Xref: utzoo sci.electronics:18129 comp.dsp:1323 Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!helios!bcm!rice!uw-beaver!milton!whit From: whit@milton.u.washington.edu (John Whitmore) Newsgroups: sci.electronics,comp.dsp Subject: Re: A question about the Nyquist theorm (CD sampling/filtering) Message-ID: <17510@milton.u.washington.edu> Date: 2 Mar 91 01:23:13 GMT References: <11515@pasteur.Berkeley.EDU> <1991Mar1.004711.15100@watcgl.waterloo.edu> Organization: University of Washington, Seattle Lines: 54 In article <1991Mar1.004711.15100@watcgl.waterloo.edu> cjwein@watcgl.waterloo.edu (Chris J. Wein) writes: >In article wilf@sce.carleton.ca (Wilf Leblanc) writes: >>jbuck@galileo.berkeley.edu (Joe Buck) writes: >>>Example of CD salespeak: pushing oversampling as an advanced technical >>>feature. Oversampling is simply inserting zeros ... >>>Still, some sales critters think it's an advanced technical extra. It isn't, mainly because it's NOT an extra. All CD players oversample, and use a FIR filter (i.e. digital filtering). The cheapest, oldest ones use 2x oversampling, and don't advertise the 'feature'. >>This kills me too. Especially 8x oversampling ! >>If you want to spend enough money, you can get very near linear >>phase response with an analog filter. So, you are right). Not exactly. You'd have to spend money on things like thermostats to keep a really high Q filter (for a brickwall filter), because REGARDLESS of cost, ALL the components in an analog filter are poorly controlled. Typical digital frequency accuracy (if you aren't trying hard) is 0.001%; that kind of performance is equivalent to an analog pole with a Q of 100,000 (which is not feasible). >I also understand that oversampling increases the 'transition region' >of the filter thus allowing for lower order filters. > >As for the CD's that do not oversample, what type of filter is generally used? > [more about why this would be a difficult filter to make} As I mentioned above, there ARE no non-oversampling CD players, for exactly the reasons you gave. A steep-slope cutoff filter with good passband linearity is going to require a bevy of trimmed capacitors/inductors, and would (1) cost mightily, (2) require difficult factory adjustments, (3) not survive shipping without needing readjustment, (4) fail when the room temperature changed. The last 2x-oversampled CD player I took apart had four trimmer components (inductors) in the filter. 4x-oversampling filters typically have merely selected components (not trimmed). I'd trust the long-term reliability of the 4x models before the 2x ones. IMHO, 8x and 16x oversampling are simple hype; the available capacitors and other filter components cannot be characterized accurately for the range of frequencies involved (20 Hz to 200000 Hz) so if I were designing a filter for 'em, I'd just use the same design as for the 4x units. The part of the filtering that ISN'T hype is the digital part (the number of taps in the FIR {Finite Impulse Response} filter is a major factor in both cost and performance). None of the sales info ever mentions this, though... John Whitmore