# Compensating for MEMS Gyroscopic drift

I'm new to IMUs and know that they are prone to high drift, which depends on Temperature, White Noise and Time, but the most uncertain drift is caused by time, and under certain circumstances, drift can be taken as a function of time as we are integrating the raw values.

Here I'm specifically taklking about gyroscopes.

I came up with a solution, which may look absurd but bare with me please.

Let us add two gyroscopes to a rigid metal rod, each at both the ends of the rod, now any of the gyroscope cannot move independently so let us assume an axis of rotation on the rod, doesn't matter where.

To make things simple let us assume orientation with respect to one axis only, let us say x-axis, now the rod is rotating about its axis and we are measuring the orientation of each gyroscope w.r.t x-axis.

Because we only care about orientation and not position, so let us assume the axis of rotation is fixed to one point and only allowed to rotate about it.

I'm in high school so my knowledge of co-ordinate geometry says that the sum of the angles of both the gyroscopes will be same, and does not depends on how much the rod is rotated about its axis, of course the rotation will change the angles of individual gyroscope but the sum will remain same.

So we can take and add readings(angles) as we turn on the gyros now as drift is a function of time, drift will be minimum at the first reading and will be stored as a constant.

Now that constant will be compared with every next reading, deviation from constant value will determine the deviation/drift and we may be able to compensate for it.

What I described above is possible to do or not? I want your thoughts please.

• Foolish question? What question? There doesn't seem to be one. Jun 1, 2017 at 9:26
• We really don't care what you were doing in the school. If you have some question, then delete 90% of the text and ask a question. Jun 1, 2017 at 9:48

No, you can't do that.

MEMS gyros are rate gyros — the output represents angular change per unit time, not total angular change.

Two gyros that are rigidly connected will give the same readings, not complementary readings, so you can't cancel out correlated errors that way.

Most of the errors in the readings from the two gyros aren't correlated anyway. These include offset and scale factor errors caused by manufacturing variations, as well as noise.