Sorry headroom. Could you step back a little. Theory is nice in many ways, but come back when you have used a DSD recorder in earnest. Things do not behave exactly as a non-complete theory predicts.
We dont listen only the Recordings, the real acid test is the Original Sound versus the recorded one.
AND:
The inherent flaw with the DSD amplitude quantization however,
is that it is much noisier than the quantization noise in a PCM system. This problem is addressed later though, with respect to each systems frequency bandwidth, dither and noise-shaping abilities all tools in diminishing the effects of quantization error noise to a practically inaudible level.
Whichever value the system assigns will obviously be a distortion of the signal, and the ensuing quantization error is audible as a rough, granular soundî (Harley 538). DSD unfortunately suffers from more of this effect than PCM; however, as seen later, is able to combat the issue much more effectively.
192 kHz PCM and DSD can displace the majority of this noise above the audible spectrum completely, reducing the audible effects by much more than the 3 dB that lesser quality PCM produces. As previously mentioned,
one major pitfall during DSD quantization is that ìthe quantization error and noise associated with sigma-delta modulation is very large, resulting in significant quantization noise, however, unlike the PCM structures
this ìnoise can be shaped such that virtually all noise power falls outside
the range 0-20 kHz, well above the range of human aural perception
(Nuijten 27). An even further distinction separating DSD and PCM is the ability of DSD filters to shape the dither with a much less complicated linear equalization curve that is not forced to rely on less audible frequency regions below 20 kHz.
Since the sampling process is essentially a quantization of samples on a time scale rather than on an amplitude scale, signals occurring between those instances are cropped. In a 44.1 kHz system the time between each sample is fairly large, with approximately 22.68 μs between each sample. On the other hand, 192 kHz PCM samples are only 5.1 μs apart, and DSD has an even greater time domain resolution with a mere .357 μs between samples. As a result, the 192 kHz PCM and DSD systems are better suited to respond to transients accurately. How this response time actually affects the systems ability to reproduce a transient, as each system was fed a -6 dB block input (click) of a 3 µs duration. The resulting graph shows that the DSD and 192 kHz PCM systems respond the fastest and most accurately; whereas the 48 kHz PCM system not only distorts the signal, but also takes a much longer time to even react. This distinction, most visible in the large width of the 48 kHz sampling frequency reproduction, is audibly apparent in transient events as a ringing or bell-like sound.