Comments on: Quantization and SQNR
https://dwellangle.wordpress.com/2008/06/04/quantization-and-sqnr/
Sporadic musings of an electrical engineerSun, 05 Feb 2012 04:08:13 +0000
hourly
1 http://wordpress.com/
By: dwellangle
https://dwellangle.wordpress.com/2008/06/04/quantization-and-sqnr/#comment-51
Sun, 05 Feb 2012 04:08:13 +0000http://dwellangle.wordpress.com/?p=253#comment-51This is not really a problem of sample rate, but of modeling the quantization error, which is always signal-dependent. Modeling it as an additive source of independent errors facilitates the use of classical linear techniques for circuit analysis.
What is discussed here is the observation that the distribution of the quantization error looks the most uniform when the signal amplitude occupies the full conversion range, but becomes less uniform (or more obviously signal dependent) as the signal amplitude gets smaller.
]]>
By: Georg Stadler (@georgst)
https://dwellangle.wordpress.com/2008/06/04/quantization-and-sqnr/#comment-46
Tue, 31 Jan 2012 17:11:17 +0000http://dwellangle.wordpress.com/?p=253#comment-46Would it really be so much more expensive to use a different pdf for the error?
From a Bayesian perspective (yeah, I am one of those Bayes fans) you are making the argument that the influence of the distribution choice (that’s the prior in the Bayesian context) decreases as more data through the higher sampling rate is available. Do I get that right?
I work on problems where getting those data points is very expensive (solution of systems of differential equations) so the prior has a significant influence on the error distribution.
]]>