Reply to thread

Message

<blockquote data-quote="arion_p" data-source="post: 642704" data-attributes="member: 45945">Normally descent D/A converters will do at least 3rd order oversampling to at least 96Khz or even 192Khz (while Sigma-Delta converters usually oversample to at least 768Khz). If however the D/A does not do any oversampling, or does it poorly, doing our own upsampling might help a bit. To understand why it sounds better, consider a sine wave at 16Khz sampled at 48Khz, a single period of the sine wave consists of only 3 samples! The end result is a square wave with severe intermodulation distortion. Now if you resample that to 192Kh using a high quality 3rd or 4th order filter or even better using wavelets the result is 12 samples per period that more closely match the sine wave. That being said, a high quality resampler is hard to implement, especially in real-time. So I still think resampling in the audio renderer is not much use and it is better left to be done by a real DSP in the amp.</blockquote>

[QUOTE="arion_p, post: 642704, member: 45945"] Normally descent D/A converters will do at least 3rd order oversampling to at least 96Khz or even 192Khz (while Sigma-Delta converters usually oversample to at least 768Khz). If however the D/A does not do any oversampling, or does it poorly, doing our own upsampling might help a bit. To understand why it sounds better, consider a sine wave at 16Khz sampled at 48Khz, a single period of the sine wave consists of only 3 samples! The end result is a square wave with severe intermodulation distortion. Now if you resample that to 192Kh using a high quality 3rd or 4th order filter or even better using wavelets the result is 12 samples per period that more closely match the sine wave. That being said, a high quality resampler is hard to implement, especially in real-time. So I still think resampling in the audio renderer is not much use and it is better left to be done by a real DSP in the amp. [/QUOTE]

Verification