The ICM-42688-P is a TDK InvenSense six-axis motion sensor combining acceleration and angular-rate measurements. It belongs in an IMU selection, rather than an FPGA selection. For a new board, the useful questions are whether its measurement range, timing, noise and host interface fit the motion being recorded—and whether the firmware preserves those settings when decoding data.
Electrical and mechanical reference
| Measurement channels | Three accelerometer axes and three gyroscope axes |
|---|---|
| Accelerometer ranges | ±2, ±4, ±8 and ±16 g |
| Gyroscope ranges | ±15.625, ±31.25, ±62.5, ±125, ±250, ±500, ±1000 and ±2000 °/s |
| VDD and VDDIO | Both 1.71–3.6 V; 1.2 V is outside the specified supply range |
| Package | 14-pin LGA, 2.5 × 3.0 × 0.91 mm |
| Specified temperature | −40 to +85 °C |
| Host interfaces | I²C up to 1 MHz; SPI up to 24 MHz; I3C up to 12.5 MHz |
| FIFO | 2 kB; packet format determines the sample capacity |
These values are summarized from TDK DS-000347, revision 1.6. The current TDK product page also lists motion functions and links the current datasheet revision. Bus clock, measurement output rate and signal bandwidth are different settings; a faster host clock does not itself improve sensor accuracy.
Data format: keep resolution separate from accuracy
The normal sensor registers hold 16-bit data. A higher-resolution FIFO format carries 19 meaningful gyroscope bits and 18 meaningful accelerometer bits in 20-bit fields. The parser must follow the enabled packet format, rather than treating each axis as an interchangeable 16-bit word. Resolution describes the encoding; it does not remove offset, noise, temperature drift or mounting error. See the FIFO packet descriptions.
As a software check, choose a full-scale setting and create a conversion test with positive, negative and zero raw values. Record the range beside every captured dataset. A range change without a corresponding conversion change can produce a clean-looking trace with the wrong physical scale. Test byte order and sign extension separately before introducing filtering or orientation calculations.
A worked transport budget
Consider an illustrative logger that writes a 20-byte record at 1,000 records/s. The payload alone is 20 × 1,000 = 20,000 bytes/s, or 160 kbit/s. If firmware requests service after 40 records, that represents 40 ms of acquired history and an 800-byte transfer. This is a proposed host record format, not a claim that every FIFO configuration produces that packet.
The engineering consequence is that the scheduler must reserve time for transfer overhead, other bus users and interrupt latency. Measure the longest service delay, not just average CPU load. Choose a watermark below the usable FIFO limit and log overflow events. A bus that passes a short bench read can still lose data when flash writes or radio transactions delay the sensor task.
Pinout and board integration
Use the manufacturer pin table and package drawing as the schematic and footprint authority. In particular, INT2, FSYNC and CLKIN share a function selection on this device. An external clock can improve sample timing, but it needs the documented pin configuration and timing settings. Do not add an assumed exposed thermal pad or ground unspecified contacts because another LGA component uses that arrangement. Pinout, circuit and assembly reference.
Check power at the sensor pins during startup and during bus traffic. Follow the reference bypass network, then verify placement and return paths on the actual PCB. Pull-up values are a bus-level choice: consider capacitance, rise time and sink current instead of reusing one resistance for every board. A logic analyzer should confirm transaction timing; an oscilloscope is needed to assess analog edge quality and rail disturbances.
A practical motion-data acceptance test
- Start with raw acceleration and rate records plus timestamps. Confirm that moving one physical axis produces the expected signed channel response. Store the board-to-sensor coordinate transformation explicitly.
- Capture stationary records in several orientations. A stationary gyro should be evaluated for bias and drift; acceleration includes gravity, so zero on all three axes is not a sensible stationary acceptance target.
- Repeat at the application temperature limits and with nearby motors or radios active. Separate vibration entering the mechanics from communication errors and from software conversion errors.
- Only then introduce an orientation estimator. Six-axis measurements do not directly provide an absolute magnetic heading; yaw drift needs to be addressed by the system design.
Frequently asked questions
Can I power its I/O from a 1.2 V rail? No: the specified VDDIO minimum is 1.71 V. Evaluate a suitable level-translation arrangement if the host uses a lower voltage domain.
Does a larger FIFO data word guarantee more accurate motion estimates? No. The word format can preserve finer changes, but the complete error budget still includes sensor behavior, sampling, mechanics and the estimator.
Selection and purchasing
Compare a proposed replacement against the complete schematic, package drawing, interface protocol, driver and measurement requirements. A shared sensor category or package outline does not establish interchangeability. Record the candidate and the test evidence in the engineering change record before substitution.
For ICM-42688-P, send the complete ordering code, quantity, required delivery date and any packaging or documentation requirements through the WWDParts RFQ list. Price, availability and delivery terms are confirmed in the written quotation.
Manufacturer references
Technical content reviewed October 12, 2026. Worked examples are design calculations, not measurements of a customer board. Consult the manufacturer documentation index for newer revisions and applicable errata before design release.