ADXL213 possible noise sources?

I'm using an ADXL213 in an existing design, and I think I'm seeing more noise than I had expected. The ADXL213 is set up with Cx and Cy both 1uF, and Rset is 5.1M. Cdc is 0.1uF. 

I'm using a MCF54415 to measure the output PWM of the ADXL at rest, and taking 100 samples and averaging them. I'm seeing a RMS error in the duty cycle of 0.000715, and I was expecting to see a RMS of 0.0001316. The expected value is based on the 160ug/rtHz value, with BW=4.7, resulting in expected 438.76 ug. At 30% duty cycle per g, that gives me an expected RMS error in the duty cycle of 0.0001316.

I'm going to get the Vcc scoped out, to make sure there's no real noise there. What other common sources of noise can I look for? Or am I calculating the expected RMS error incorrectly?

Thanks!

  • Hi Michael,

    The following document might help, unless you have reviewed it already:

    http://www.analog.com/media/en/technical-documentation/application-notes/AN-603.pdf

    Regards,

    Venkat

  • Hi Venkat,

    Thanks for taking the time to reply.

    I looked over the app note. I had ignored it previously because it discusses a simpler, low-accuracy algorithm, while I'm looking for high accuracy.

    I am taking timestamps at the rising and falling edges of the PWM waveform and post-processing them in order to verify the accuracy that I think I am seeing. I am seeing jitter in both t1 and t2, and although much of it cancels out, there is some residual variation in the duty cycle. When I try to calculate the RMS of the duty cycle variation I come up with a value that is 5 to 10 times higher than what I had expected the RMS error to be. I'm using the noise density (160ug/rtHz) and bandwidth (5 Hz, earlier I stated 4.7 which was incorrect) to derive RMS error in ug, and then I'm assuming 0 tilt and sensitivity of 30%/g when converting ug to change in duty cycle (which introduces some error, but not 5X).

    I was originally asking about other noise sources that could contribute to the observed variation in duty cycle. Vcc scoped out clean, and I can't come up with anything else. Why am I seeing more variation in duty cycle than what I can attribute to the noise density value?

  • Hi Michael,

    A few considerations:

    What is your Vcc - is it 5V? If it’s under 5V the noise gets worse.

    Are you measuring T1 and T2 in the same period otherwise the ratiometricity doesn’t work out.

    Also, can you try and measure the analog output at Cx and Cy and verify if the RMS noise is as expected.  If those are ok then we can look into the PWM measurement by driving the Cx and Cy pins by applying Vcc/2 on them (make sure the voltage source is <<32k output impedance) and check if the PWM output is clean.

    Regards,

    Venkat

  • This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

    Thank you,
    EZ Admin