These days, we all use USB cables of different lengths and varieties for high-speed devices like hard drives and generally either it works or it doesn't. USB protocol sends data in packets which consists of not just the audio data itself (up to 1024 bytes at a time for high speed mode), but CRC check to detect errors, flow control, and also address information to direct the data to the appropriate device (remember, you can attach up to 127 devices to each host controller). The low-level details of this communication including timing are addressed by the hardware and generally outside of the purview of the software we install on the computer. Logically this means that fine timing issues (like data flow control and scheduling of the data packet delivery) would be outside of the effect of things like audio player software. Furthermore, with modern DACs, the machine will use its own internal memory buffer and very fine timing (like pico- and nanosecond jitter) will be derived from the accuracy of this internal clock.
USB data transfer rate for audio is much less demanding than something like a hard drive. Whereas we can easily transfer >200Mbit/sec to the hard drive using a simple generic USB2 cable, sending standard 16/44 stereo audio to a DAC is only ~1.5Mb/s, 24/96 only needs ~4.6Mb/s, and 24/192 ~9.2Mb/s. If we go all out with the TEAC UD-501 DAC with DSD128 or 32/384, even that "only" takes up to ~25Mb/sec.
Not unexpectedly, some message forums got a bit heated when the USB cable test post was published since obviously the results deviated from expectations or experiences of subjective evaluation or maybe the person is trying to sell something related. I have seen no good evidence from controlled tests demonstrating otherwise.
Shortly after that post, this set of measurements came out from one of the principles of Empirical Audio (I thought he promised to measure the AudioQuest Diamond USB cable?). Nice to see he measured the down-to-earth Belkin Gold anyhow. I will let you, fine readers, examine the data and determine if you think this correlates with audio quality, or just demonstrating minute differences between lengths of copper with little correlation to digital data transfer much less the analogue output from a DAC! (Note that he's measuring 5m [~16.5 feet] lengths of USB cable to get those picosecond numbers.)
For this installment of MUSINGS, I wanted to have a look at a recent article from the July 2013 issue of Hi-Fi News where they did a "Group Test" looking at USB cables. Since I do not subscribe to this magazine, I want to thank "Mushroom_3" and "Julf" off the Slim Devices/Squeezebox forum for bringing up this article.
As you can see on the cover - "USB cable sound - Fact or fantasy? p43" was advertised. So from page 43-51 (with 2 pages of ads), they tested 10 cables.
Here's the first 2 paragraphs of the article (as usual, I reproduce only small portions out of fair use for the purpose of discussion, education, and critique):
That's a pretty gross generalization and assumption to start off with isn't it? The only thing I think "every seasoned audiophile" knows is that without cables, there can be no sound - and that's assuming your system isn't a boombox :-). That's about the extent of discussion as to whether USB cable sound could be "fantasy" (I think the better word would be 'inaudible'). From there on, it's all about oscilloscope measurements (they give data risetime measurements for the cables and use this and the eye pattern as their objective data), and minimal detail was provided as to how they conducted the "blind" subjective test - we do not know how many subjects participated, how they were selected, in what way was the 'blinding' accomplished, duration between cable swaps, etc. Furthermore, we do not know how the participants scored the cables or what the statistical outcome is - it may be just a "blind" solicitation of subjective opinions on those 10 cables (a huge number to test properly!).
I appreciate that at least Paul Miller the editor spent a page talking about the technical bits of USB cabling on page 98. At least the physical cabling characteristics were described in detail but strangely nothing on the packetized nature or how they would correlate audible differences in such a system of data transfer since IMO this would be much more educational.
The comments below will hopefully make sense to those who have not read the article.
1. From what they wrote, the test setup went like this:
Laptop --> Test USB Cable --> Musical Fidelity M1 S-DAC --> coaxial cable --> Devialet D-Premier --> speakers
Why was the decision made to use the Musical Fidelity M1 S-DAC as a USB-to-S/PDIF converter? Seriously, you're testing USB cables but introducing a coaxial S/PDIF interface into the middle of the signal path for no clear reason when the M1 is already a fine DAC?! We know that a good USB asynchronous interface has lower jitter in general than most S/PDIF, so why not use the analogue output of the M1 directly to a preamp/amp? I would not be surprised if the Musical Fidelity is a better DAC technically than the Devialet (see below). There was no mention of what coaxial cable was used - surely this must be important since this is the digital cable directly connected to the DAC and these guys believe S/PDIF cables make a difference as per the first paragraph quoted above. How is it possible to test USB digital cables if there's potentially an even more jittery (yes, the dreaded jitter bogeyman yet again) digital coaxial cable/interface in the audio chain?
2. Why was the Devialet D-Premier used? I suspect why they used the USB-to-S/PDIF interface is because the Devialet internally resamples analogue to digital as per the discussion in the Stereophile review, plus it doesn't have a USB interface. Furthermore, the measurements for the D-Premier isn't all that remarkable as a DAC. It achieves a respectable 18-bit dynamic range (remember, the SB Touch already can do 17.5 bits) but like John Atkinson say - it's "not quite up to the standard set by the best-measuring standalone processors, such as the Bricasti M1, MSB Diamond DAC IV, NAD M51, or Weiss DAC202". I don't know about the M1 S-DAC specifically, but the Musical Fidelity M1 appears to outperform the D-Premier already (Stereophile data) and "offers performance that is close to the state of the art."
3. Why was the "grotty giveaway USB cable" measured on p.98 not part of the blind listening test? I wish they included a picture of this "grotty giveaway" - what makes it "grotty" anyways? Why not also tell us what that cable's data risetime is while you're at it? Doesn't this often seem like the case with these audiophile "shootouts"? They have all these high priced options (well, at least they included a £18 QED Performance Graphite) but neglect to include the real competition - something that is essentially free and works! Another idea - how about testing a reasonable quality "non-audiophile" USB cable like the Belkin Gold for <£5?
4. USB data risetime & eye pattern: is it even relevant to the analogue waveform coming out of the DAC? Are they measuring something which actually COULD have sonic impact? In terms of this article with the risetime derived from the eye pattern as the prime objective measure, what evidence do we have that this actually correlates with the subjective impression? So what if the £18 QED scores 12.8ns, £60 Kimber B BUS Cu scored 12.4ns, £139 Audioquest Carbon scored 11.9ns (hey folks where's the Diamond?), and the insanely priced £6,500 Crystal Absolute Dream scored 11.0ns? What does this risetime have to do with the final DAC analogue output anyways given the nature of packet digital transmission, asynchronous protocol, and a cable that can provide much more bandwidth than required for audio - especially since you're passing the bits off through a digital S/PDIF coaxial in the audio chain?! I question if taking "a good few months" (p.98) to develop this test was time well spent!
5. Why does there appear to be so little correlation between sample rate and audio quality through the USB cables? Surely, a poor USB cable with "slow" risetime should sound worse with 24/96 or higher bitrate music, right? Yet the majority of the subjective complaints focused on 'Hotel California' off Hell Freezes Over CD or Oscar Peterson's We Get Requests FIM K2 CD. Heck, even The Beatles' Abbey Road 24/44 barely taxes the USB interface. The only really "hi-res" track was the 24/192 Helge Lien Trio Natsukashi which was mentioned 3 times total in the whole subjective write-up. Surely, to get a sense of how well USB cables work, we need to grab some DXD and DSD128 material, right? If timing/jitter were that important, there could be big issues with DSD sampled in the MHz range, don't you think?*
6. In the subjective analysis conclusion, you see the following comments: the QED is described as "almost sounds louder", Transparent Performance's cable had "rich and warm sound proving a little too luxuriant at times", Wireworld's Starlight 7 had the "woomph of air that would normally accompany the deepest bass was subjectively filtered away", or the Crystal Cable's "greater extension" - why not measure those things? Surely you can easily detect amplitude changes, differences to frequency extension, tonal changes to add "warmth" in the DAC output, right? Yet, for the objective test all we get is a risetime measurement of a piece of wire and the subjective testing done through a second digital interface of unclear quality and unclear blind testing protocol. Therefore neither the subjective or objective results appear all that convincing.
Assuming we do acknowledge this article as having validity and not just an example of a majorly flawed study (needless to say, the setup using a digital coaxial cable is a BIG problem IMO), there is one thing we learned in this piece for sure that could be of value and a good reminder. Don't buy the £70 Vertere Pulse D-Fi USB cable - apparently it's constructed out of "twin coaxial cables" (err, impedance matching anyone?), has bad rise time (27.6ns), and sounds questionable with "softening of its extreme treble and loss of atmosphere". I wonder if connecting this cable to a hard drive might show slower transfer speed due to high data error rate. Evidence that you can spend more and get significantly less quality than the 'freebie'?
Assuming we don't believe the listening tests are valid because of the methodological issues, then perhaps this article serves as an example of the unreliability of subjective listening. Isn't it possible they're all just listening to the same sound with the "jittery" coaxial digital cable and interface and coming up with different impressions? Since we have no concrete idea of how the "blind test" was conducted (especially what level of statistical confidence we can expect), it's quite possible the term 'fantasy' would be appropriate to describe the results (but verboten).
Enjoy the music folks... IMO, there is still no good evidence that USB cables make a difference to sound (well, at least with decent modern gear!).
My musical selection tonight: C.C. Colletti Bring It On Home (HDTracks 24/96 Binaural recording). Wonderfully spatial sonics and details from the ASUS Essence One + Sennheiser HD800! Have a listen to that title track.
* For the record, I do not believe there is any reason to think DSD would be affected by USB cable jitter to an audible degree with good modern DACs. Just wanted to throw out a thought which the jitter-fearing-audiophile might bring up one day.
0 comments:
Post a Comment