# Speedmeter using acceleration sensor

#### kubeek

Joined Sep 20, 2005
5,785
Few months ago, I got some samples of a two-axis accelerometer, which has linear outputs of 1V per g.

I want to feed the output of one axis to an opamp connected like integrator with a low time constant. I am not too sure, but the output of the integrator should be proportional to actual speed.

Do you think it would be possible to create a speedmeter like this?

problems that probably will occur:
1) need of reseting to zero every time the motion stops.
2) problem with centrifugal force when the meter is not exactly in axis with the motion

#### hgmjr

Joined Jan 28, 2005
9,027
kubeek,

Can you provide a part number for the accelerometer that you will be using?

hgmjr

#### kubeek

Joined Sep 20, 2005
5,785

#### beenthere

Joined Apr 20, 2004
15,819
The problem is the signal going to zero whenever a change in motion is not ongoing. You might find a microprocessor is handy to keep track of the incremental changes in acceleration for the speedo function. No matter how good your analog stuff is, it will have trouble holding a peak value and responding to small + and - changes.

Using the other axis lets you keep track of velocity. If your A to D and computer are really good, you can use it to navigate ( submarines do this, with a third axis for up and down).

#### kubeek

Joined Sep 20, 2005
5,785
Using the other Y axis is a problem, because when the motion is moving in a circle at constatnt speed, the Y axis will keep + signal, so if I computed the total acceleration like sqrt(X*X+Y*Y), the velocity made out of total acceleration will keep increasing even though the real velocity is constatnt.

I don´t have the funds nor experience to use A/D and processor, also it is just my "theoretical excercise".
Can you suggest some really good opamp to be used in the integrator?
What type of cap should there be used? What about using artificial cap made of opamp?

I was thinking to use magnets to attach the future device to a wall of a tram, and see how precise it is and if it has to be zeroed often.

#### hgmjr

Joined Jan 28, 2005
9,027
When the velocity is constant then the acceleration in the Y and X direction are both zero therefore the magnitude of the acceleration as computed by sqrt(X*X+Y*Y) will equal zero. This seems to be the result you would expect.

hgmjr

#### kubeek

Joined Sep 20, 2005
5,785
When the velocity is constant then the acceleration in the Y and X direction are both zero therefore the magnitude of the acceleration as computed by sqrt(X*X+Y*Y) will equal zero. This seems to be the result you would expect.

hgmjr
Lets say the sensor moves on a circle. X axis is tangential to the circle, so Y axis points to (or from) the center.
The magnitude of velocity is constant, in czech we call it uniform motion on a circle.

The X sensor will show a=0, the Y sensor will show acceleration proportional to speed and radius.
The acceleration showed by Y will be non-zero, so the magnitude of acceleration will also be non-zero.
If I integrated the magnitude of acceleration, it will produce wrong numbers, because the magnitude of speed is constant.

#### hgmjr

Joined Jan 28, 2005
9,027
From your description, it sounds as though your frame of reference is always changing. You stated that Y is always pointing to and from the center of the circle.

If your frame of reference is always changing then I think you will need to take that into account in your calculations. Once you account for the constant change in the frame of reference I think the calculation error will disappear.

hgmjr

#### kubeek

Joined Sep 20, 2005
5,785
If I correctly understand, than the frame of reference is the sensor, not the point it is circling around.

If you instead of the sensor used a ball on a spring, the spring will became longer when the ball is rotating, because it indicates the force being applied to the ball. And this force is created by the of acceleration.

#### hgmjr

Joined Jan 28, 2005
9,027
I was thinking that a more appropropriate frame of reference would be centered on the object in motion but its X-axis and Y-axis would always point in the same direction relative to a fixed frame of reference located elsewhere on the same plane as the one containing the circle.

hgmjr

#### beenthere

Joined Apr 20, 2004
15,819
The output from Y in a circular path is analogus to a Z-axis oriented toward the center of the earth. The ball is a good indicator of acceleration, including on the Z axis, if one includes a spring scale for that motion. Sensing the displacment of the free-hanging ball gives the product of the X- and Y-axis accelerations.

#### kubeek

Joined Sep 20, 2005
5,785
This is in fact the way how this sensor works.

So does anybody have suggestions for the parts?