Accelerometer How to calibrate the baseline

[koukileo]

I have an acceleration data, but because the sensor's skew will cause the baseline offset (0g), the method I thought of is to collect the background noise first, then find the average maximum and minimum of the signal, and finally subtract it. Obtain the intermediate value c, so I can calibrate by subtracting c from the new data (as shown in the figure). Is it feasible? Or is there any other method?

thank~

-----------------------------------------------------------------------------------------------------------------

My English is not good, the following is my original text

我有一個加速度的數據，不過因為感測器的偏斜會造成基準線偏移(0g)，我想到的方法是，首先採集背景雜訊然後找到訊號的平均最大值跟最小值，最後相減得到中間值c，如此一來我將新數據減去c就可以校準(如圖)，請問是否可行?還是有其他方法呢? 感恩

[koukileo]

If the above method is feasible, how to achieve it?

如果上述方法可行如何實現?

[altenbach]

Can't you just take the mean of all data shown?

(we can typically offer more help if you attach your VI and some data. What is the datatype (array, waveform, dynamic, etc.?)

[koukileo]

Just now I found out that I got the wrong unit when I converted m/s^2 to cm/s^2, which caused me to get the wrong average. Now I have successfully used the average to correct the baseline. I am too stupid. I even proposed such a strange method XD

Thank you for your reply~

[JÞB]

---

@altenbach wrote:

Can't you just take the mean of all data shown?

(we can typically offer more help if you attach your VI and some data. What is the datatype (array, waveform, dynamic, etc.?)

---

Better yet,  just assume the mean is 9.8m/sec^2.  In this case I would even use a Global Variable! Since, it is not going to change appreciably on this planet.

Oh, yeah, the observed "Offset" is actually the same 1gravity of Acceleration that makes us all fall down.

[Bob_Schor]

There are no Stupid Questions!

Some Biomedical Engineering Graduate Students came to me with a little triaxial accelerometer they wanted to use for a Project.  The specs of the device was that each channel (X, Y, Z) had a gain of 0.3 ±10% V/g, and a bias of 1.5 ±10% V.  It didn't bother them that if they placed this device on a table top and measured the "acceleration" along the three axes, they got average values that might be 0.02 g for X, -0.04 g for Y, and 1.05 g for Z, even though the device was not moving!  They'd never heard of (or didn't learn about) "Calibration".

Is your accelerometer 1 or 3 axes?  Do you know the simple way to calibrate the Z Axis?  [Hint -- it takes two measurements, as you need to determine both the bias, the offset when the accleration is 0, and the Gain, how many volts per g.  Use the convenient 1g (or pretty close, but largely time-invariant) constant acceleration available in the Lab.]

Bob Schor

[altenbach]

---

@JÞB wrote:
Better yet,  just assume the mean is 9.8m/sec^2.  In this case I would even use a Global Variable! Since, it is not going to change appreciably on this planet.

Oh, yeah, the observed "Offset" is actually the same 1gravity of Acceleration that makes us all fall down.

---

Yes, I did not look at the  actual values, but if you are planning to use this device exclusively on the earth surface (details) and not in a location where it significantly differs (e.g. on Mars), just assume it as a constant and make it a diagram constant (not a local variable).

As Bob already mentioned, since your mean values seem to be slightly different from the expected value, maybe have exclusively a gain error and could use the knowledge of the gravity to actually calibrate the gain of the sensor in all three orientations. I am sure the device came with instructions for a calibration procedure. And yes, once the gain is properly calibrated, you can subtract the gravity using a diagram constant.

[JÞB]

---

@altenbach wrote:

---

@JÞB wrote:
Better yet,  just assume the mean is 9.8m/sec^2.  In this case I would even use a Global Variable! Since, it is not going to change appreciably on this planet.

Oh, yeah, the observed "Offset" is actually the same 1gravity of Acceleration that makes us all fall down.

---

Yes, I did not look at the  actual values, but if you are planning to use this device exclusively on the earth surface (details) and not in a location where it significantly differs (e.g. on Mars), just assume it as a constant and make it a diagram constant (not a local variable).

As Bob already mentioned, since your mean values seem to be slightly different from the expected value, maybe have exclusively a gain error and could use the knowledge of the gravity to actually calibrate the gain of the sensor in all three orientations. I am sure the device came with instructions for a calibration procedure. And yes, once the gain is properly calibrated, you can subtract the gravity using a diagram constant.

---

A little bit backward on that CA, use a knowledge of earth's mass and a little trigonometry to perfectly level the test bench.  I'd still use a global variable rather than a constant just in case 😉

[altenbach]

---

@JÞB wrote:
A little bit backward on that CA, use a knowledge of earth's mass and a little trigonometry to perfectly level the test bench.  I'd still use a global variable rather than a constant just in case 

---

Don't forget the forces due to the earth rotation and the position of the moon. :D.

Still, looking at the 20% noise in the graph shown, I would say a constant is sufficient. all other effects are <<< noise  😉

[JÞB]

Noise? More likely vibrations.  Sound pressure alone can account for that tenth of a "g".  That's actually, a rather quiet workplace when calculated as 2 to 20 mPa and assuming the sensor itself isn't even a gram of mass. In  reality an accelerometer is just two capacitors in series sharing a common center plate.  Since C=k,the dielectric coefficient,*(area/distance). Since air is easy to come by we use that as a dielectric and k=1.   As the center plate deflects, due to acceleration, the capacitance away from the force increases and the response to the force decreases the distance between the other plates.  The ratio of the total applied voltage and the voltage across either capacitor is nicely linear over Acceleration. 

I only point it out because phyisical environment effects and the simple physics are so commonly overlooked by software developers.  You HAVE TO follow Newton's laws!