As an example. ESS DAC chip has self-noise / output impedance equivalent of 600 ohm resistor. This is the ultimate analog limitation. Your digital domain noise floor is well below of that, no matter how your set the digital volume control.
Now if you have analog volume control with 10 kOhm potentiometer, that will be the one to spoil your SNR by large factor, not the digital volume control...
If you take let's say ASDM7EC-ul at DSD256, it'll will beat your analog domain by large margin...
Hi Jussi, my thinking is based on this:
Lets have a 24bit recording and DAC with 20 bits of analog domain resolution. Lets have a chain without analog preamp, so using HQPlayer volume control as preamp (preatt) with power amplifier directly connected to DAC. So no analog preamp with volume pot is used.
Now I will attenuate to -60 dB, what approx. means shifting the audio data about 10 bits right, with adding 10 zeros at beginning.
Both HQPlayer oversampling filters and modulator increase dynamic range. It is hard to know for me how much exactly, but since it is mostly enough to use -3dB to -6 dB of attenuation to avoid signal limitation, for simplicity I will count with 1 bit of improvement over 10 lost bits in my thought example.
Valid audio data of original 24 bit recording:
rrrrrrrr rrrrrrrr rrrrrrrr
After 60 dB of attenuation:
00000000 00rrrrrr rrrrrrrr rrrrrrrr rr (10 zeros at beginning)
Valid audio data after improving about 1 bit with upsamling/modulation (simplified for the purpose):
00000000 0rrrrrrr rrrrrrrr rrrrrrrr rr (9 zeros at beginning)
Now when we limit the output to 20 bits of DAC resolution, we get only the upper part, containing only 11 bits of valid audio data (max. 66 dB of dynamic range):
00000000 0rrrrrrr rrrr
And the rest of valid audio content appears now lost below DAC analolg noise floor.
And now, after such a strong attenuation, power amp is amplifying analog signal, which contains only part of original audio information, since low level details were lost below DAC noise floor.
Now let think about the opposite situation - small digital attenuation to avoid limitation in HQPlayer and rest required attenuation done by analog preamp.
The goal is to fill upper 20 digital bits with valid audio data so DAC will use its full 20 bits of resolution to create analog output. Then let the required attenuation to be done in analog domain.
If we would attenuate only about 6 dB (instead of 60) digitally and if we would be getting the same 1 bit of dynamic range back by means of upsampling and modulation, we get:
Original recording
rrrrrrrr rrrrrrrr rrrrrrrr
After 6 dB of attenuation:
0rrrrrrr rrrrrrrr rrrrrrrr r (1 zero at beginning)
After improving about 1 bit with upsamling/modulation (simplified):
rrrrrrrr rrrrrrrr rrrrrrrr r (no zeros at beginning)
Now when we limit the output to 20 bits of DAC resolution, we get 20 bits of valid audio data, so 9 bits of audio information more than in the 1st case.
rrrrrrrr rrrrrrrr rrrr
Now if we would further attenuate with analog preamp, we attenuate analog audio content containing much more low level details (maximum what DAC is able to provide). So the signal at analog preamp output will contain also lower level details, which were missing at DAC output in the 1st case, but attenuated. Of course, analog preamps don't behave ideally, so part of that low level detail could be shifted below preamp noise floor. But we cannot generalize how much - it depends on preamp implementtaion.
My example is simplified to be illustrative but you get my point. What's wrong in my thinking?