A pressure sensor can be useful in a water-level logger, but the board you tested on your desk is only one part of that device. You still need to decide which surface touches the water, how the electronics stay dry and whether the pressure range covers the depth you want to measure.
On our bench we connected a GraftSense MS5803-01BA module to a GraftPort-RP2040 controller, ran a MicroPython driver and read temperature and pressure. That test happened in air. It is a useful first step toward a wet-environment prototype, and it is not an underwater test.
Builders run into this question regularly. In a June 2026 SparkFun discussion, someone measuring pressure in a water column asked whether the whole breakout board could go underwater or only the sensor's circular face. That thread concerns the MS5803-14BA, a different range variant, but the distinction between the sensing surface and the surrounding electronics applies here too.
Check the suffix before planning the depth
MS5803 names a family, not one pressure range. Our module is the MS5803-01BA. TE's datasheet for it specifies a full-accuracy range of 300 to 1100 mbar and an extended range of 10 to 1300 mbar. Its much higher overpressure limit is about survival under stated conditions, not accurate measurement up to that limit.
This sensor measures absolute pressure, which includes the air pressure at the water's surface. You cannot spend its whole range on the water above it. With a freshwater illustration and 1000 mbar at the surface, each meter of depth adds roughly 98 mbar. One meter puts you near 1098 mbar, already close to the top of the full-accuracy range. That is an explanatory calculation, not a depth rating for the module.
A MicroPython builder considering this exact variant for a 10 m lake logger raised the selection problem in 2022, and the thread does not report a successful final test. Ten meters adds roughly another bar to surface pressure, beyond the 01BA's stated measurement range. A different variant or sensor would need its own range, media and software checks.
Write down the intended minimum and maximum pressure before designing the enclosure, including surface pressure and any conditions beyond a quiet water column. "It survives immersion" and "it measures the required pressure accurately" are separate requirements.
Reproduce the dry connection first
We used a four-wire HY2.0 cable from the module to the controller's I2C0 port, with GP4 as SDA and GP5 as SCL. The carrier has a regulator producing 3.3 V from its 5 V connector supply, and the I2C lines are pulled up to 3.3 V. Match the carrier's actual connections rather than applying the connector voltage to the sensor chip directly.
The program created the bus at 400 kHz and scanned for 0x76 or 0x77; our module answered at 0x76. On this chip, the PS pin selects the communication protocol and, in I2C mode, CSB selects the address. Check those when a module answers at a different address from the one a library expects.
We then copied the MS5803-01BA driver package to the controller. The package files we used were labeled 1.0.1, while the install command in the accompanying notes named 1.0.0. Keep the actual driver files with your test record, so a later change in output can be traced to a specific version.
Follow the reading all the way to its unit
The test program initialized the sensor, sent a reset and requested three successive readings: one without unit conversion, one in Celsius and pascals, and one in Fahrenheit and bar, at the default oversampling setting of 256. That sequence is useful because the same physical pressure should stay consistent when only the displayed unit changes.
It did not. One reading showed about 10,032 Pa and the next about 1.0032 bar. Those differ by a factor of ten: 1.0032 bar is about 100,320 Pa. Reading the driver, we found an extra division by ten in its pascal conversion. Its compensation path also needs checking against the 01BA datasheet before its output can serve as a reference measurement. We are reporting the readings as recorded, not substituting corrected values we did not measure.
For a new prototype, verify the driver first and compare one compensated reading expressed in both units. Capture pressure, temperature, timestamp and the actual settings. Getting numbers on the terminal is a real milestone, but a depth calculation built on the wrong scale will still be wrong.
The chip's factory coefficients let software compensate its raw measurements. They do not remove the need to use the right calculation for the fitted variant. Once that path is checked, take repeated readings in air before changing the mechanical setup. That gives you a baseline to compare the assembled device against.
Draw the wet boundary around the sensor, not the whole board
The metal cap and protective gel belong to the sensing component. The carrier's connector, solder joints, regulator and controller are separate items. Our bench setup had no seal around any of them, and we did not run a water test.
TE's datasheet describes sealing at the outside diameter of the metal cap with an O-ring, keeping the electronics protected from water and humidity while avoiding mechanical stress on the sensor. Treat that as an integration approach to develop and verify for your component and enclosure, not as a complete seal design for this carrier.
Before making the housing, mark the intended wet surface, seal contact, cable exit and dry electronics space on a drawing. Check how the sensor is supported and how assembly tolerances affect the seal. A coating or adhesive that worked on somebody else's board does not establish performance for your geometry, materials or exposure time.
The older datasheet we worked from also ties its accuracy specifications to a wet/dry condition, stating that the sensor should typically be dried once a day. Continuous immersion therefore needs explicit confirmation for the exact variant you order and its current documentation. It cannot be inferred from general statements about water-resistant watches.
Turn the bench result into a limited, checkable trial
Once the pressure scale and mechanical design are ready, run a controlled trial that stays within the confirmed pressure and exposure limits:
- Record surface pressure, liquid type, temperature, known water height and the corresponding sensor reading.
- Repeat measurements after returning to the starting condition.
- Inspect the dry side for moisture after each exposure.
For an open water container, the quantity tied to water height is the pressure above the current surface-pressure baseline. A sealed container with changing gas pressure needs a different treatment. Define that environment before choosing a reference or promising users a water-level number.
The useful outcome is a small set of comparable measurements and an intact, inspected assembly. From there you can decide whether the sensor range, seal and sampling behavior suit the product. A polished enclosure or a screen showing centimeters comes after those decisions.