Post Go back to editing

How precise is the Dualaxis AD22302 accelerometer?

Hello everyone

I am a mechanical engineering student currently working on a project,

wich involves working with an AD22302 Dual-Axis accelerometer.

It is capable of measuring +/- 70g on the x-axis wich is the one I am currently using.

My question is:

what is the accuracy of the device? I am able to find the non-linearity in the datasheet (attached to the post)

but that is about it and I need the overall accuracy. Any kind of help will be highly appreciated

- Jesper

  • Hi Jesper,

    Thank you for your post. As you can see in the datasheet, there are a lot of different condition-dependent error sources. Applying them to a composite error estimate requires understanding of those conditions.  With more information on what you are trying to measure and accomplish, I will be glad to walk you through the process of estimating this accuracy.  Can you offer a more detailed explanation of what you are trying to accomplish, along with any specific details you have on the actual acceleration profile (magnitude, time, frequency, if available/relevant).

    Look forward to the discussion.



  • Hi Mark,

    Thank you for the fast response, I am trying to measure the impact force of a punch. I have built a device for this purpose and the basic idea is that you:

    1. Hit the black black plate (Wooden plate with 2.5 inch soft padding to increase contact-time between fist and plate).

    2. The arm will swing back and the accelerometer (yellow plate on top) will measure the acceleration.

    3. The spring under the axle will return the arm to it's original vertical position.

    The magnitude is about 10 g's and I know this is a obvious problem, but my testing rig will not go much higher. I have thought about lowering the weight of the plate though.

    The contact time of a punch is about 50 ms (measured using a highspeed camera), so to peak acc. there is roughly 20 ms.

    Regarding frequency, I have assumed that the pulse from the punch is a step-input and calculated the step-response of the filter and everything seems to be okay with no loss in output signal. If you need futher details let me know and I will do my best to retrieve them.

    Best regards - Jesper

  • OK, this helps! Thank you.  On this type of signal, 70g/50ms, shape probably between a pulse and a half-sine, is the peak acceleration the key objective?

  • Yes, at the moment it is. If I have enough time I will try to find the energy transfered aswell but finding the peak acceleration has highest priority right now.

  • And by the way, I read my peak value wrong, I am reaching about 20 g's in peak.

  • How far off is it from your expectation?

  • If we were being very conservative, I would say that the difference in the peak shock, with respect to the no-shock output, should be better than 10% accurate, assuming that the majority of the energy is coming from the axis in question.  The cross-axis behaviors present another, independent error source that can reach 5% as well.  Since you are measuring 2x off of your present expectation, I would suspect that there are other errors in this process, which are not related to the sensor's behaviors.

  • Hi Mark,

    Last week I made a couple of tests were the difference between expected output and the measured input was 5-20 % off but in the mean time I found that a lot of noise were coming from vibrations from the device itself after it had been hit. And the method of wich I were obtaining expected input acceleration were too inaccurate. So I will start testing again tomorrow and see how much it helped, hopefully I will stay under a 10 % difference in expected and measured acceleration.

  • So I will start testing again tommorow and see if any of my improvements (software filtering and dampening of the metal parts) have helped ** - and then I will return and tell you have it went.

  • That is great!  We look forward to hearing how that goes.