I’m currently using a pre-amp with 2 mono power amps. My DAC has variable volume control. 99 is the highest setting, which is so when fixed volume with my pre-amp. If I’m using it direct to my mono power amps I tend not to turn it up above 25. I heard Paul (PS Audio) say DAC volume controls are like an accelerator being as they are in digital domain (like mine), whereas analogue volumes are like brakes. Does this matter?
The other thing I’ve read, is that lowering a DACs variable volume less than 75% starts stripping bits. How true is this?
I do notice a difference in SQ with and without my pre-amp. Just not sure which I prefer.
Shorter signal paths surely yield better performance.
Open to views and any technical data ref bit stripping if true.
(Smoke me a kipper, I'll be back for breakfast!)
I believe it is correct that most digital volume controls tend to strip bits when turned down significantly.
At 25% I would say the sq would not be optimal.
Whether the DAC strips bits or not depends on how it implements the volume control. DACs usually have an internal up-sampling filter, so they use floating point samples internally (32 bits or more). Applying volume to floating point samples doesn’t strip anything; that’s how Roon implements volume in DSP, which, according to measurements, doesn’t affect quality. That’s part of what makes a DAC ‘good’.
Sorry, this does not make sense. PCM is an integer, equally-spaced amplitude code. The floating point volume computation has to be converted back to the integer code, with the fractional part being thrown away (maybe by dithering, which reduces aliasing, but still reducing the information in the signal).
After up-sampling and volume control, the floating point samples go through a delta-sigma modulator, which increases the sampling rate further and reduces the bit count to a very low value (say 6 bits, sometimes even one). The original number of bits is not relevant anymore.
This is where a ‘good’ DAC comes into play, i.e. a DAC that accepts at least 24-bit samples. With 24 bits, the quantization errors in the absence of any dithering are no more than -144dB (24 * 6). Not only is that not audible, it’s a few orders of magnitude below any analog component’s performance.
DSD is 1 bit, but sampled at 64 times the rate of CD (that is, at 2.8MHz, as in ‘mega’). In order for the dynamic range not to be affected, a reduction in the number of bits must be accompanied by an increase in sampling rate and by noise shaping to push the additional quantization noise outside of the audible range (i.e. above 20kHz).