Hacker News new | past | comments | ask | show | jobs | submit login

Note that the anti-aliasing filters necessary to cut off audio at 22 kHz distort the audible range, too. Also, what they don't cut off gets introduced into the lowest frequencies as noise (this effect is called aliasing). With a higher sampling rate, you can use less steep anti-aliasing filter slopes that introduce less distortion and noise. Furthermore, that distortion can be limited to frequency ranges above the human hearing threshold. Hence, higher sampling rates can make a huge difference.

There is no such thing as a perfect, distortion-free filter.




Right. Which is why it's a shame that CDs aren't 16/48 rather than 16/44 -- it allows a filter with half as steep a slope.


This is where oversampling comes in, and you get CD players with 196khz 1 bit DACs - we mathematically shift samples to a much higher bitrate (at lower sample width), and then run it through a dac - that allows for a much, much gentler filter, as artifacts are shifted way higher in the spectrum. This is already a solved issue......(even if sales guys a decade or so ago tried to claim it increased resolution - oversampling was all about gentler filters and cleaner sound, not increased resolution, as we know that's impossible)


Oversampling only improves DACs (CD-plyers, sound cards), it does not solve anything for ADCs (mastering). When mastering a CD, the antialiasing filter still needs to shut off at 22.050 Hz and there is nothing oversampling can do to improve this.

You can use really long, linear-phase look-ahead FIR filters that are way better than the analogue IIR filters of the olden days, but the filter slopes still have to cut off between 18-22 kHz, which means some ~60dB+ per octave. A filter like this will always produce audible artifacts. This is pure physics and there is nothing we can do about this.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: