Upconverting and Resampling Questions

[MaxwellsEq]

@GXAlan @MaxwellsEq another plot, this time to illustrate quantisation noise at low levels like -80dB. The samples are stored as integers so we get the red dots instead of the orange dots we would like. This is why dither is applied - it adds noise but improves resolution. To reiterate, this process doesn't add information, but it does turn a correlated source of distortion (quantisation) into an uncorrelated one (noise)

View attachment 409365
If we go to -60dB the quantisation noise is almost gone:

View attachment 409370
And here's -80dB again but with 10 rounds of random dither - note how the white noise is increased but the underlying signal is much clearer:

View attachment 409375

Thank you. I'm fully aware of these facts. But, now, look at the graph again, check the levels and count the bits in play.

 

[mcdn]

Thank you. I'm fully aware of these facts. But, now, look at the graph again, check the levels and count the bits in play.

Reconstructing a nice looking 1KHz sine wave from 4 bits of sample resolution is just how this stuff works. I'm obviously missing something important, would you care to spell it out instead of making vague allusions?

 

[MRC01]

@GXAlan @MaxwellsEq another plot, this time to illustrate quantisation noise at low levels like -80dB. The samples are stored as integers so we get the red dots instead of the orange dots we would like. This is why dither is applied - it adds noise but improves resolution. To reiterate, this process doesn't add information, but it does turn a correlated source of distortion (quantisation) into an uncorrelated one (noise)
...

The same can be done with signals below -96 dB - so low that without dither, all samples would be zero (in 16-bit). An intuitive way to think about this is that dither randomizes the LSB, so 0 and 1 are equally likely. But when you add a tiny signal to this it slightly biases the outcome. When the signal swings positive, even though it's too small to flip the LSB, the sum of signal + random LSB is slightly biased toward 1, meaning it’s slightly more likely to be 1 than 0. When the signal swings negative, the opposite happens.

A few years ago I created an experiment demonstrating this with sample files.

 

[Kiko1974]

I've been an upsampling advocate for years. I upsample all my CD Rips to 176.4 kHz/24 bits using Weiss Saracon. This has given me good and noticiable better sounding CD Rips, until now...

Last August I got SMSL'S D400 PRO DAC that uses the latest AKM chipset which is a two IC combination. The AK4191 does oversampling and Delta Sigma modulation, and the AK4499EX that is the actual converter.

The AK4191 has what is called "Sound Colors". These are not some kind of equalizations or simulations on the Sabre fashion. These Sound Colors give the choice of x128 or x256 (5.6 Mhz and 11.2 Mhz) oversampling and each the choice of two different Delta Sigma modulation.

With both x128 oversampling Sound Colors, upsampling using Weiss Saracon IS really worthwhile and audible better. With x256 Sound Colors, upsampling is NOT worthwhile, the native 44.1/16 and the 176.4/24 upsampled sound the same, maybe even the native ripped played with both x256 Sound Colors sound a bit better and transparent than the Weiss Saracon upsamples, but I still have to test this with more music of different genres.