Mazz
100+ Head-Fier
- Joined
- Apr 13, 2008
- Posts
- 295
- Likes
- 10
Quote:
If you educate yourself about digital audio and how it represents an analog signal and what the sources of error are, you'll understand that there are errors in every single step of the process. The trick is to keep them low enough to be inaudible. In digital processing phases, the main aim is to keep low enough that they don't change the value of the least significant bit.
I'd imagine what ATF are saying here is that their implementation is designed to do the upsampling process itself in such a way that the resulting errors (due to computation, rounding, perhaps others) are less than one significant bit in a 24-bit value. This is as good as you can get if you're producing a 24-bit result - because there's no way to express any theoretically more accurate result in the 24-bit output format.
Quote:
If you have a true 24/192 recording, you don't need to upsample because you have a source that's theoretically a little better than an upsampled lower resolution/frequency source. If all you have a 16/44.1 source, you might wish to upsample for reasons covered in other posts.
But the proof is in the pudding. Can you hear a difference, and is it big enough to care about? Different people may have different answers to that question.
Originally Posted by Herandu /img/forum/go_quote.gif Thanks for the ATF link, but they do say in their write up: "...extremely small residual error that is left is designed to be beyond the dynamic range of 24bit audio..". Right. Extremely small... So there is an error as I suspected |
If you educate yourself about digital audio and how it represents an analog signal and what the sources of error are, you'll understand that there are errors in every single step of the process. The trick is to keep them low enough to be inaudible. In digital processing phases, the main aim is to keep low enough that they don't change the value of the least significant bit.
I'd imagine what ATF are saying here is that their implementation is designed to do the upsampling process itself in such a way that the resulting errors (due to computation, rounding, perhaps others) are less than one significant bit in a 24-bit value. This is as good as you can get if you're producing a 24-bit result - because there's no way to express any theoretically more accurate result in the 24-bit output format.
Quote:
Originally Posted by Herandu /img/forum/go_quote.gif Also, why use ATF upsampling to 24/192 when the Wolfsson WM8740 chip used in the DacMagic can already support 24/192 |
If you have a true 24/192 recording, you don't need to upsample because you have a source that's theoretically a little better than an upsampled lower resolution/frequency source. If all you have a 16/44.1 source, you might wish to upsample for reasons covered in other posts.
But the proof is in the pudding. Can you hear a difference, and is it big enough to care about? Different people may have different answers to that question.