — |
| 13mm Press | 94 | +0.072 | 0.155 | 0.103 | — |
| 16mm Press | 57 | -0.055 | 0.944 | 0.227 | — |
Takeaway:
In the validated 0–29 bar range (raw 100–3000, sensor not saturated),
13 of 14 sensors track true pressure within a mean absolute error of
< 0.13 bar (< 1.3 m water).
The best-performing sensor achieves MAE = 0.054 bar.
ID 135 (16mm Press) is the outlier with MAE = 0.345 bar
and high variance (std = 1.346 bar) — it warrants individual inspection.
Above 29 bar the sensor output saturates at raw = 3000 (290 m depth);
readings in that range are flagged and excluded from calibration statistics.