This is a picture of the corner of the room where I performed the measurements for the post last week. Notice that Intel i7-3770K computer in the left corner I use for gaming? Within that box are 16GB DDR3 RAM, both a SanDisk Ultra II SSD and a recent 2TB Firecuda drive, plus a rather powerful nVidia GTX 1080 graphics processor which is what is driving the display on the 4K TV. Inside the case, it's all powered by a 6-year-old Antec 650W switching computer power supply these days. The computer is about 6 feet away from the TEAC UD-501 DAC.
Suppose I take a 16' USB cable and connected this computer to the DAC and compared the measurements with the low-power Raspberry Pi 3 B+ streamer... What do you think the result would be from the perspective of distortion and noise?
Before we start with results, here's the testing setup. Very similar to what I did last week.
Computer / Streamer (i7 computer or Raspberry Pi 3 B+ with JustBoom Digi HAT) → 16' generic USB cable or 12' generic TosLink / coaxial → DAC (TEAC UD-501) → generic 6' XLR balanced cable or 6' shielded RCA for unbalanced → RME ADI-2 Pro FS ADC → generic USB → Windows 10 measurement laptopI don't think it'd be a surprise to anyone if I said that based on conventional "audiophile logic", the i7-3770K computer "should" be a terrible music player/streamer. Using plain ol' Foobar, Windows 10 ASIO driver for bit-perfect playback is nothing special. The computer's power supply should be rather poor for high-fi audio, and that GTX 1080 graphics card would make the noise even worse. Furthermore, shouldn't HDMI 2.0 be rather noisy, hooked up to a 75" 4K/60Hz TV sitting about 2 feet from the DAC?!
Let's see what the RightMark 24/96 results taken from the TEAC DAC's XLR output look like compared to the low-power Pi 3 with USB output or using the JustBoom Digi HAT...
|SPS = Switching Power Supply for the Pi, Bat = RavPower 22000mAh Li battery, Wi = WiFi, Eth = ethernet, Scr = touchscreen ON|
Notice particularly how the IMD+N sweeps dissociate based on the interface used. Whether from the Pi 3 or i7 computer, the USB interface resulted in lower distortion on the sweeps (as I mentioned last week, I suspect this is correlated to jitter with the TosLink > Coax > USB).
And if I overlay the i7 computer noise floor playing digital silence over the previous Pi 3 log tracings shown last week (the yellow tracing pointed to being the i7 computer)...
There's nothing bad at all! Within the audible spectrum, the i7 computer noise profile stayed below -140dB. No terrible noise spikes or unusual ultrasonic noise differentiating the i7 despite the fact that we're looking at a moderately high-powered, older CPU using a typical multi-rail PC switching power supply with a powerful GPU inside with no attention paid to software audio or OS optimizations.
Okay. What if we stress the computer's i7 processor, nVidia GPU and power supply with a heavy processing load? Surely there must be all kinds of electrical noise that gets generated and passed down across the USB cable to the DAC, right? The i7-3770K has a TDP of 77W and at full load, the GTX 1080 GPU sucks up to 180W!
For this next set of results, here's what my computer is doing in the background while we run the RightMark tests to the TEAC DAC:
As you can see with the Task Manager window, the CPU is running at 100% (Prime 95 64-bit "Mixed" Stress Test in the background), while I'm running the FurMark GPU Stress Test with the GTX 1080 graphics card sucking up "99.4% TDP" at 78°C after >20 minutes. Remember, this is all running at 4K/60Hz over HDMI 2.0 on an i7 computer while sending the audio digital data over to the TEAC DAC! That's about as extreme as one can imagine playing music on a computer... Obviously, no sane audiophile would be doing this while enjoying music. Considering how extreme this is, what does objective testing show?
Nothing to suggest more noise or distortion! Surprised? Of course we really should not be by this point after all the measurements I've shown and discussions we've had over the years since 2013. Once we control for pitch inaccuracies (highly unlikely with modern digital), temporal fluctuations (eg. jitter, typically well controlled with asynchronous USB), and the DAC manages to reject noise reasonably well, guess what... "Bits are bits" even with generic cabling :-).
And just to be really complete... Here's the i7 jitter profile under full load:
From my testing, jitter anomalies are primarily a function of the DAC and asynchronous USB interfaces have served us well, the J-Test looks great even from a 5-year-old DAC. 100% CPU and 100% GPU processing did not affect the J-Test when transmitted over asynchronous USB. I have listened to the system and it sounds just great. The only thing is that after a few minutes the CPU and GPU fans will rev up which of course is intrusive for high-fidelity enjoyment, this is primarily why I prefer a fanless solution like the Raspberry Pi.
Notice however that so far I have been using the XLR balanced output in my measurements last week and so far with this post. While the DAC in balanced mode appears resilient to noise, what if we now switch over to the unbalanced RCA output? What kind of noise do we now see while using the i7 computer compared to the Raspberry Pi 3 B+ "Touch" streamer?
As we know, RCA unbalanced outputs are noisier than the balanced XLR results on good equipment. To the left we see 3 measurements with the i7 computer connected to the TEAC DAC by USB; we have the XLR results, followed by playback through RCA without increased CPU/GPU load, and then the RCA playback under 100% CPU & GPU load. On the right side, we have the cluster of results from the Raspberry Pi 3 B+ connected to the TEAC DAC through USB with a comparison of the XLR output compared to RCA either with switching power supply (ethernet + screen on) or lithium battery (WiFi + screen off).
Looking at the numbers, it's clear that XLR provides lower noise level, lower crosstalk, and consistently less distortion. This can be seen on the graphs:
There's a slight 60Hz hum along with a few noise peaks picked up using the RCA connection (highest component at 180Hz, -122.5dB). The IMD+N sweeps show a difference between the RCA and XLR outputs. Scouring through the graphs, the only other interesting thing here with the RCA output is the 2 low level peaks we see in the noise level that sticks out when the i7/nVidia computer is under high load as identified with the red arrows. Otherwise, there's not much else to see!
So yes, we can say that a very busy and electrically noisy computer can add to anomalies in the RCA output when connected to the DAC by a USB cable... But let's keep in mind the magnitude we're talking about. To demonstrate the extremes, here's a direct comparison of the noise floor up to 192kHz between the battery-powered Raspberry Pi 3 and the i7/nVidia computer running at 100% load (WaveSpectra, 384kHz sample rate, 128k FFT points):
As per the red arrows, there are small amounts of noise at 8kHz (possibly exacerbation of USB PHY packet noise?), 16kHz, plus a little bit at 45-50kHz. I think the rational audiophile will look at this and just say "What's the big deal!?". Well dear audiophiles, as far as I can tell, despite all the hype made about the need for expensive and supposedly quiet power supplies, or benefits of computer optimizations, there's basically little difference between a low-powered Raspberry Pi 3 B+ and a power-hungry Intel i7 computer with high-power graphics card even with unbalanced RCA output in my system!
Remember, the 8kHz, 16kHz and 45-50kHz noise above from the i7 computer under full load is only found with the RCA output. Balanced XLR output was successful in keeping out the noise:
Summary:Well then... Let's wind this down and draw a few conclusions based on the data:
1. If you're an audiophile who wants low noise performance, go for a DAC with balanced XLR analogue output (and make sure the rest of your system is up to par).
2. While my experience has been that jitter is rather inaudible, if you want low jitter performance, asynchronous USB is generally better than S/PDIF (TosLink or coaxial). Newer DACs like the Oppo UDP-205 and RME ADI-2 Pro FS have excellent S/PDIF jitter rejection so even this difference is minimal. The better your DAC, the more "bits are bits".
3. I remain skeptical of claims that digital streamer devices can affect sound quality significantly unless they are actually changing the data sent to the DAC. As per the tests today, we see that even a computer that's not "optimized" for audio with ostensibly "noisy" hardware does not actually deteriorate measured playback quality nor actually worsen the noise floor much at all even when put under a very intense processing load. Furthermore, this was tested with the 5-year old TEAC UD-501 DAC rather than newer and potentially better devices.
4. This continues to add to the evidence that "audiophile" computer-based digital "transports" do not make a significant difference to the sound. I cannot subjectively hear a difference either when I play music as I switched around the configurations while measuring. What's most important remains the quality of the DAC itself. In my opinion, other than the user interface, supported features or the "non-utilitarian" benefits like the esthetics and the "wow factor" of an expensive system, "high end" devices like the Melco previously discussed will make no appreciable difference to the sound quality itself compared to an inexpensive Raspberry Pi connected to the same "good" DAC. As usual, I would like to see evidence to the contrary and challenge the manufacturers of such products (Aurender, Auralic, SOtM, Sonore, Fidelizer, Baetis, Antipodes, Innuos, etc.) to demonstrate objective benefits from the DAC output using their products (please also identify the DAC used and under what conditions).
5. Having said the above, remember that there are indeed subtle differences we can see on objective testing but I certainly would not consider these significant in my system. Be reasonable with your set-up and there should not be any concerns. Obviously, do not do things like put your DAC in close proximity to a powerful computer as the DAC and any poorly shielded cables could pick up electrical interference (one example is the old measurement using my ASUS Xonar Essence One back in the day sitting close to the computer). Remember the importance of low ambient noise in your sound room when enjoying music among other things like room treatments and decent reverb time. EQ'ing and digital room correction make huge differences as well. Also, especially if you have a complex component audio system, make sure to address ground loops if you hear hum.
Well, the Christmas audiophile sales drive is in force now with "Best Of" lists, "Gift Guides", music recommendations, and advertisements dessed up as reviews on-line :-).
I noticed recently that the idea of audible noise and jitter from different USB cables is alive and well in late 2018... Here's Darko's take on the benefits of the Curious USB cable expressed as "the double trouble of electrical noise and jitter that you dropped US$350 on a USB cable...". Personally I would not drop more than US$50 on a USB cable, but I'm cheap in this way :-). Obviously, I'd love to see some evidence in light of the post today for either "troubles". For the record, the 16' generic USB cable I used in the tests today was the US$6.00 AmazonBasics USB 2.0! Remember also my USB cable measurements back in the day.
I noticed that at least he made some comparisons and claims that the Curious USB cable beats out the AudioQuest Carbon. Of course this is a "safe" conclusion and rather nondisruptive to the Audiophile Industry Nature of Things which seeks to correlate sound quality with MSRP because the Curious cable is about twice the price of that AQ Carbon. Considering the findings in this post, what then does it mean if indeed audiophiles like Darko are able to hear a difference between USB cables because of "noise" and "jitter"? Does it mean that they can hear below the noise level of the RME ADI-2 Pro FS ADC!? Is the equipment he uses so noisy and jittery compared to even an i7 computer and a 5-year old DAC such that I would have needed those cables in order to detect the benefits? If indeed the AudioQuest cable sounded worse than the Curious cable, does that mean there's something defective with the AudioQuest because even a 16' $6 AmazonBasics USB cable did not show any unusual noise nor jitter!? Is he even sure that noise and jitter has anything to do with the reported sonic differences? May I humbly suggest that last explanation to be the most likely...
Other than here, perhaps Audio Science Review, and maybe other objective forums, when was the last time you saw a magazine or mainstream audiophile web post show overlays of the noise level using different cables? When did you last see a direct comparison of jitter being different using various cables connecting the same pieces of gear? I certainly do not recall anything in the last few years even though a magazine like Stereophile routinely shows much more sophisticated measurements. Why are such basic claims never investigated? So many questions about the way things are done in subjective audiophilia land yet so few answers.
I don't think it's any surprise that manufacturers (especially of snake oil) prefer objective measurements be avoided yet IMO clearly this information can be very useful for consumers. I don't think it's any secret that things like cables have fantastic markups to the price and are potentially very profitable products. The fact that mainstream audiophile magazines do not do objective comparisons and "investigative reporting" IMO is simply a reflection of the influence of Industry sponsorship. At least to some extent, the "press" functions as the advertising arm of the Industry and certain rather straightforward truths must be kept hidden so that myths like "Expensive USB cables make a huge difference!" remain within the realm of the possible to drive sales.
Look folks, I'm not insisting that people should not / must not buy audiophile USB cables; enjoy the freedom to do so! Gift them as stocking stuffers to your audiophile loved ones. Savour the unwrapping of a beautiful set of cables on Christmas morning! But just remember to take a few moments to compare with a decent run of generic USB 2.0 and honestly consider whether you hear a difference. Remember, it's OK not to hear a difference.
As I have said before, we don't always have to buy expensive audiophile toys because they "sound better". There's nothing wrong with a nice looking set of luxury cables, just like there's nothing wrong with wearing a nice suit, Italian leather shoes, fancy cufflinks, or a meticulously designed wristwatch. Nothing wrong with being a proud owner of a $5000 set of cables! Be proud if that's how they make you feel! Of course, remember that if you don't feel there's good value in those cables, it's great these days as customers that return policies are quite good and one is free to exercise that option as well. One could always use the money towards good music in 2019 and I'm sure the artists and record industry would appreciate that as well.
Have a good week everyone. Hope you're enjoying some holiday cheer especially accompanied with great music. Gonna take some R&R time soon :-).