In Reply to: You can lead a horse to water ..... posted by Wavy Davy on August 27, 2003 at 19:54:26:
"many digital filters fill out the extra bits in the expanded word length with some form of dither"That is not how it works. The extra bits provide intermediate points, between bit increments of the base data, in the oversample process. These points are *calculated* by a process called "convolution," which takes a set of values of the media data, seen in real time within the FIR filter's "window," then multiplies those values by the corresponding window coefficients defined in the filter (it digitally depicts the filter "function"- most often the "sinc" function), and finally adds these multiplied values to determine the value of the interpolation point. If the base data is 16 bits, and the coefficient set is 8 bits (it could be more than that and then "truncated"), the resultant values would be 24 bits. No dither is involved.
Now a delta-sigma DAC *could* use dither in the one-bit pulse-density modulation conversion, but I personally do not see reason to this, unless the bitstream itself is *unable* to capture the full resolution of the signal. Dither is useful when the conversion is to a resolution **lower** than before, and the dither uses noise to mix in with the higher resolution media to create LSB "triggers" for the lower-resolution media, to improve the signal linearity below the lower-resolution media's least-significant bit. But it *must* be a "downconversion," for this to have benefit. If the delta-sigma conversion *does* use dither, that is an indication that the one-bit stream *lacks* the resolution of a conventional D/A conversion, and needs the dither to minimize the loss of resolution.
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- But Can't Make Him Drink.... - Todd Krieger 08/27/0322:10:25 08/27/03 (5)
- Re: But Can't Make Him Drink.... - Werner 22:19:27 08/27/03 (4)
- High Resolution Signal versus Upsampled Signal - Todd Krieger 22:52:10 08/27/03 (3)
- Re: High Resolution Signal versus Upsampled Signal - Werner 23:02:15 08/27/03 (2)
- Think About It... - Todd Krieger 00:57:23 08/28/03 (1)
- Re: Think About It... - Werner 01:38:26 08/28/03 (0)