I have been plowing through this technology for about six months with the following observations:
1) A good asynchronous USB -> SPDIF interface box (I use the ART Legato) gives me the best sound for RBCD (better than either an SPDIF or AES/EBU dedicated audio card). The things the interface has to do (besides low-jitter buffered and re-clocked output) include isolating the computer from the DAC. I find streaming-mode USB interfaces and interfaces that use USB bus power sound inferior (almost always, jitter) to an asynchronous interface with its own clean power.
Interestingly, USB cables *do* make a difference; the best sound I get is from an ART cable that carries the signals but not the power lines. The presence of the power wires in most USB cables (most of which, excepting audiophile types, make no effort to shield data signals from power lines) had a deleterious effect on the interface, even though the interface made no electrical contact with the power lines. I also had an improvement from an optical USB (self-powered) cable that provided complete electrical isolation.
2) I have built three servers; one is a 'loss' in that it does not produce SQ at the level of the other two, whether by sound card or USB. I thought it was the power supply, but upgrading it did not help. The difference could be in the motherboard construction or power handling or it could be that the two much more powerful servers sound better because they have more RAM, more CPU power, and SSD for the OS (Win7). The 'loss' is a Core2 E5300 with 4GB RAM. The two better-sounding boxes are I5-650's with SSD and 8GB ram. All three use Corsair power supplies, wildly over-spec'd (650 and 850w).
Not sure we understand yet why one computer sounds better than another; laptops at least can be kept off the power lines and in many cases sound better than desktops. While I am no fan of the 'fanless' server with linear DC power (too many clocks in the PC for this to matter much), I do suspect that the more powerful processors produce better sound, especially in a Windows environment when the operating system always (always) has some busy-work to do. De-configuring services, indexing, and restore point processing is tedious and error-prone. Reduced O/S busywork in may have a lot to do with the excellent reputation of Apple boxes as music servers.
1) A good asynchronous USB -> SPDIF interface box (I use the ART Legato) gives me the best sound for RBCD (better than either an SPDIF or AES/EBU dedicated audio card). The things the interface has to do (besides low-jitter buffered and re-clocked output) include isolating the computer from the DAC. I find streaming-mode USB interfaces and interfaces that use USB bus power sound inferior (almost always, jitter) to an asynchronous interface with its own clean power.
Interestingly, USB cables *do* make a difference; the best sound I get is from an ART cable that carries the signals but not the power lines. The presence of the power wires in most USB cables (most of which, excepting audiophile types, make no effort to shield data signals from power lines) had a deleterious effect on the interface, even though the interface made no electrical contact with the power lines. I also had an improvement from an optical USB (self-powered) cable that provided complete electrical isolation.
2) I have built three servers; one is a 'loss' in that it does not produce SQ at the level of the other two, whether by sound card or USB. I thought it was the power supply, but upgrading it did not help. The difference could be in the motherboard construction or power handling or it could be that the two much more powerful servers sound better because they have more RAM, more CPU power, and SSD for the OS (Win7). The 'loss' is a Core2 E5300 with 4GB RAM. The two better-sounding boxes are I5-650's with SSD and 8GB ram. All three use Corsair power supplies, wildly over-spec'd (650 and 850w).
Not sure we understand yet why one computer sounds better than another; laptops at least can be kept off the power lines and in many cases sound better than desktops. While I am no fan of the 'fanless' server with linear DC power (too many clocks in the PC for this to matter much), I do suspect that the more powerful processors produce better sound, especially in a Windows environment when the operating system always (always) has some busy-work to do. De-configuring services, indexing, and restore point processing is tedious and error-prone. Reduced O/S busywork in may have a lot to do with the excellent reputation of Apple boxes as music servers.