Hi,A question of how one should interpret Figure 9 and 12 of the ADIS16460 data sheet:One shows the gyro bias with std. dev over temperature when cycled cold to hot, while the other show when the sensor is cooling off from hot to cold.
The figures are nearly identical. Can one conclude that e.g. if a sensor has 0.12 deg/s offset at 0 deg. C when going from cold to hot, that it will be near the same value when going hot to cold? Or could this just as well be -0.07 deg/s, (i.e. no correlation )?If the latter, and I want to characterize the sensor's offset over temperature, I would need two sets of compensations; one for cold-to-hot and another for hot-to-cold. Would it then be sufficient to look at a low pass filtered value of the sensor's temperature sensor to determine which sets of compensations to use? Does the rate of change of the temperature affect the offset?
Sorry for our gap in response. The short answer is that these plots are trying to communicate how a population of parts will behave, across a common set of conditions. We have found that on individual units, the bias hysteresis can be in the 0.1 deg/sec region for these devices. I hope that this helps.
Hello Mark,No problem. I think I'll go for a calibration approach where it is cycled low to high and high to low, while recording the offsets at short intervals. When I then apply the offsets, I will just determine which one to use based on if the temperature is rising or decreasing. I hope the slope of temperature increase/decrease does not affect the offset, though.An extra question: Would it matter if the measurements were just performed without ever resetting/restarting the ADIS16460? I'm guess what I'm asking - is the offset over temperature depending on the starting temperature of the sensor?
If you feel you can't answer these questions in this public forum, please let me know - maybe it would be better to write through email.
We tend to quantify bias drift over time, not specifically through power cycling. Over normal use conditions, the behaviors tend to follow a Rate Random Walk (RRW) behavior, which is one reason that we have not been able to support tighter error metrics on the product datasheet.
For these devices specifically, they do have some drift at turn-on, which follows an exponential decay model, which has an initial error of +/-1dps and has a time constant of 10 minutes. The root cause is a charging mechanism, which also holds true, from the time of power removal. In other words, if a device is fully settled, then you turn it off, it will follow (approximately) the same profile to settle back to its "off state" bias.
I hope that this helps!
Thanks again Mark, it is much appreciated. What I want to avoid by a temperature calibration of the ADIS16460 gyroscope is to fight against some internally applied algorithm inside the IMU which compensates the output of the gyroscope based on how fast the temperature changes. Can I be sure that no such compensations are carried out and that what I'm seeing is the true nature of the (gyroscope) sensor?
The charging mechanism and time constant you describe, this is also the trend described by the Allan Variance of Figure 7, right?
I think those were the last two questions I had.