What is "USB done right"? Does USB have a sonic signature?

caesar · Jan 5, 2018

Some folks, who are really into digital, say to those still using SPDIF or AES connections: "You haven't really heard USB until you have heard it done right!".

Yet at the same time we all have heard a slight variation of the above quote - "doing DSD right", "doing tubes right", "doing SS right", etc. Despite the proclamations, ones preferences for the sound of a particular technology usually triumphs.

Does USB have a sonic signature, like tubes vs. SS, analog vs. digital, or DSD vs. PCM?

Are people talking about subjective elements of USB or is this purely a technical subject?

What exactly does doing USB "right" mean?

Hifi Boy · Jan 5, 2018

Honestly, I'm not sure what done right means, but there sure are differences between different USB cables.

Just recently, I purchased a Hugo 2 and its stock USB was clearly inferior to a Moon Audio Silver Dragon which I also bought additionally. The change was not night and day but you could tell the difference. The Moon Audio cable was cleaner sounding.

Empirical Audio · Jan 5, 2018

Unfortunately, there are a lot of things that can affect SQ with USB, including:

1) USB computer or server port quality
2) playback software APP and even the rev. level - particularly if it uses the OS software stack or has a poor volume control DSP function
3) computer or server power supply
4) USB cable
5) Background apps running on the computer
6) USB interface on the DAC or converter - the Master Clock and associated circuits, whether it is galvanically isolated, does it use the power in the USB cable and if so, is there a better supply available for that, does the interface have good power delivery to keep the jitter low, is the implementation good (no cuts in the PC board ground-plane etc...

With everything else the same, it's usually #6 where you will hear large differences.

This is one of the reasons that I have started offering Ethernet interfaces in addition to an XMOS USB interface. There are still a few hoops to jump through to get optimum SQ, but much less than with USB IME.

Steve N.
Empirical Audio

analogsa · Jan 5, 2018

Empirical Audio said:
1) USB computer or server port quality
2) playback software APP and even the rev. level - particularly if it uses the OS software stack or has a poor volume control DSP function
3) computer or server power supply
4) USB cable
5) Background apps running on the computer
6) USB interface on the DAC or converter - the Master Clock and associated circuits, whether it is galvanically isolated, does it use the power in the USB cable and if so, is there a better supply available for that, does the interface have good power delivery to keep the jitter low, is the implementation good (no cuts in the PC board ground-plane etc...

An excellent list. Once the music reproduction through usb, or for that matter any other interface becomes independent from those factors, one may claim it has been done right.

Perhaps a dac with a large enough buffer to store entire albums may be such a solution. Especially if the time to fill up the buffer is not excessive. Only this will kill an entire industry for interface improving devices

Legolas · Jan 5, 2018

USB done right, I haven't heard it yet, not if you compare it to the best CDP sounds quality. Remember USB was never in intended for high quality audio, it was to connect cameras and small ext drives etc. Yes Ethernet wasn't either but it was designed for fast data flow over long distances.

Possibly there is another connection that is even better? I2S? Dunno. A new one may be designed in the future, and if it is, then at least it will be designed just for high quality audio, not some hotch botch multi connection fits all.

My opinion, drop USB altogether.

Brucemck2 · Jan 5, 2018

astrostar59 said:
My opinion, drop USB altogether.

My experience has been the opposite. For all it's faults, I've found USB sounds better than very high end compact disk players (the last time I did a head-to-head the then top of their line 800 series Meridian transport), and, the user interface is considerably better. YRMV

microstrip · Jan 5, 2018

caesar said:
Some folks, who are really into digital, say to those still using SPDIF or AES connections: "You haven't really heard USB until you have heard it done right!".

Yet at the same time we all have heard a slight variation of the above quote - "doing DSD right", "doing tubes right", "doing SS right", etc. Despite the proclamations, ones preferences for the sound of a particular technology usually triumphs.

Does USB have a sonic signature, like tubes vs. SS, analog vs. digital, or DSD vs. PCM?

Are people talking about subjective elements of USB or is this purely a technical subject?

What exactly does doing USB "right" mean?

Good subject, but you are mixing apples and oranges in the question. We can not compare digital to analog conversion or amplification, that result in different electrical waveforms that sound subjectively different, with digital transmission, that results in equal digital contents and is theoretically perfect.

IMHO the USB does not have a sonic signature per se - but in practice asynchronous implementations have a wide margin of variation and non-predictability, and noise from them leaks to the DAC's, subjectively affecting sound quality.

Ethernet has signals of very different bandwidth - probably for technical reasons it is easier to deal with their noise or maybe it simply has a spectra of noise that does not affect the DAC so much. It is not difficult to design a box that converts USB in Ethernet - would it still have an USB signature?

Berkeley Audio have a separate box to convert USB in AES/EBU - should we consider it "right" USB? I used it with Devialet and it sounded subjectively better than the Devialet USB implementation when using just a cheap portable PC as a server - never tried the top optimized servers.

Another big obstacle to establishing rules is that the DAC performance is strongly dependent on the system where it is connected - and our systems, even non tweaked ones, are far from uniform in terms of ground and noise.

Empirical Audio · Jan 5, 2018

analogsa said:
An excellent list. Once the music reproduction through usb, or for that matter any other interface becomes independent from those factors, one may claim it has been done right.

Perhaps a dac with a large enough buffer to store entire albums may be such a solution. Especially if the time to fill up the buffer is not excessive. Only this will kill an entire industry for interface improving devices

It is not only jitter at fault, so the buffer might help, but not fix it entirely. The DSP functions and noise across USB interfaces is also the problem. Common-mode noise is definitely a factor in USB interfaces that are not galvanically isolated. There are quite a few filters available. I even sold one for a long time.

Ethernet has both transformer isolation, although not perfect, and packet buffering with error checking. It is also less likely that the player software uses the audio stack in the OS. That is the reason that it is less problematic. It still has the jitter issue. Some of us have dealt with that effectively and others have not.

Steve N.
Empirical Audio

analogsa · Jan 6, 2018

An album sized buffer implies the interface can be disconnected during playtime. Neither incoming jitter, nor CM noise can then be an issue. Or OS/player. But there is the inconvenience of waiting for the buffer to fill up.

As for jitter using async usb, cannot see how incoming jitter can be an issue at all. The galvanic isolators with the ridiculous and dramatic cuts in the pcbs otoh... not so sure they isolate cm noise at all.

Ethernet is probably a better kludge on the road to buffer play than usb but the issue of keeping the listening space clean from rfi remains. I routinely switch off routers/switches whenever LPs get played and the improvement is not subtle. Even though all power supplies are linear.

Empirical Audio · Jan 6, 2018

Brucemck2 said:
My experience has been the opposite. For all it's faults, I've found USB sounds better than very high end compact disk players (the last time I did a head-to-head the then top of their line 800 series Meridian transport), and, the user interface is considerably better. YRMV

I believe that good USB can beat virtually any CD transport, but it takes a bit of effort and the right choices in USB interface, USB cable computer, server and software.

Steve N.
Empirical Audio

Empirical Audio · Jan 6, 2018

Berkeley Audio has a separate box to convert USB in AES/EBU - should we consider it "right" USB? I used it with Devialet and it sounded subjectively better than the Devialet USB implementation

This is not unexpected. The Berkeley is a galvanically isolated XMOS-based USB interface with separate power supply. My own XMOS-based USB interface is also galavanically isolated. This eliminates the common-mode noise that I talked about in my list above.

The XMOS interface will always sound better and the galvanically isolated interface will usually sound better than the typical USB interface. Not to mention the other design features like low-jitter clock and associated circuitry.

Steve N.
Empirical Audio

Empirical Audio · Jan 6, 2018

analogsa said:
An album sized buffer implies the interface can be disconnected during playtime. Neither incoming jitter, nor CM noise can then be an issue. Or OS/player. But there is the inconvenience of waiting for the buffer to fill up.

CM noise is still an issue. It's essentially ground-loop noise on a differential signal. The OS can change the data, which can be audible, both by DSP and changing the offset.

As for jitter using async usb, cannot see how incoming jitter can be an issue at all. The galvanic isolators with the ridiculous and dramatic cuts in the pcbs otoh... not so sure they isolate cm noise at all.

Internal jitter is the problem here, not jitter on USB.

Ethernet is probably a better kludge on the road to buffer play than usb but the issue of keeping the listening space clean from rfi remains. I routinely switch off routers/switches whenever LPs get played and the improvement is not subtle. Even though all power supplies are linear.

Buffer play of the type you describe is only useful if you are playing an album. If you are playing 1000 tracks, it will overrun or underrun.

The alternative is a clocking system that "brackets" the nominal frequency by clocking slightly higher and lower to keep the FIFO from over or underrunning. The problem with this solution is that the clock will never have the low jitter of a fixed free-running master clock. There are already devices like this out there and I have measured their jitter:

http://www.audiocircle.com/index.php?topic=154498.0

Steve N.
Empirical Audio

analogsa · Jan 6, 2018

Empirical Audio said:
The OS can change the data

The OS? You mean a naive user using the OS sound system? It has been trivial to achieve bit perfection for a long time...

As for the digital pll clocks, after a bad experience with a Si570 i have discarded the idea of an adjustable clock, at least for the time being.

Empirical Audio · Jan 6, 2018

analogsa said:
The OS? You mean a naive user using the OS sound system? It has been trivial to achieve bit perfection for a long time...

As for the digital pll clocks, after a bad experience with a Si570 i have discarded the idea of an adjustable clock, at least for the time being.

The one I measured is not a PLL clock, it is a bang-bang bracketing clock. I also designed a product like this, prior to my Synchro-Mesh reclocker.

Bit perfection has nothing to do with offset.

Steve N.
Empirical Audio

Pb Blimp · Jan 6, 2018

"USB done right" is ethernet direct to I2S. :b

Search

Search

What is "USB done right"? Does USB have a sonic signature?

caesar

Well-Known Member

Hifi Boy

New Member

Empirical Audio

Industry Expert

analogsa

Well-Known Member

Legolas

VIP/Donor

Brucemck2

Member Sponsor

microstrip

VIP/Donor

Empirical Audio

Industry Expert

analogsa

Well-Known Member

Empirical Audio

Industry Expert

Empirical Audio

Industry Expert

Empirical Audio

Industry Expert

analogsa

Well-Known Member

Empirical Audio

Industry Expert

Pb Blimp

Well-Known Member

Similar threads

Members online