cliveb;655992 Wrote: > Great post, adamdea. You have very eloquently presented the points that > I feared I might have to try and put across, so I am relieved of that > task. However, I do have one nitpick regarding the above statement... > > Digital filtering has probably come on from when I last investigated > it, but my understanding is that filtering per se is a mathematical > treatment of the signal. Whether you implement a filter in hardware > (analogue) or software (digital), you end up with the same artefacts > predicted by the mathematical model, viz. phase anomalies in the > passband (and the steeper the cutoff, the worse those anomalies). In > other words, if you implement a brick-wall filter at 22kHz, you WILL > get phase errors below 20kHz, regardless of whether the filter is > analogue or digital. That's why oversampling is useful in both the A/D > and D/A processes. > > But there remains the issue that even if you record at high sample > rates, when you resample down to 44.1 for final delivery you have to > put in place that 22kHz brick-wall filter to avoid aliasing. This is > the one aspect of Redbook that I might conceivably be persuaded is > worth eliminating by increasing the sample rate of the delivery > format. > > That said, there is little evidence that even quite serious phase > errors in the top octave are actually audible, so I don't think it > matters in practice. I'm still 95% certain that 16/44.1 is good enough. Thanks for your kind words.
Is the effect you refer to specific to downsampling high sample rate recordings. I can't find a reference to it at the moment but I can't quite see why it should be different from decimation following oversampling where the analog filter doesn't have to be brick wall. With oversampling and digital filtering you only need an analog filter before the A/d and after the d/a. The effect of the digital filtering and oversampling being that the analog filters don't have to be brick wall but can be gentle. In the case of the anti imaging filter it only has to remove images centred around the oversampling frequency which is many times the Nyquist frequency: the other images between are removed digitally before D/a Pohlman (5ed) p. 111 says that for 4x oversampling a 12dB /octave response and -3dB point between 30 to 40 kHz is suitable and will yield no more than +/- 0.1 degrees phase distortion across the audio band. Presumably with higher o/s rates the analog filter's job is even easier. Apologies if this is different from the situation you are describing. -- adamdea ------------------------------------------------------------------------ adamdea's Profile: http://forums.slimdevices.com/member.php?userid=37603 View this thread: http://forums.slimdevices.com/showthread.php?t=89733 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
