Bartok as a preamplifier

On preamps, and digital volume control… Considering for a moment the role of a pre-amp in a digital audio playback system:

  • To select different sources.
  • To level match source and power amplifier.
  • To set the listening volume.

When comparing a DAC connected to a power amp directly vs with a preamp in line, there are potential benefits to the addition of a preamp:

  • Interfacing between a source which has only an unbalanced output to an amplifier that only has a balanced input or vice versa.
  • To allow a “difficult” amplifier input and/or cable to be driven. This is more likely to be an issue if the cable is long and, of course, different preamps will have different drive characteristics.
  • The preamp has tonal characteristics the listener likes.

The output drive issue is not a problem with a dCS DAC. Our balanced output stage has very low output impedance (less than 1 Ohm) and very high current delivery. It’s really very unlikely that any preamp would out-perform in these regards and any input/cable combination that caused it difficulty would be questionable in itself.

The preamp categorically cannot reveal more information (be it frequency content, dynamic resolution etc.) from the signal than by connecting DAC to power amp directly – it simply adds in noise, harmonics, frequency response aberrations etc. that aren’t in the original recording that the listener finds pleasing – euphonic distortion. Of course what I cannot do is tell you that your ear does not prefer a particular sound. That is very a very personal preference – but from a purely technical perspective, it is a degradation to sound quality (albeit a potentially psychoacoustically pleasing one). I would hope that we could all agree a DAC should be a transparent link between music and amps/transducers, so attempting to voice the unit to produce such distortions is counterintuitive.

Level matching is less easy to dismiss. If the amplifier were very sensitive, to the extent that our volume control was having to be backed off by 30dB or more, there might be a case for inserting a preamp. In all other circumstances, our digital volume control will introduce less crosstalk, noise, distortion and gain and frequency response errors than a preamp. In the vast majority of cases however, lowering the volume control by more than 30dB would suggest a gain issue elsewhere in the system which would need to be addressed.

Provided this is the exception not the rule (such as night listening), use of the digital volume control inside the dCS DAC would be technically preferable to then permanently having a preamp in the chain for the sake of volume control. Analogue volume controls are definitely no free lunch and come with their own limitations to consider - such as that more analogue componentry in the signal chain affects the frequency response in some way, adds noise and likely distortion as well. Finally, contrary to popular belief, an analogue volume control does not remove the problem of signal to noise - any real system will have some degree of self-noise.

Now, the points that have been made about digital volume controls throwing away bits and decreasing dynamic resolution – these are not correct. The idea that a digital volume control truncates the signal and throws away bits to achieve a volume reduction would only be true if the volume control was to take place before the filtering stage in the DAC, which is not the case – at least with a dCS DAC. Considering how a digital filter operates, there are hundreds, thousands or potentially even millions of multiples/accumulates throughout the signal path. The volume control is effectively just the last one in this chain.

The architecture of a dCS DAC also means that the interface between the output of the filtering stage and the input of the DAC is carried out inside the main FPGA, so we don’t have to deal with fixed width interfaces which are present when for example sending digital audio signals between ICs. We have much more control over the whole process.

As an example, consider a 16 bit sample from a CD, along with a 16 bit volume control. All possible permutations of volume and sample are guaranteed to fit within 32 bits. Even if the signal was upped to 24 bit, there is still a digital noise floor of -192dB, with zero distortion or quantisation artefacts, assuming it is dithered properly. This is way below any real-world analogue noise floors.

The point to consider related to volume control however is that the DAC itself has an analogue noise floor, which is fixed. Reducing the signal by 30dB for example means the ratio of signal to the analogue noise floor of the DAC is reduced. This does not mean the DAC is simply throwing away bits to attenuate the volume. Another point worth considering is that with some DAC types (notably ladder DACs) the distortion products of the D/A conversion structure are essentially fixed. They become proportionally bigger as the signal gets smaller (i.e. when the volume is turned down), so poor low-level linearity is an issue. Not the case for the Ring DAC, as the distortion goes down faster as the signal gets smaller.

16 Likes