TBH, I find that "corrected" audio more offensive than the original, in most cases. The problem with the kinds of noise that people generally want gone -- hiss, HVAC, etc. -- is that it's mostly random and broad-spectrum. Hiss is normally caused by things like Johnson noise, which is actually useful to repurpose into the entropy for a random number generator. So by definition fairly unpredictable.
I can't recall ever hearing digital noise (bus activity, etc.) on any of my Creative cards. I did hear it on some cheap motherboard (and laptop) sound chips in the early 2000s. My Sound Blasters were just a little bit hissier than my HiFi audio components. If I had my stereo cranked and flipped over to the PC audio input, I would hear the noise floor. Meh. At that point, it was generally loud enough that fidelity wasn't my goal at the moment.
I don't have any super high-class audio interfaces, but I do have a Presonus Digimax FS rack. I guess I should make an evening out of it, and play back some sample clips on all my cards and record the difference on a card-by-card basis. I just never noticed anything being out of sorts back when I was using these as my daily driver rig. Could be rose tinted glasses I suppose.
rasz_pl wrote:used car salesman snakeoil. Neo, there are no stair steps. […]
Show full quote
NewRisingSun wrote:A2. The high quality 16-bit CODEC plays back every detail of the 8-bit sample including the coarse resolution of the staircase waveform.
used car salesman snakeoil. Neo, there are no stair steps.
https://www.youtube.com/watch?v=cIQ9IXSUzuM
https://www.youtube.com/watch?v=pWjdWCePgvA
I admit up-front I haven't followed these links yet. But I'm going to assume this leads to the Xiph guy. This is one of those cases where a little bit of knowledge can cause some great misunderstandings.
The waveform is stair-cased. Digital audio is quantized in the X and Y axis, so I don't even understand how anyone can fathom it as anything else. The lower the resolution on either axis, the more stepping occurs.
If you have a DAC that does not have the reconstruction filter integrated, and you can probe the point where the data becomes an analog waveform again, the stair-case is quite literally right there. I did this on my MT-32, for example, since it uses an old Burr-Brown DAC with an external S&H circuit, and finally the discrete analog reconstruction filter. The point of that filter is to reduce the slew rate and thereby remove the upper harmonics that are not part of the desired signal. They don't go away completely. No brick-wall filter is literally a brick-wall. It's a slope, of varying degree, and there are negative consequences when you push it to higher orders - namely transient response and ringing artifacts.
The Xiph guy is going to show you an oscilloscope image where you can see for yourself that the resulting waveform (with only two data points per cycle!) is not stepped. But you have to understand, this is because the stepping has been filtered out. It becomes a semantic argument "whether they exist at all, then" I suppose, but the difference is mainly important due to the discussion topic here:
Old sound cards play samples at various rates, but didn't always have a filter with a -3dB cutoff at the Nyquist frequency. That can change the sonic profile of sampled audio whose Nyquist is below the fixed filter frequency. You will get aliasing artifacts in this case. Guess where those come from.
Now, most DACs today are more sophisticated than those early ones using resistor ladders and/or S&H latches. So there is no point where you'll be able to measure those steps. It's rather dubious to claim they don't exist at all, though.