Contact Us

IMU Selection Guide

IMU Selection Guide

Inertial measurement units and sensor ICs supplied by Superb

An inertial measurement unit request often starts with a short list: six axes, SPI, small package. The first datasheet that matches the list can still be the wrong device, because an IMU is chosen by how it behaves over temperature and vibration, not by the number of axes in the title. Two parts with the same range and interface can differ tenfold in bias stability, and that difference shows up as drift in a heading that was accurate on the bench.

The unit also has to survive the process. An IMU is one of the few components whose performance changes with the way it is mounted and the stress the board carries after reflow, which makes the choice a board level decision.

Where an IMU choice goes wrong

Bias stability and noise density are confused first. Noise density describes short term noise and averages down with filtering; bias stability describes how far the zero rate moves over temperature and time, and no averaging removes it. A part chosen for low noise can still drift out of the heading budget within a minute.

Range is the second trap. A gyro with a plus or minus 250 degree per second range has better resolution on paper, and it clips during a fast rotation that the application considers normal. A part with a wider range has coarser resolution and may need a different filter in the control loop.

Temperature behaviour is the third. Zero rate offset and sensitivity both move with temperature, and the motion differs between parts of the same family. A unit calibrated at room temperature that then sits in a sealed enclosure at 60 degrees starts the mission with an error that has nothing to do with the algorithm.

Vibration is the fourth. Linear acceleration couples into the gyro output, and the coupling depends on the package, the mounting and the internal resonator. A device that is quiet on a desk can produce a rate error on a vibrating frame.

Interface and timing come fifth. Output data rate, filter settings, FIFO depth and the time between a sample and its availability decide whether the control loop can use the data at all. An interface that shares a bus with a high rate sensor can starve the loop when it matters most.

Finally, lifecycle. Inertial sensors are retired like any other part, and a device that is available today with no successor doubles the cost of the next revision. The second source question belongs in the first conversation, not in the last one.

The parameters that decide an IMU choice

ParameterWhy it decides the outcomeWhat a weak comparison looks like
Axes and full scale rangeRange has to cover the fastest rotation without clippingSame axis count, range not compared with the motion profile
Gyro bias stabilitySets the drift that no filter removesZero rate offset at 25 degrees quoted as stability
Noise densitySets the short term resolution after filteringOutput noise quoted without the filter bandwidth
Accelerometer range and resolutionDecides both motion sensing and vibration toleranceRange matched, resolution and bandwidth not compared
Temperature behaviourOffset and sensitivity drift decide the warm start errorRoom temperature data only, no curve over temperature
Vibration and shockRectification error and survival decide field reliabilityShock rating quoted, vibration coupling not mentioned
Interface and output rateBus, rate and FIFO decide what the control loop can useInterface matched, timing and FIFO depth not compared
Supply and powerRails and current decide loop budget and self heatingAverage current quoted at the lowest rate only
Package and mountingPackage stress and axis alignment decide the calibrationSame footprint, mounting and alignment not considered
Grade and lifecycleTemperature grade and successor parts decide the riskCommercial part offered for a sealed outdoor enclosure

How SUPERB PCBA supports an IMU choice

Send the motion profile, the accuracy the product has to hold, the loop rate, the rails, the footprint, the temperature inside the enclosure and the annual quantity. The reply compares candidates against the table above, marks each difference as compatible, needs a check, or needs a board change, and returns every open point as a question.

Families are grouped under sensors and detectors, with the flight control case in the 6 axis IMU sensors for flight control and navigation page. Where a candidate is pin compatible on paper, pin to pin cross reference confirms the pin map before anything is ordered, and end of life routes are collected in the obsolete part cross reference hub.

Provenance is checked before the sample. Reels are counted by X-ray, markings and date codes are inspected, and decapsulation is used where the risk justifies it, because inertial sensors are often re marked in the open market. Verified lots carry a 730 day warranty, and a substitution only enters the bill of materials after written approval.

Proving an IMU on the board, not on the bench

An IMU is qualified after assembly, because the process changes it. Reflow stress, board warpage and the stiffness of the mounting all shift the zero rate, and the axis alignment depends on how the part sits on the copper.

SUPERB Assembly runs that qualification: first article inspection against released data, placement down to 01005, 3D SPI before reflow and 3D AOI after it, X-ray in 2D and CT mode with voiding below 5 percent under the package, and functional test on a fixture approved by your team that logs zero rate, sensitivity and noise at the temperature corners the product sees. The work runs under IPC-A-610 Class 2 and Class 3, ISO 9001, and ISO 13485 for medical programmes, with boards supported from 500 MHz to 77 GHz, lines down to 0.05/0.05 mm and impedance held to plus or minus 5 percent.

SUPERB Tech keeps the calibration and test record against the build, so the next batch is compared with the first one rather than with a datasheet.

IMU questions

Is a wider range always safer?

It protects against clipping and costs resolution. The right range is the one that covers the fastest motion with margin, and the resolution left has to meet the heading budget.

Can two IMUs from the same family share a calibration?

No. Zero rate offset and sensitivity are part per part, and they move with temperature in a way specific to the individual device. Calibration is written against the unit, not the family.

How is an IMU substitution proved?

With zero rate and sensitivity logged over temperature on the assembled board, a noise measurement at the loop bandwidth, and a vibration check at the frequencies the product actually sees. A stationary bench read proves very little.

What if the replacement has a different axis orientation?

Then the firmware mapping and the mounting both change, and the change is quoted with the re calibration and the test plan before the order.

Next step

Send the motion profile, the accuracy the product has to hold, the loop rate, the rails, the footprint and the annual quantity. You get the comparison with each difference marked, the candidate list with availability and lifecycle status, and the test plan needed to release the change.