Post Go back to editing

ADIS16203 Measure stability

Dear all,

I'm using the ADIS16203 inclinometer and he works fine but I have some problem with the actual "measure stability". When the MEMS is fixed to the plane, the angular data oscillates on 7 - 10 measure steps (max) and it don't have a null average.

I tried to reduce the bandwidth but the oscillation is more or less the same.  I tried to average the results (up to 128 samples) inside to the  microcontroller but the problem persist.

My questions:

What is the best precision (not the accuracy) performance of the ADIS16203? 

From Datasheet:

- Output noise =  0,1 degree rms

- Relative accuracy =  +/- 0,6 degree

How is the register configuration to reach the best precision (= measure stability)?

If I don't need others functionality of this MEMS, there are a possibility to "not use them" and increase the performance of precision? (Do you have some workaround?)

My ideal condition is that the measure oscillates between 1 - 2 step maximum.

My actual registers configuration is:


ADIS_ALM_CTRL : 0x0002

ADIS_MSC_CTRL : 0x0407

ADIS_SMPL_TIME : 0x001A (or Max value)

ADIS_AVG_CNT : Default (I tried 0x0008)

(The Others registers not are changed)

Thanks for your support,


  • Thank you for your post.  Let's start by making sure that I am understanding what you are saying.  Does 10 measurement counts describe the peak to peak variation in the measurement?

    10 counts x 0.025 deg/count = 0.25 degrees (peak-to-peak).

    If we look at the 0.1 deg (rms) noise specification, we can approximate the peak to peak content to be closer to 0.66 deg, so it would appear like you have reached the capability of the device. 
    Let me give this a little thought and get back to you, as I think that it might be possible to achieve your objectives with your own signal processing of the ADIS16203's outputs.

  • Just as a side note.  You probably already know this, but the ADIS16209 provides ~500% improvement in accuracy performance over temperature, yet only has a 50% cost trade-off for that benefit.  Since it uses the same package, has the same pinout and only has minor changes in its register structure, some customers have used both of them on the same board. The noise density is the same but the ADIS16209 has a 50Hz filter in its signal chain so that can reduce the burden on your signal processing in some cases.  I will respond with more insights on processing the ADIS16203's data as well.

  • Hello NevadaMark,

    your statement is correct: 10 steps are peak to peak and reach more or less the stated rms noise.

    I need to achieve the best performance on this specific ....

    Thanks for your support and look forward to your response,


  • Just doing some quick math:

    • 2 counts = 0.05 degrees
    • sin(0.05 deg) = 0.0008g
    • RMS -> 0.0008/6.6 = 0.00013g

    If we review the Allan Variance curve for the ADIS16362, which uses the same sensor and signal chain as the ADIS16203, we can determine that this level of uncertainly is possible, if you average the output data for approximately 2-3 seconds.  I would recommend setting the sample rate to its maximum and reading the data at this maximum sample rate, to make sure your noise is not degraded due to under-sampling.

    SOURCE: ADIS16362 Datasheet, Rev E, Page 8

    Does this answer your question?  If so, our system admins would appreciate you marking the most appropriate response with "Correct Answer" when you get a chance. If this does not answer your question, we look forward to further clarification and discussion.

  • Hi NevadaMark,

    I want to inform you that I'm implementing in the product, the algorithm for the floating window average up to mediate 2-3 seconds of data samples.

    Thanks for your support and as soon as I have more news, I will inform you.

    Sincerely, Rocco

  • Thank you for the affirmation and recognition.  We look forward to hearing about how this works for you!