ADIS16210 accelerometer resolution

I'm evaluating ADIS16210. In fact I want to use pure accelerometer data from device. When I recorded the data I've found strange thing. As you can
see on the picture, there is kind of quantization of the accelerometer values with the step around 40 (values are row counts, not
in g or m/s2). If it is true the quality of accelerometer in ADIS16210 is close to ADXL362, but in the documentation you can find resolution of ADIS16210 equal 61ug/LSB! Could you explain me what is going on? What is quality of accelerometer inside ADIS16210? Is it the same as  ADXL362 and is added CPU for calculating average acceleration  and tilt?


  • 0
    •  Analog Employees 
    on Feb 20, 2014 6:43 PM

    Thank you for posting this.  What sample rate setting are you using for this data?  You statement/question about the ADIS16210 and ADXL362 having the same quality of accelerometer is not true. They use different accelerometer cores; the ADIS16210 should perform much better, especially over temperature and time.  

  • 0
    •  Analog Employees 
    on Feb 21, 2014 7:12 PM

    On the ~40LSB gaps in the histogram:

    This product uses a 12-bit ADC, but captures the bit growth (up to 16-bits) from the average/decimation filter.  The AVG_CNT register controls the average/decimation filter.  Click on the following link to jump to the location in the ADIS16210 datasheet, which describes this register's function:

    http://www.analog.com/static/imported-files/data_sheets/ADIS16210.pdf#Page=12

  • 0
    •  Analog Employees 
    on Feb 25, 2014 1:39 AM

    Comparing the ADXL362 to the ADIS16210

    For the intended purpose of the ADIS16210, high-accuracy tilt sensing, one key specification in the ADXL362, initial bias, offers a key comparison point.  In the ADXL362, this specification is +/-50mg for the z-axis. The core accelerometer in the ADIS16210 will be on the order of 1-2mg.  While some of this behavior can be calibrated, some of it is related to temperature hysteresis and longer-term drift factors.  While I am not in a position to conduct an appropriate analysis of this parameter to support a direct comparison, I suspect that the ADIS16210 will be ~10-20x better, in comparison.  Both products perform their intended purpose, with industry-leading performance for their respective classes.  They are just focused on different markets and deliver different functions/performance levels. I hope that this helps.

    Best,

    NevadaMark

  • I'm using max 512 SPS. You are right, my fault - the quality of this two accelerometers is different, but I afraid, that ADIS is worst because has higher noise density, equal 190 ug/sqrt(Hz) (page 3 of documentation), when ADXL362 has 175 ug/sqrt(Hz).

  • ...

    In the documentation you cannot find information that accelerometer has 12 bit ADC and is converted to 16 bit. Only information you can find is that sensitivity is ~61 μg/LSB on page 9 (http://www.analog.com/static/imported-files/data_sheets/ADIS16210.pdf) . There is only written, that temperature is coded as 12 bit value (page 11). Dominance of ADIS16210 is in the added CPU which calculate average value from the programmed set of samples, but it is only statistical trick.

    The cost of ADIS16210 is 20x higher then ADXL362, but quality is less... I expected very good device according to the price, and I loosed a time and money...

    Regards

    Janusz