You must be logged in to post to the forum.

PC: today's grey area of hi-fi experience

Today's asynchronous USB D/A converters, with galvanic isolation from PC, were supposed to make PCs sonically invisible, provided the PC output the bit-perfect data. The things however did not turn out really like that. Even though such DACs are less critical regarding the PC qualities than the earlier generation of non-isolated adaptive USB converters was, the PCs still did not become really unimportant. Some might say, the problem has just moved to another level.

Usually, and for the good reasons, technically oriented audio devotees consider two major contributors to the digital source qualities: the one is the data quality (bit-perfection or bit-transparency), and the other is the signal quality (signal integrity). I spotted these issues some time ago in a separate article.

Indeed, by solving known problems associated with the data and signal integrity, it is possible to achieve significant sonic improvements, and cure many, or maybe even the most of the traditional problems we experienced in the digital domain. Namely, by providing bit-perfect data, and feeding the D/A converter directly by the good clock, and keeping the noise inside the unit low, the digital audio is able to push its subjective sound quality way ahead of what was achieved in the old, and not that good, "perfect sound forever" days. (Notice: The analog audio parts of the D/A conversion process, such as I/V stages, are out of the scope of this topic.)

Still, some of us dealing with the digital audio are aware of certain ways to change the sonic properties of the digital audio source, apparently not changing any performance mentioned above.

And this problem – and yes, this is the "problem" – is older than USB audio. For instance, back in CD days, we wanted to know why numerically identical CDs can sound different. So, before we proceed, may I recall one excellent article from the 90s, by Julian Dunn et al?

Now, back to PC. How can we change its sonic properties?

We can do it both by the hardware, and by the software.

So, firstly about the hardware.

Here is the mail I got some time ago from the Model S USB customer, once he tried the USB conditioner. (Forwarded by Jim's permission.)

"It works! The noise floor is reduced, voices and instruments sound more clear and refined, and bass is more articulate.

I use a Macbook, and I think these types of units work best with non-audio (noisy) sources. I am guessing that is why there is a significant improvement in my case. With a dedicated audio music server, my guess is there would not be much of an improvement."

It was also the time I've been considering making some improvements in this domain. And, as you might be aware, the Audial USB design was slightly revised in 2017, with the second run of Audial USB interface board. In addition to previously used stand-alone (local, or self-) supply for audio clocks, it also brought a stand-alone supply for the USB decoder. Basically, this USB decoder stage is really and purely digital stage, whose output is isolated and re-clocked, so this has no effect on jitter performance or noise levels in the audio part. And that was why this stage was using supply from the USB line.

However, more people were reporting success by experimenting with Vbus conditioners. And once I tried to go with local supply for this stage too, I had no problem to hear the change.

Moreover, I could easily hear the quality of the supply used here. My first and quick effort to experiment with this part was to connect my lab supply, which is the linear supply, so I set it to 5V, and hooked the USB decoder up to it, instead of the Vbus. Now, please note that the PC supplies are usually switching supplies, and switching supplies usually have good bass and dynamics, and smooth tone. As for their bad sides: they are dry, so brass instruments for instance never get lush and fresh, and you might even get tired after some time. With this lab supply, I could easily hear the change. It was more clean and I liked the resolution, but it lacked the scale in comparison to Vbus, and was too thin, up to the point of fatigue. But it was still only the lab supply. So I connected the "real thing" instead. And then there was no way back.

Physically, the USB interface consists of four connections: apart from the supply provided to the peripheral device (Vbus), it has the ground connection, and differential data pair (D+, D-).

A ground may seem most closely related to the Vbus – if we suppose that, in some peculiar way, the noise gets through to the device via Vbus, we can also suppose that it gets through the ground connection too, as a common mode noise. And if this was correct, how would we deal with it?

One may think about two ways. The one is to break the ground connection completely, however there are numerous reasons (too many to mention) why the host and peripheral device, unless they are truly isolated, should share the same ground reference, so that's not what I'd recommend here.

The other is to try to force the stray ground currents to run the other ways. This can be done by inserting a small value series resistor into the ground line. The bad side is that, inserting the resistor here will violate the USB (applies to USB 2.0 and later) impedance requirements (90 Ohm differentially, expectedly set by the 45 Ohm between each data pole and the ground), of course assuming that this impedance is designed properly and is uniform at all, all along the line. But, if we do try this, what does happen, does this change the things sonically? Well, the effects might not be immediately that obvious as with the above described Vbus change, but the answer is positive: it is often quite audible.

So, to picture this:



The same can be also done before the connector, say as a part of the USB cable. Also, a ferrite bead, with its own advantages and shortcomings, may be used instead of the resistor.

(To be continued...)