DocumentCode
1967540
Title
Improving high speed analog to digital converter dynamic range by noise injection
Author
Estra, Anthony
Author_Institution
BAE Syst., San Diego
fYear
2007
fDate
17-20 Sept. 2007
Firstpage
669
Lastpage
676
Abstract
The ideal analog to digital converter (A/D) has a dynamic range set only by the number of bits of the converter. The general dynamic range roughly follows the theoretical limit of DR = 20log102n in dB. In this case dynamic range is defined as the ratio between the desired signal and the top of the noise or spurious signal level. For an 8-bit A/D, the dynamic range is close to 48 dB before any additional signal processing techniques. However, as our digital processing systems demand higher processing speeds, the ideal A/D model does a poor job of predicting the real world dynamic range. More and faster high resolution data means that each generation of A/D is asked to run at higher clock rates and with more bits. In order to meet the speed demands, A/D designers must use architectures that have inherent discontinuities in the digitizing structures. The dynamic range model is not a simple digital model. It is now an analog system with all of the black magic associated with Radio Frequency (RF). This non-ideal A/D generates spurs along with the desired digital bits. In the digital world, these spurs show up as mal-converted bits or missing states in the conversion process. The missing states cannot be repaired without redesigning the A/D. Most of the A/Ds in use are already about as good as the vendor could make in the first place. Barring a significant technology advance, the converter will not improve dramatically. However, what if the holes in the desired waveform could be recovered by randomizing the nearby data? This improvement concept is called dithering. It is by no means new. Audiophiles have been using this technique for years. This paper will show that the A/D´s dynamic range can be vastly improved by smoothing out discontinuity boundaries via dithering - a mechanism that removes sharp, discrete discontinuities by injecting white Gaussian noise. Through careful use of noise at the correct level, and in controlled bandwidths, discrete s- purs can be reduced by as much as 15 dB. That is more than two extra bits worth of dynamic range without buying a new A/D converter.
Keywords
Gaussian noise; analogue-digital conversion; white noise; analog to digital converter dynamic range; discontinuity boundary smoothing; dithering; white Gaussian noise injection; Analog-digital conversion; Clocks; Dynamic range; Noise level; Predictive models; Radio frequency; Signal processing; Signal resolution; Signal to noise ratio; Smoothing methods;
fLanguage
English
Publisher
ieee
Conference_Titel
Autotestcon, 2007 IEEE
Conference_Location
Baltimore, MD
ISSN
1088-7725
Print_ISBN
978-1-4244-1239-6
Electronic_ISBN
1088-7725
Type
conf
DOI
10.1109/AUTEST.2007.4374283
Filename
4374283
Link To Document