• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 46 subscribers
  • Views 667 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.

Precision ADC for audio?

Andor Horvath
Prodigy 60 points
Other Parts Discussed in Thread: PCM4220

Hi!

Can you tell me, why a high-speed precision ADC, like ADS1610, couldn't be used with an additional noise shaping, in order to create a nice audio ADC?

This converter's output features 16-bit with still almost 4^4 oversampling over the 48 kHz audio standard sampling frequency. Which means after further decimation filtering a S/N ratio of a 20-bit converter could be obtained. If an additional external noise shaping feedback could be added, the results would possibly surpass a 24-bit PCM converter, wouldn't they?

If such an application exists, could you give me some directions where to find information about it? Thanks a lot, best regards,

Andor

  • 0
    TI__Guru**** 189401 points

    Hi, Andor,

    Hmmmm, let me move your post to the Precision ADC forum to see their thoughts.

    -d2

  • 0
    TI__Expert 6925 points

    Hi Andor,

    You are right that random broadband noise can significantly reduced by oversampling with decimation and digital filtering.

    However, even without random broadband noise, SNR is limited by the converter accuracy set by LSB-size and non-linearities.

    For example, an ideal 16bit converter with zero noise and zero non-linearities will have inaccuracies with input voltages that lay between codes. This accuracy error produces what is known as quantization noise in AC signals. Its approximated value is LSB/SQRT(12) Vrms, which is the value used to derive the well known equation: SNR = 6.02dB*Bits+1.78dB.

    Then, in non-ideal converters, non-linearites affect accuracy even more increasing quantization noise a little or a lot depending on INL/DNL values.

    In my opinion, 16bit accuracy is what doesn’t allow a 16bit converter to be used as 20bit converter for AC signals even if the signal is oversampled many times per effective conversion to remove all random broadband noise.

    Best regards,
    Rafael

  • 0 in reply to Rafael Ordonez
    Prodigy 60 points

    Hi Rafael,

    Thanks for your answer! I might not understand everything clearly, but as I learned the complete equation was

    SNR=6,02dB*Bits*(0,5*L)+1,78dB

    where L is the number of octaves of oversampling. That's why audio delta-sigma-converters have 30-40 dB SNR, and even 1-bit converters with 2...3 MHz internal sampling rate have around 20-30 dB SNR. This noise is then shaped to be mostly far outside the audible range, which gives a far better result (the incredible 123 dB in case of the TI PCM4220). In case of the PCM4220, the distortion is at -120dBFS for most input levels, which shows how much this oversampling thing helps with quantization noise.

    As far as I learned, the problem that quantization noise remains even at high sampling rates, occurs in digital-to-analog conversion, where oversampling is only imitated by ways of interpolation, which doesn't create new information.

    My question is, in a somewhat other way, if it is possible to use a 1...6-bit delta-sigma converters with 30-40 dB SNR for audio, why not use a 16-bit converter with much more SNR, and do further decimation and noise shaping? Best regards,

    Andor