HMC5883L

No — it is almost certainly a QMC5883L from QST, which has replaced genuine Honeywell silicon on most GY-271 and GY-270 breakouts since the HMC5883L was discontinued. The two parts are not register compatible: the QMC5883L puts X/Y/Z data at 0x00 to 0x05 with control registers at 0x09 to 0x0C, while the HMC5883L uses configuration A at 0x00, configuration B (gain) at 0x01, mode at 0x02 and data at 0x03 to 0x08 in X, Z, Y order. Use a QMC5883L driver, or buy verified original Honeywell stock. An I2C scan is the definitive check: 0x1E means HMC5883L, 0x0D means QMC5883L.

Honeywell ended production, so for new sockets we recommend the TDK/MMC5983MA (±8 gauss, low noise, tiny WLP or QFN package), the ST LIS3MDL (±4 to ±16 gauss selectable, I2C and SPI, 1.9 V to 3.6 V), the PNI RM3100 for high-precision industrial heading, or the ST IIS2MDC. If you are supporting an installed base, keep buying verified original HMC5883L and plan a documented migration; do not design a new product around broker-market HMC5883L without a second source.

The 7-bit address is 0x1E (0x3C for write, 0x3D for read) and it is fixed — there is no address pin, so you cannot put two HMC5883L devices on one bus. Registers: 0x00 configuration A (averaging MA, output rate DO, measurement mode MS), 0x01 configuration B (gain GN), 0x02 mode (continuous 0x00 or single 0x01), 0x03 to 0x08 data output in the order X MSB, X LSB, Z MSB, Z LSB, Y MSB, Y LSB, 0x09 status (data ready, data output register lock), and 0x0A to 0x0C identification A/B/C. Default gain is ±1.3 gauss at 1090 LSB/gauss; use ±8.1 gauss (230 LSB/gauss) near motors or speakers where fields are strong.

Two separate effects. Hard-iron offsets come from permanent magnets and magnetised steel near the sensor — speakers, motors, screws, steel enclosures — and are removed by a fixed offset per axis found by rotating the board through a figure-of-eight and taking the midpoint of the min and max readings. Soft-iron distortion comes from ferromagnetic material that bends the field and needs a 3×3 correction matrix. Neither can be trimmed out in silicon. Also keep the sensor away from current-carrying traces: a straight conductor carrying 1 A produces about 0.2 gauss at 1 cm, which is a large fraction of the Earth's total field (roughly 0.25 gauss to 0.65 gauss).

Yes. The HMC5883L uses set/reset strap pulses to degauss the AMR elements and cancel offset, and those pulses are roughly 1 A peak. C2 (0.22 µF nominal) is the set/reset capacitor and C1 (4.7 µF nominal) is the reservoir that supplies the pulse; both must be low-ESR ceramic (under 200 mΩ recommended) and the traces to them must be wide enough to carry the pulse with low drop. Omit them and you get degraded offset stability, increased noise or a device that fails self-test. Do not hand solder this package — Honeywell specifies reflow only, MSL 3 with a 260 °C peak.

Run four checks. Read identification registers 0x0A, 0x0B and 0x0C — genuine parts return 0x48, 0x35, 0x38 (ASCII H, 5, 8). Trigger the built-in self-test with positive bias at GN = 5 (configuration B = 0xA0) and confirm the response falls between 243 and 575 LSB on all three axes. Confirm the I2C address is 0x1E and that the field vector magnitude sits near 0.25 gauss to 0.65 gauss when level and away from ferrous material. Physically, check laser marking, LCC pad finish and the 18 mg package mass. Clones and QMC5883L substitutions fail one or more of these.

Yes, but only from last-time-buy inventory and audited broker stock — there is no factory channel any more. We hold verified original tape-and-reel stock and screen every lot with the identification-register and self-test checks described above. MOQ is 500 pcs on tape for production or 50 pcs loose for prototyping and repair, with 1-3 business day dispatch while stock lasts. Quantities are finite; if this is a long-life product, start the migration to MMC5983MA or LIS3MDL now.