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.

ADS8167: Noise on internal reference, and on REFby2

Part Number: ADS8167

For a design using the ADS8167 we are looking at using the output of the REFby2 pin into a single supply op-amp, somewhat like Configuration2 of Figure 38 on page 27 of the datasheet. However, we are implementing a band-pass filter with gain of about 50 on the op-amp. When we analyze the noise gain for our REFby2 input we see peak gain of about 62 just above our filter center frequency and our noise equivalent bandwidth is about 1.2 kHz.  We would really like to understand the expected noise on the REFby2 pin when using the internal reference. The datasheet doesn't provide much past Figure 22 and Figure 24. The graphs under typical characteristics state "REFIO configured as output pin, … (unless otherwise noted)" which imply the internal reference is used. However, clearly Figure 24 is not using the internal reference, leaving me concerned that neither is Figure 22.

 

Specific questions (feel free to add more information, I am sure I am overlooking something):

  1. What is the noise density of the internal reference? (at REFIO and at REFP, we are interested mostly between 10 Hz and 10 kHz.)
  2. What is the noise density of the REFby2 output?
  3. The REFby2 has a minimum capacitance of 1 µF, how much higher can this go and how does a higher capacitance effect the noise.

 

If the REFby2 has more noise than we can handle (we are targeting about 14 ENOB at a 3000 Hz sampling rate) then we have various design options. It would be good to have this information while we are still working through some architectural details instead of discovering an issue during testing. (We could shift some gain to an earlier stage, add a separate unity gain stage for the single supply offset, and/or look at external references. All of these options have trade-offs in our system. The relative benefits of each option depend on the noise levels.)

 Thanks, Chip Weller

  • Hello Chip,

    Most of the datasheet measurements include the internal reference, including Figure 22.  As you point out, Figure 24 uses an external reference; we missed adding a note to this drawing.

    Regarding your specific questions, we do not have this data readily available, and may not have any data for noise density.  The design contact is out of office this week and it will be late next week before we can get any additional information.

    The REFby2 buffer was intended to bias the external signal chain to the middle range of the ADC input, with a typical external gain on this node of 1.  

    The reference noise will likely be dominated by 1/f in the 10hz to 10khz range, so a single noise density number will not be helpful.

    Would total noise over 10khz bandwidth help for the reference and REFby2 buffer noise evaluation in your system?

    Regards,
    Keith Nicholas
    Precision ADC Applications

  • Any noise data would be helpful. We have an eval board on order and will try to test the REFby2 noise. 

    Thanks, Chip Weller

  • Hello Chip,

    I will see if I can get some measurement data in the lab over the approximate 10hz to 10kHz range.  Please give me a few days to report back with some data.

    Regards,
    Keith

  • Hi Chip,

    I was able to take some measurements on the EVM.  The approximate measurement bandwidth used was 10Hz to 10khz.  The 10Hz HP filter was implemented using a singe pole RC filter, and the 10kHz filter was implemented with a brickwall digital filter.

    Using this approach, I was only able to get total integrated noise over this noise bandwidth.  For both the REFIO pin (4.096V reference output) and the REFby2 pin (2.048V output), the total noise in this bandwidth was approximately 5.2uVrms.

    I hope this helps.

    Regards,
    Keith

  • Hi Keith,

    Thank you very much for getting me these values. Based on those I think we should be find with our current design approach. (We have bigger noise sources in the system.)

    Thanks,

    Chip