• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 2 answers
  • Subscribers 47 subscribers
  • Views 321 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

ADC32RF44: Noise Performance

Danilo A.
Mastermind 28850 points
Part Number: ADC32RF44

Hi Team,

A customer wants to know the noise performance of ADC32RF44 similar to the information that can be found at Table 4 of USB-1608FS in this link.

https://www.mccdaq.com/pdfs/manuals/USB-1608FS.pdf#page=20

Is this the same as the signal to noise ratio? 

Regards,

Danilo

  • 0
    TI__Mastermind 18095 points

    The RF ADC noise is specified as a noise spectral density relative to full scale voltage.  The NSD for the ADC32RFxx can be found in the datasheet of the device in table 7.6.  For example, the datasheet specifies -152.3 dBFS/Hz.  NSD represents the normalized noise power that is dominated by thermal noise when the signal is very small or not existent.  SNR performance is typically a bit worse since there is thermal noise plus aperture jitter contributors (since the signal is present and near full scale).

    --RJH

  • 0 in reply to RJ Hopper
    Prodigy 10 points

    Hi , thanks for the response, the question was not about the definition of noise or how it is evaluated.

    The issue was how to equate a specification stated in noise count as shown in the example data sheet to the way

    the TI specify; i.e a 6 counts = xyz dBm rms, when input is shorted to gnd. when connected to a system

    with no input signal, the "noise"  counts goes up to 8, hence the system contributed 2 bits of noise power

    which xyz/2^6 x 2.  How will the TI part shown this when connected to the system with no input.

    I understand there is the background details for the SND, will the TI part shown 3 bits of noise

    and what is cal showing that 2 bits of noise count = say 3 bits of SND as an example.  Apologies if the

    question was not clear.

  • 0 in reply to Philip Cheung
    TI__Mastermind 18095 points

    The noise count is an odd way to specify noise performance.  Perhaps it is sufficient for data acquisition systems like the example you provide, but it does not make sense for a signal chain type device like the ADC32RF44.

    From what I could tell, the noise count method captures a bunch of samples with no input and determines how many iterations the ADC detects a signal above a certain LSB threshold.  I don't know how to equate the RF ADC noise to a noise count, and I would not want to.  Instead, use the noise spectral density and the bandwidth of the system to determine the noise voltage.  For example, with a NSD of -152.3 dBFS/Hz and say a BW of 200 MHz (=83 dB-Hz) then the noise voltage is -152.3 + 83 = -69.3 dBFS.  If needed, this can be converted to voltage given the full scale operating voltage of the device.  The BW is key here as the overall noise voltage is determined by the system BW. 

    Noise is a statistical phenomenon.  In a time domain viewpoint under a normal situation you will see the LSBs toggling "randomly" due to noise. This phenomenon is captured by the ENOB (effective number of bits) spec.  The ENOB spec is 10 bits at lower input frequency with a near-full-scale signal.  It is likely 11 dB or so with no signal.  That means you will see some toggling of the 3 LSBs due to noise.  Statistically, you will see some toggling of higher bits with decreased probability.  For a given threshold,  a "count" for a given sample size could be measured.  But again, I think that this is not the proper way to evaluate noise with this type of device that is suited for signal capturing and not data acquisition.

    --RJH