Digital volume controls SUCK
There's something that's been bothering me for quite some time: some of my analog versions of digital recordings consistently sounded better than my digital, with more resolution. I had convinced myself that redbook can be really good, and even surpass analog in some cases. So something wasn't adding up. During my recent DMA-500 demo I was consistently hearing the very fine details I should be hearing, along with smoother more natural sound, and I know all these existed in the program because the LP reproduces them in my system, but not my digital, at least with the software I played. I was not prepared to attribute these improvements to the DMA-500s, and this is why I have not posted details yet.
To make a long story short, I started exploring details about digital volume controls. This lead me to
http://thewelltemperedcomputer.com/Intro/SQ/VolumeControl.htm which is making claims also mirrored in the opening of this WBF thread
http://www.whatsbestforum.com/showthread.php?3813-Digital-Volume-control (which I was hoping had received more attention).
The basic claims are that resolution is lost with 16-bit volume controls, not so much with 24-bit (the claim is that the first 8 bits shifted - which translates to ~48dB volume attenuation - comes with no original-signal loss), and much less with 32-bit volume controls. What is indisputable is that S/N is affected, but the wider the word the lower the inherent noise floor (true, but relative). There is also plenty of discussion on the net around quantization errors and noise, which are also true, though it's claimed that quantization errors are reduced or eliminated as the volume control word widens (don't know to what degree this is true, but my gut feeling is that this claim cannot be made universally).
So here's where my experiments came in: I was able to increase resolution in my modified Alpha DAC by bringing the volume up from 54.0 to 55.0 (the recommended setting in the manual), to 57.5 and the max at 60.0. But at the same time, the bass over 56.0 or so became overpowering. I had a really hard time deciding which setting offered the highest resolution, but 54.0 would not do, nor did 55.0. So I decided to take a more scientific approach and consider the widely claimed 2V 0dBFS reference (clipping) point by the redbook specification (TRUE???), and the goal was to get 2V out of the DAC with a test tone recorded at 0dBFS. However, all I have is a 1kHz test tone at -20dBFS on a Stereophile CD, but that's also good enough, as long as the measured output is adjusted to 0.2V (0.1V output of this DAC's RCA outputs).
So the volume setting that gave me 0.100V from RCA-out was 55.5, very close to what the manual suggests. The same setting measures as expected 0.2V from the two phases of the XLR output.
The sonic results I get at that level are simply phenomenal, and more importantly, consistently repeatable:
- Exceptional low-level resolution (I am hearing all the details I heard during the DMA-500 demo), which outperforms ANY other volume setting I tried (even higher ones)
- Very smooth highs
- You-are-there presence with great hall ambience
- Natural voices
- VERY articulate and oh-so-complete notes (though not the very last word)
- Lower distortion with orchestral climaxes (fascinating rendition of the Mahler 2nd's finale)
- Vividness
- Equally important, no so-called digital glare
- (and the list goes on and on)
I heard the same drop in resolution with various volume settings when I had the Vivaldi 2.0 in for evaluation, and that DAC also sounded better at max volume (2V output configuration, if I remember correctly). All of these make me want to say the following:
- When we make claims that digital has problems, in any environment we may have been in, I would first question the overall digital set-up, including now the digital volume control setting. That's not to say digital cannot sound bad, but rather, was the set-up optimal to begin with. Next, I would question the transparency of downstream equipment.
- Whether what I am hearing at home is the lack of quantization errors & noise and/or preservation of bit resolution remains to be seen, but the sonic results are unmistakable and repeatable
- I would never drive amps direct with just about any DAC, and would rather figure out the optimal voltage output and then concentrate on a transparent preamp - the former assumes the designer knows what he's doing inside the DAC to start with (and that's a big IF), and the latter is easier said that done. However, I have been convinced that, unlike previous generations of Spectral equipment, the 30SV/400RS are more than capable of passing through the most resolving input signal they can receive - that's a major engineering feat
- Kudos to those manufacturers who eschew the digital volume control, like Spectral and Schiit
- Most important of all, whether every manufacturer adheres to the 0dBFS "standard" is yet another open question, so we really need to know what they suggest for an optimal digital volume setting themselves
Pleased to know what I've done wrong, but my ears leave little room for error
PS: Can anyone point me to the actual detailed redbook specification?