Hello,

I'm confused about the terminology of in-run bias and Noise Density.

I need to know what they mean in practice.

Lets use the example of the ADIS16405.

I have tested this board over a table without moving it and I got that it rotates something about 3-4 degrees each minute. So I can say that, for my results, the DRIFT of the board is 3-4degrees per minute.

So far so good. Now...

That's the information of the data sheet:

Typical bandwidth (KHz): 0.33

Noise Density (°/s/rtHz): 0.05

Initial Bias Error: 3°/sec

In-Run Bias Stability: 0.007°/sec

My question is: Based on these values, how is possible to calculate the drift of the board? I want to see the math that shows me the same values that I got in my test.

Any help would be extremely appreciated!

Dan

There is no single number that you can use to calculate drift over time for an inertial sensor.

A good start is to look at the Allen variance chart. On the X axis go to the time period you are interested in and see where the curve intersects the Y axis. Generally for a gyro the units in the Y axis will be in °/s, so make sure you multiply this number by the number of seconds you are measuring (the value in the X axis). Note that is is likely the very best case (stable temperature, no movement - so no error due to acceleration sensitivity, etc) and it's unlikely you'll get this performance outside the lab.

Next problem is that in the real world temperature rarely stays dead stable over many minutes. If the temperature bias drift of the gyro is 0.01°/s/°C a 1°C change in temperature over that 10 minutes can result in a 6° heading error!!!

Next, add in real world stuff like g-sensitivity....

Of course, you will not likely get most of this information from data sheets. Most low cost gyro makers do not publish Allen variance charts or histogram data of bias tempco - it's quite expensive to measure a reasonable population of devices and the data can be rather embarassing. So you'll have to do this yourself. If you do not get the results you expected, best you contact the manufacturers Applications team and make sure you have configured the device optimally for the test.

Hope this helps