What does Bias Repeatability tell me in an IMU's spec sheet?

Question

If there is a standard interpretation of bias repeatability as it applies to a gyroscope, what is that interpretation?

---

This document indicates that it represents the average change in bias between successive turns-on. So if I estimate bias to be 10 deg/hr and the bias repeatability is 1 deg/hr, I could power off the gyroscope, power it on, and expect the next bias estimation to return something between 9 deg/hr and 11 deg/hr (in 68% of trials).

This forum implies that it reflects how bias changes as the gyroscope ages, but it isn't obvious to me why. If initial bias is changing over years of use, but I'm correcting that bias with proper bias estimation each time, this aspect of aging should not affect performance.

---

Alternatively, this spec sheet (for this tactical grade IMU) describes its bias repeatability with this footnote:

Bias repeatability provides an estimate for long-term drift in the bias, as observed during 500 hours of high temperature operating life at +105C.

This seems to estimate the change in bias over a single 500-hour run. So if I leave my gyro on for 500 hours after bias correction, I would observe the bias after that time period to be 252 deg/hr (value taken from spec) on average. If I don't set aside moments in that 500 hours for bias re-estimation, this results in severe gyro inaccuracy.

---

This other IMU shows its "Bias repeatability (1 yr)" as 1800 deg/hr, which could be (1) the average bias change in a continuous 1-year test, (2) the average change in turn-on bias over a year of usage, or (3) something else entirely.

These different interpretations have drastic differences in implied performance. So while I'm analyzing spec sheets to find the right part for my use case, is there a uniform definition of bias repeatability, or does each manufacturer design their own test to reveal different information about their device?

Answer 1

• No, there is not a uniform specification.
• I've worked with electromechanical gyros, and those did have a true turn-on to turn-on bias change. So there would be one bias, then you'd cycle power and there'd be another, etc.
• If you're buying enough, you can always ask the manufacturer for clarification. Sometimes this works if you're a student, too. Otherwise, just accept the fact you're comparing apples to oranges and do your best.

It is typical that if gyro bias is truly that important you either find a way to calibrate it on the fly*, or you go buy a bigger, more expensive gyro.

* If you're doing sensor fusion, you can make this part of your state estimation.