jj@alice.UUCP (07/29/85)
>From allegra!ulysses!mhuxr!mhuxn!charm!prk Wed Dec 31 19:00:00 1969 > >2. About oversampling: The"staircase" waveform that results from >straight reconstruction fron discrete samples exhibits white noise >all over the frequency spectrum, not just at the high end as he >suggests. This noise is inaudible with 14-bit sampling, so it's >not the issue when it comes to oversampling. Oversampling does >not shift this noise to higher frequencies, as he suggests. It merely >allows for less phase distortion. A second thought: why they call >the output filters "anti-alias filters" I'll never know. The aliases >are either prevented by filtering before digitization, or else they can't >be removed at all. Whoa! Now, hold on here! You make a whole new set of aliases when you reproduce the signal in the analog domain. You have to remove those, as well as the aliases caused by digitization. >Oversampling introduces new aliases, but they are >above the original Nyquist frequency and are removed by output filters. Oversampling makes new aliases just like reproduction through a DAC at the original frequency makes "new aliases". When using oversampling, the "new aliases" are removed by digital filtering (which has a linear phase characteristic because of the particular filter design) and the new alias created by the oversampled conversion (which also provides more noise rejection in the original passband) is then removed by a much simpler analog filter. See "Theory and Application of Digital Signal Processing" by Rabiner and Gold for more details. I have work to do. -- TEDDY BEARS MAY BECOME EXTINCT! HELP AN ENDANGERED SPECIES! "It was great when it all began, ..." (ihnp4/allegra)!alice!jj