|ODROID-C2 & unused "Fujifilm" 5V 1.0A wallwart ready for testing...|
Before I begin, let's discuss what we're trying to figure out objectively here... It has been claimed by some that the USB interface (which the ODROID will be streaming through to the DAC) can be "bad". Supposedly, it's electrically "noisy" (of course a noisy fan and spinning hard drives are also very bad). And, this is especially "bad" with computer audio because computers are electrically noisy devices and will end up polluting the output from your DAC. Hence, by this reckoning, a general-purpose computer isn't supposedly a good thing to have doing high-fidelity audio duties. Furthermore, it has been surmised by some that power supplies can be problematic. Specifically, an inexpensive switch-mode power supply (like the one in the picture above) can be noisy and again, will have deleterious effects on the sound quality.
It has been said that devices like the recently released Sonore microRendu among more fancy audio streamers like the Auralic Aries line of devices can "sound" better because they reportedly take into account the various sources of noise and supposedly have optimized hardware/software. True or not, the price of such "audiophile" grade devices can be rather substantial especially when it's a device without internal storage and acts as a digital conduit to feed a DAC.
So, what then should we evaluate objectively for an audio streaming device? If you look at Stereophile's measurements for the Aurender N10, evidently not a heck of a lot :-). So long as it's reliable, does the streaming job without undue hassle, and for me sonically unobtrusive (ie. no noisy fans disrupting the playback), then the likelihood is that there's not much that can go wrong unless one can find noise anomalies as per the above perfectionist-audiophile claims.
Here's then what I am going to do... Let's put this little ODROID-C2 + Volumio streamer through the test bench with my USB TEAC UD-501 DAC. Over the years, I have measured the sound output of this excellent DAC with all kinds of computers. Can I find any difference between machines? Let's also include the usual Dunn J-Test to look for evidence of jitter/time-domain issues. And finally, let's check and see if I can detect the 8kHz PHY microframe packet noise as I did last year due to sensitivity of the Emotiva XSP-1 preamp's RCA analogue bypass input which was able to pick up noise in my more complex home-theater setup.
Part I: Differences between computers compared to ODROID-C2 driving the TEAC UD-501 DAC (USB)Here's the setup for these measurements:
"Streaming Device" [Computer / ODROID-C2] --> generic shielded USB --> TEAC UD-501 DAC --> shielded 6' RCA --> E-MU 0404USB ADC --> USB cable --> Windows 8.1 computerAll upsampling and DSP in JRiver 21 turned off of course for the testing (I have already shown previously the frequency response when the convolution DSP with JRiver is activated). Volumio 2 using its underlying Linux mpd and ALSA infrastructure streaming through gigabit ethernet. All Windows 10 machines are using the latest TEAC ASIO driver (v108e from late 2015). For some (like the Surface Pro 3 and Pentium G3220 computers), the test signal was on the SSD in the computer so not streamed through the network. The Gigabyte HTPC is playing off my Windows Server "NAS" through Samba over ethernet. And of course JRiver sending through DLNA/UPnP for the ODROID-Volumio 2.
Let's dispense with 16-bit audio, doubtful that "standard resolution" signals will yield anything when we can examine higher resolutions for anomalies... So starting with 24/96 RightMark test signal, this is what we see as a numerical summary:
To the left is the ODROID feeding the TEAC UD-501. Then next is the new Skylake Gigabyte Z170X-Gaming 7 motherboard HTPC. This is followed by my Microsoft Surface Pro 3 running battery powered feeding the UD-501. The last two columns were measurements done months ago with my previous Pentium G3220-based HTPC computer with ASUS B85M motherboard - playing 24/96 straight from Windows 10 or Windows 10 plus Fidelizer and JPLAY software optimizations as described previously.
Numerically, despite some expected inter-test variability and the fact that these measurements were done over months, there is no significant difference... At most <1dB difference here and there with key measures like the noise floor way down below -110dB! Notice too the lack of any difference with the measurements of distortion. As demonstrated in the past, I have never been able to show that software optimization like Fidelizer and JPLAY have ever changed the output of a good external DAC.
A few choice graphs to examine for those inclined:
Okay... Let's go to 24/192 resolution:
And more choice graphs for your consideration:
Again... As you can see, there is no difference to be found whatsoever with the different computers I have feeding the TEAC UD-501 DAC. Whether it's a latest generation Skylake i5-6500 computer, or a much slower Pentium G3220, or a battery powered Surface Pro 3 with i5 processor, or the little low-power ODROID-C2 ARM-based processor. The sonic output is a reflection of the quality of the DAC itself, NOT the streaming device.
Part II: JitterAlright, how about using the J-Test measurements for hints of time anomaly?
Anyone see a difference they think may be audible between the ODROID, Surface, and Gigabyte motherboard computers? :-)
As demonstrated time and again, the asynchronous USB interface dissociates whatever timing irregularity might be in the USB stream from the clocking of the DAC playback. The effect from a streaming device is minimal unless you're using an interface like S/PDIF which has the clock signal embedded... To demonstrate a gross example, we can look at the 24-bit J-Test accelerated to 96kHz from 48kHz comparing the asynchronous USB conversion from the ODROID and Surface Pro 3 with what I showed previously with the Google Chromecast Audio and TosLink S/PDIF interface:
|Note the X-axis was a little different for the Chromecast Audio. The ODROID-C2 and Surface 3 plots are essentially the same...|
I've seen recent claims that USB --> S/PDIF converters are somehow beneficial to good sound. I suppose this only makes sense if one absolutely needs galvanic isolation with optical TosLink and willing to pay the price of potentially higher jitter. Remember that I looked into one of these inexpensive devices (CM6631A-based asynchronous USB converter) back in 2013 and found that it did work well. But I would not advocate adding yet another device in the digital chain unless it's clearly necessary.
Part III: SMPS's (Switched-Mode Power Supplies) Are Horrible! (So says some perfectionist-audiophiles...)I know... We all know that linear power supplies are "much" better, right? Well, alas, I don't have a linear power supply handy (though I was thinking of getting a lab unit for experiments). However, because the ODROID-C2 can be powered by the micro-USB cable, why not just use a simple 5V Li-ion battery?
As you can see, the little 2200mAh Duracell (5V, max 1A output, presumably typical LiCoO2 chemistry) I got from Costco as a bulk pack is capable of running the ODROID with no problem (for maybe a couple hours, didn't bother timing exactly)! Let's have a look at what the measurements look like with this:
As you can see, both the 24/96 and 24/192 measurements are here ("ODROID-Battery" obviously indicating the measurement with the Duracell battery used to power the machine). Essentially no difference here folks! Here are the 24/96 graphs...
As you can see, using a Li-ion battery with the ODROID as streamer made no difference to the TEAC UD-501 output... If in fact there was a difference, then it's below the noise floor of the E-MU 0404USB which means it's below -110dB with this DAC and I can't imagine how this could make a true audible difference! Clearly, we see the 60Hz powerline hum in these measurements. The DAC output as one would expect is logically more affected by the power supply inside the DAC rather than whatever power supply is feeding the digital streaming device upstream.
I had a listen to a couple tracks off Andrea Bocelli's recent album Cinema (2015) with the ODROID powered by the Li-ion battery versus the SMPS wallwart. No discernible difference as far as I can tell... Of course, there was a delay in shutting down the machine and rebooting so it's not a quick A/B comparison. Even if there were a difference, it wasn't obviously better that I felt compelled to run the device in a battery operated mode (or to grab a linear power supply and try given what I see here).
Part IV: Search for 8kHz PHY Microframe NoiseRemember back last year, I showed that there was an 8kHz noise I could hear out of my preamp's analogue "home-theater bypass" input when the HTPC USB was plugged into the DAC? Recall from that post that I was able to attenuate it with the Corning USB3.0 Optical Cable to a significant degree.
Let's see if I can measure it with the ODROID-C2:
As you can see, taking the analogue output from the Emotiva XSP-1 preamp set to the "home theater bypass" analogue input with the TEAC DAC plugged into the Skylake HTPC directly, we can pick up the 8kHz microframe noise at a low level (on the top left we see the loudest tone is measured at -102dB at 7998kHz). In comparison, the ODROID-C2 was nice and quiet with the "loudest" level being a 60Hz powerline hum at -105dB.
That's encouraging... I suppose it says something about the low-power ODROID not emitting as much electrical interference. Remember though as I warned previously, I feel this is a rather idiosyncratic finding likely related to the complexity of the home theater system and the sensitivity of this "home-theater bypass" analogue input on the pre-amp. It's a way to show myself that the 8kHz noise can be amplified in certain circumstances and probably has to do with how I'm hooking up the HTPC to my system (just like ground loop hum from the computer was an issue and needed to be addressed in that previous post).
Part V: ConclusionI hope this post answers a few questions you might have been curious about and provides answers about the role of digital streaming devices on sound quality when used as a USB source. In general I can state the following:
1. Yes folks... "Bits are bits" when dealing with accurate digital data transfer to an asynchronous DAC. Some people don't like to believe this but it is consistent both in my personal subjective observations and consistent with the objective results. Indeed, the universe seems to "work" as expected within the domain of engineered electrical devices meant to reproduce audio frequencies :-). Measurements like frequency response and distortion characteristics do not change based on what computer or streaming device is used to feed the DAC asynchronously. These qualities are inherent to the DAC itself. Also, the sound of the DAC of course does not change based on the streaming device since they're all feeding the same (bit-perfect) data.
2. Time-domain jitter is not an issue with asynchronous DAC's as I have shown many times before irrespective of the source. I hope the jitter "boogey man" can be laid to rest when it comes to most reputable modern asynchronous DACs. And even when there are differences in jitter, audibility is unlikely unless very severe.
3. While a switched-mode power supply (SMPS) might be noisier than a linear one (high frequency switching noise for example), I fail to see how going with a linear power supply will make any difference when used to power the digital streaming device unless there is evidence that this noise affects the digital interface and the DAC is further susceptible. I don't have any qualms with advocating for a good power supply in the DAC of course (for example, it would be nice to not have that 60Hz hum in the output of my TEAC DAC). In any event, I found no difference going from the "freebie" SMPS I have been using for the ODROID-C2 compared to a Li-ion battery on the DAC analogue output...
Needless to say, I am rather impressed by the little ODROID-C2 computer for audio streaming. It's very easy to set up, for weeks it has worked without an issue in the background sucking up little power and always on the ready to play whatever is requested through JRiver 21. I have not experienced any instability while running Volumio. The quad-core 64-bit processor is speedy for audio-related tasks. And like I showed last week, the gigabit ethernet is really quite capable! By the way, Volumio operates in a "headless" fashion so while it's running, the HDMI video output is turned off for those who are concerned that somehow this may affect sound quality. Remember that although I'm testing Volumio here, Archphile already has an image out and a RuneAudio release is imminent. Always good to have options!
For significantly less than $100, I think one can't really go wrong with this little unit or the Raspberry Pi 3 or numerous others. The benefit of the ODROID-C2 is its speed, 2GB RAM, and the potential to be a good audio/visual streamer in the future if you want 4K/60 HDMI 2.0 plus excellent gigabit ethernet speed.
Looking around at the audiophile offerings, I'm scratching my head at the mark-up and seriously questionable claims out there. For example, what's with the redacted texts in the Sonore microRendu "review"? Is that not just a bit pretentious considering that we're talking about a rather pedestrian dual core i.MX6 Cortex-A9 machine!? BTW, anyone see the clockspeed of the processor listed anywhere or how much memory? And can we also dispense with the name-dropping and cult of personalities as if there is some kind of "wizardry" at play? It's one thing to romance sentimental old technology like vinyl, turntables and glowing vacuum tubes, but can we not see how ridiculous this all sounds with contemporary ARM-core computers running off Linux and using mostly open-source code? Do we have to euphemistically justify the use of a slow SD card interface (the SDHC slot has been there since the dawn of the Raspberry Pi)? If it doesn't seem silly already, look back on articles like this in a year's time. The magical thinking which has been applied to traditional "High End" audio really does not work when applied to computing appliances like this. IMO, simple digital streaming devices like this should be inexpensive since the technology advances quickly and there's just no "luxury" value to the hardware. I bet just reading articles like the microRendu "review" will turn off many technologically savvy audiophiles as being yet another example of how out of touch this hobby can be at times.
As usual, it would be great if manufacturers of some of these streaming devices could publish some evidence of claims. How does that expensive linear power supply plugged into the streamer affect a good quality DAC's sonic output? Where is the noise/EMI anomaly we should be concerned about in the USB output? In sum, where are the benefits to be found based on the eventual analogue output from the DAC to justify the expense considering that claims may be biased by financial incentives? I think this is a very fair question. Of course I would not be holding my breath in anticipation of any actual objective results beyond likely wordy apologetic testimony of supposedly "complicated" details from the manufacturers and faithful practitioners.
Okay, got a few "real work" projects on the go for the next couple of weeks. Also have a few new albums to check out plus playing catch-up on the latest The Walking Dead and Game of Thrones episodes :-). Hope you're all having a great time with the music!