• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 1 answer
  • Subscribers 46 subscribers
  • Views 250 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADS8900B: Noise Density

Erik Anderson41
Intellectual 915 points
Part Number: ADS8900B

Hi,

Planning to oversample with this ADC and use a digital low pass filter, so I'm trying to understand how much the noise will reduce.  Is there any way I can estimate the noise density of this converter?  For example, using the transition noise spec I get near 11.5uV-RMS.  If I assume this is wideband (white) noise, and I assume that the small signal bandwidth is valid (23MHz), then I get 11.5uV-RMS / sqrt(23MHz) = 2.4nV/sqrt(Hz).  Is this a valid estimate, or is there a valid method to estimate the impact of a digital low pass filter given the transition noise?

Thank you,

Erik

  • 0
    TI__Mastermind 36150 points

    Hi Erik,

    Take a look at the shorted inputs histogram; see figure 20 in the datasheet.

    This data was taken with Vref=5V.  The full scale input range is +/-5V in this case.  The equivalent voltage noise is:

    2ppm*2*5V=20uVrms

    As you can see, this noise follows the expected bell curve.  If you used a simple averaging filter, then you would reduce the noise by 1/SQRT(N), where N is the number of samples averaged.

    Regards,
    Keith Nicholas
    Precision ADC Applications

  • 0 in reply to Keith Nicholas
    Intellectual 915 points

    Hi Keith, thanks for the correction, I wasn't thinking VREF*2 (differential).  Let's say then we oversample 64x and get the 20uVrms down to 2.5uVrms.  But if we implement a brick wall digital low pass filter at say 4kHz, how do I determine how much further the noise will go down?  Without noise density specs I don't know how to figure it out.

  • 0 in reply to Erik Anderson41
    TI__Mastermind 36150 points

    Hi Erik,

    The total noise, 20uVrms, is spread over the Nyquist bandwidth of 500kHz (in the case of 1MSPS).  You are correct that the thermal noise is bandwidth limited by the 23MHz analog bandwidth of the ADC sample and hold, but once sampled, any noise above 500kHz (up to 23MHz) is aliased back down.  

    I think for a first order estimate, you can use 20uVrms/SQRT(500kHz), or 28.3nV/rt-Hz for the sampled data.

    I hope this helps!

    Regards,
    Keith

  • 0 in reply to Keith Nicholas
    Intellectual 915 points

    Ah ok now it makes more sense with the noise folding back down.  Thank you.

    Erik