Preamps built Into DACs

.... I believe it was probably abandoned because it is very difficult to implement. It puts another layer of complexity and linearity concerns on top of the architecture....

The worlds #1 manufacturer of cost no object DACs just released a >$100K DAC, but they abandoned the optimal implementation of the VC they used in earlier models and reverted an approach that compromises sound quality because doing it right is "very difficult to implement". Not a very plausible explaination

.....Well even the best S/W DSP volume controls seem to cause audible artifacts at more than about -9dB of volume reduction. No amount of dithering and resampling will help IMO....

There is absoltely no consensus on this, nor confirmation in listening tests. Lots of digital guru's will tell you -25db is fine. My personal experience confirms this.

....The besst overall solution is to use a good volume system to reduce the volume to listening levels and then adjust finely for each track using -0 to -9dB of digital volume. Works like a champ....

In principle this seems like a very good approach. I personally would think a 0db, -10db, -20db -30db -40db analog domain attenuation switch and an additional 25db to work with in digital domain during operations would work just as well, but we're splitting hairs.

I note the following from the manual of my Wadia 121, which uses digital VC:

"Best performance is obtained when operating the Wadia 121Decoding Computer Volume Control near the top of its range. If needed, the maximum output level of your Wadia 121Decoding Computer can be adjusted to match the overall sensitivity of your system so that critical listening will take place with the volume control operating near the top of its range. Critical listening should be done when the 4th or higher LED is lit. The maximum out level of the Wadia 121Decoding Computer is adjustable by means of a series of IR commands issued from the Wadia remote control. The Wadia 121Decoding Computer output level is factory set to accommodate the most common range of system sensitivity. If you find that your typical volume level during critical listening is below the 3rd LED on the LED display, it will be advantageous to use a different setting."

The output levels can be adjusted to 4.0v, 2.0v, and1.0v.

This seems to support Steve's statement that "even the best S/W DSP volume controls seem to cause audible artifacts at more than about -9dB of volume reduction." I wonder how he feels about Wadia's approach to allow the user to adjust the output voltage. Since my understanding of electronics is so poor, I won't pretend I understand all of this.

Seems like an excellent approach. However, I believe going from 4.0v to 1.0v represents about 15db attenuation. Most listening takes place in the -20db -40db attenuation range (if you have sensitive speakers could go all thw way up to -60db), so the Wadia would still have to apply at least -25db attenuation in the digital domain, far more than -9db. It seems like the adjustment to output level are to adjust for speakers sensitive during initial setup. These sensitivies have indeed a range of about 15db. The for day to day listening you do digital domain volume control in the -30db - 0db attenuation range.

".....Well even the best S/W DSP volume controls seem to cause audible artifacts at more than about -9dB of volume reduction. No amount of dithering and resampling will help IMO....

There is absoltely no consensus on this, nor confirmation in listening tests. Lots of digital guru's will tell you -25db is fine. My personal experience confirms this."

I have a background implementing dithering and resampling algorithms for image processing applications used by the government and military. Not sure about this statement either. I would expect different yet comparable results in terms of accuracy to the alternative of not using digital if done well. Of course, that's always the big if with anything.

Also, the algorithms needed to accomplish optimal results have been well documented in academia for decades already and are old hat in the image processing world. I have no idea why it need cost a fortune to implement decent software based dithering and resampling to lower volume in home audio gear these days. That's why I like to buy gear that likes to advertise how their gear achieves performance, not just the claimed results. Then one can really know the value of what they buy if they care enough to do the homework.

In the case of DCS, specifically the newer more mega buck than ever Debussy gear, theirs is seemingly some of the most sophisticated home digital audio processing gear and software out there from what I read (with sound quality to match bsed on actual audition) but it comes for a hefty price. I would expect a very good perhaps reference implementation there that noone could fault based on technical approach nor listening.