The four pins
Four pins, and only one of them is a decision. What is soldered around them — two pull-ups, a decoupling capacitor and a power light — is the work somebody has already done for you, and knowing which is which is what makes the rest of the board readable.
The four names
GND is the return path for everything, and it goes first. It is on the square pad.
SDA and SCL are the bus: data and clock. Both are shared with every other I²C device on the project, and both are pulled up on this board.
VCC is the one that is not free choice, and the reason is on the board rather than in the sensor. The sensor itself is happy anywhere from 2.15 V to 5.5 V. But the two pull-up resistors tie SDA and SCL to VCC, so whatever you feed VCC is the voltage your microcontroller's pins will see idling on the bus.
On an Uno that is 5 V and correct. On an ESP32, an ESP32-S3 or a Pico it has to be 3V3, because 5 V on a 3.3 V pin is a failure that takes weeks and then happens all at once.
The back of the board prints 3–5V next to that pin, which is the supply the board is designed for rather than the sensor's full range.
What else is soldered to it
Six components, and between them they are every decision you did not have to make.
- R1 and R2, 10 kΩ each, hold SDA and SCL high so the bus has a resting state. Every I²C device only ever pulls a line down; something has to pull it back up, and on most breakout boards that something is a resistor you had to buy. Here it is fitted. It is also switchable, which is the second jumper's job.
- R3, also 10 kΩ, holds the sensor's ADDR pin at ground. The datasheet is explicit that ADDR must not be left floating, and this resistor is why it never is — and why the board's address is 0x44 and not a lottery.
- C1, 100 nF, sits across the supply next to the sensor. Section 3.1 of the datasheet asks for exactly this, placed exactly there.
- LED1 and R4, 5.1 kΩ, are the power light. It means VCC arrived. It does not blink on a reading, it cannot tell you the bus is working, and it draws more current than the sensor does — which is its own article.
The part in the middle
The sensor is the small grey rectangle between the two mounting holes: a Sensirion SHT31-DIS in a DFN-8 package, 2.5 mm square, which is smaller than one of the header pads.
Inside it are a capacitive humidity element, a temperature element, a converter for both and a memory holding the calibration burned in at the factory — which the part reloads before every single measurement. That is why there is nothing on this board to trim, no reference voltage to supply and no table to look anything up in.
Eight numbered pins and a pad underneath. Six of the eight are wired: SDA, SCL, VDD, VSS, ADDR, and pin 7, which the datasheet lists as having no electrical function at all and says to connect to VSS — so the board does. The pad underneath goes to ground too, which the datasheet asks for on mechanical grounds rather than electrical ones.
The two that go nowhere are ALERT and nRESET, and in both cases that is what the datasheet recommends: leave ALERT floating if you are not using it, leave nRESET floating or tie it to VDD through a resistor. Floating is not neglect here. It is the instruction.
When it does not work
Turn the board so the sensor faces you and the header is at the bottom, and read GND, VCC, SDA, SCL from left to right. GND is on the square pad, and the square is the only marking that tells four identical holes apart — the back of the board prints the same four names mirrored, which is what catches people out.
No. The board's two pull-up resistors tie SDA and SCL to whatever you put on VCC, so a 5 V supply idles both signal lines at 5 V against pins rated for 3.3 V. It appears to work, which is the problem — the damage is cumulative. The sensor is specified from 2.15 V, so 3V3 costs you nothing at all.
The light only proves VCC and GND arrived; it is wired straight across the supply through a 5.1 kΩ resistor and knows nothing about the bus. A lit board with a silent bus is almost always SDA and SCL swapped, or a missing ground on the host side. Run an I²C scanner before a sensor sketch.
Enough. C1 is the 100 nF the datasheet asks for, placed next to the sensor, and the sensor draws so little that supply noise is rarely its problem. If a long supply wire is dropping voltage under load, that is a wire problem and a bigger capacitor will not fix it.
There is an arch of missing board round the sensor, and it is not a manufacturing artefact.
The slot round the sensor →Edit this page — content/books/sht31/the-four-pins.mdx
Questions about this product
See what other owners have asked, and read their solutions.
SHT31 Temperature and Humidity Sensor
Loading discussions…
Discuss this article
Ask about this page. The answer stays here, on the page it belongs to, for whoever hits the same wall next.