Analysis of the Analog Circuit’s SNR in the Selection of ADC Bit Resolution In the February 1991 issue of JSHR, Harris, Brey, Chang, Soria, and Hilton reported that speech intelligibility and quality are not distinguishably different when a 12‐ to 16‐bit uniform analog‐to‐digital converter (ADC) is used, or when a 5‐bit or better floating point system is employed. While we agree with ... Letter to the Editor
Letter to the Editor  |   April 01, 1992
Analysis of the Analog Circuit’s SNR in the Selection of ADC Bit Resolution
 
Author Affiliations & Notes
  • Teri A. Hamill
    Texas Tech University Lubbock, TX
  • Thomas P. Barron
    Tech-Sys Software Lorenzo, Texas
Article Information
Hearing & Speech Perception / Acoustics / Hearing Disorders / Hearing / Letters to the Editor
Letter to the Editor   |   April 01, 1992
Analysis of the Analog Circuit’s SNR in the Selection of ADC Bit Resolution
Journal of Speech, Language, and Hearing Research, April 1992, Vol. 35, 477-478. doi:10.1044/jshr.3502.477
History: Received March 8, 1991 , Accepted May 15, 1991
 
Journal of Speech, Language, and Hearing Research, April 1992, Vol. 35, 477-478. doi:10.1044/jshr.3502.477
History: Received March 8, 1991; Accepted May 15, 1991
In the February 1991 issue of JSHR, Harris, Brey, Chang, Soria, and Hilton reported that speech intelligibility and quality are not distinguishably different when a 12‐ to 16‐bit uniform analog‐to‐digital converter (ADC) is used, or when a 5‐bit or better floating point system is employed. While we agree with their conclusions, several methodological points warrant comment.
In a digital‐audio system, noise can be introduced at any step in the process whether from a digital or analog component. In Figure 1, the system steps and the potential noise sources that may be introduced by each component are outlined. Harris et al. have focused on just one source of noise that may occur when converting to and from the digital domain: quantization error due to ADC bit count. However, the digital‐audio system may introduce noise even prior to this point. Aliasing/imaging noise is a case in point. The filter used in this study would appear to be inadequate for the purposes of preventing aliasing/imaging noise. Given the chosen sampling rate of approximately 20 kHz and the selection of a 16‐bit ADC, 102 dB attenuation of the signal at and above the Nyquist frequency of 10 kHz should be provided. The Krohn‐Hite filter used by Harris et al. provides only 3 dB of attenuation at 9000 Hz, and 51 dB of attenuation at 18,000 Hz. A probable result of this filter setting is aliasing noise that would appear spectrally below the Nyquist frequency, and imaging noise above this point (see, e.g., Hussung & Hamill, Miller 1990;). Cauer elliptic low‐pass filters are generally used to prevent such noise.
First Page Preview
First page PDF preview
First page PDF preview ×
View Large
Order a Subscription
Pay Per View
Entire Journal of Speech, Language, and Hearing Research content & archive
24-hour access
This Article
24-hour access