Thursday, 11 August 2016

MUSINGS: Do "audiophile" computer-based players make a difference? On "Everything Matters" and wisdom.


Since publishing the results from the NUC measurements a week back, permit me to discuss the topic of "audiophile" music servers, streamers and transports - specifically addressing the idea some have that they make a significant sonic difference.

As audiophiles, we've all seen them reviewed in the various websites. Names like Aurender, Antipodes, SOtM, Small Green Computer, Baetis Audio for machines that run either Windows or Linux, or more customized non-PC looking devices like the Auralic streamers, SONORE microRendu (Wow! So much hype!), and established audio companies within the arena like PS Audio, Ayre, Bryston, Simaudio, etc. Even JRiver sells their customized NUC server/player, called the Id.

Although the hardware features may vary significantly (with potentially extra cost associated): Does it have an integrated CD/DVD ripper? What digital interface is available (USB, ethernet, S/PDIF)? Is there an integrated DAC?

And the software set-up may be quite different: Windows? Linux? MPD? JRiver? Squeezebox? Roon endpoint?

The fact remains that these devices are basically microcomputers at heart in typically small cases and ultimately operate in the digital domain to send data to the DAC for playback.

On Computer Transports and Sound Quality

Unless programmed to manipulate the digital data, if we're talking about "bitperfect" transmission of digital audio to an outboard DAC, there is no reason to think there is any unusual or unique "sound" to these devices as demonstrated by the NUC vs. ODROID-C2 vs. laptop vs. HTPC measurements. Asynchronous interfaces (like USB and ethernet as previously examined) have already dealt with jitter for years such that nobody to this point has demonstrated audibility differences in controlled tests, and therefore in my opinion the only thing left is the issue of electrical noise which is the bugaboo being touted as the rationale for all kinds of digital audio "problems" these days.

In "dealing" with this "issue" of electrical noise, all kinds of tweaks have furthermore been offered for sale to prospective buyers (or at least those audiophiles with enough faith to accept the "diagnosis" and "cure"). Gear like the Audioquest Jitterbug, UpTone Audio Regen, iFi iPurifier, optical ethernet isolators have been suggested (like the TP-Link MC200CM), as well as the upcoming PS Audio LANRover (reminds me of my tests using the inexpensive USB ethernet extender) are all being touted as beneficial when hanging off your USB port!

After all these years of listening and testing encompassing laptops from Apple and those running Windows, to Microsoft Surface Pro, to a recent MacBook, to Linux OS, to low-power ARM machines like the ODROID-C2, to motherboards like the Gigabyte or ASUS, to software OS optimizations and "audiophile software music players" like Fidelizer and JPlay, and the Intel NUC recently, the conclusion is rather obvious to me. Bitperfect playback from a reasonable quality computer, using an asynchronous interface, and through a reputable DAC would result in the same sonic output irrespective of claims I've heard otherwise. I have found no need for special power supplies, fancy cables, or specialty devices for example to clean up the USB signal. In my mind "reasonable" just means a device that's known to be reliable and has a good reputation, rather than something that needs be endorsed by an audiophile guru. This is the most logical position to take intellectually given how digital devices work, based on objective results I've found, and subjectively I have no cause to testify otherwise. To put it bluntly... Yes, "bits are bits" using modern digital computer playback hooked up to a good asynchronous DAC!

This doesn't mean I don't have preferences for the soundroom playback device of course. Some devices might operate faster. I might prefer a certain tablet control app. Acoustically silent fanless devices are great. I might like a certain appearance or size (sure, heavy devices created out of beautifully machined metal can look great and instill confidence). I might want a multitude of digital in/out options. I might prefer DLNA +/- Roon +/- Squeezebox playback. I might want the ability to hook up a local USB drive. These are just some of the potential features and subjective values I might place on the device and if important, I wouldn't mind spending extra money on these qualities. However these features generally would not affect the ultimate utilitarian function - "high fidelity" audio playback.

Time and again, tests show that a good DAC like the TEAC UD-501, or Light Harmonic Geek Out V2 can resolved >16-bits of dynamic range (>96dB), with no evidence of added distortion, nor change to frequency response with any of these computers serving the data. This means that in a properly functioning system, you really can't say that one bitperfect device "brought out the midrange", "controlled bass better", "lowered the noise floor", "had even lower distortion", "changed the soundstage" (ie. interchannel effect and stereo illusion somehow changed!) or "made the dynamics so much more powerful and exciting" than any other without coming to terms with the cognitive dissonance these comments would imply! If one were to truly hear difference such as these, the likelihood is that one did not set things up properly and should consider what went wrong. In fact, none of the asynchronous DACs I've tested at length (TEACLight Harmonic GO2AudioEngine D3Focusrite ForteCreative E-MU 0404USBASUS Essence One) have shown significant variation in measured analogue output using different computers over the years.**

It's fascinating watching the audiophile market, the build-up of excitement around devices, the ebb and flow of products as they work their way through the audiophile mind-share and supply chain. Like the "flavour of the year" in 2015 with the UpTone Audio USB Regen, a device which supposedly improved USB transmission through a single-port hub, I'm amazed by all the hype around the microRendu recently. Folks, it's basically a low-power ARM SBC*** with Linux-based DLNA/Squeezebox/Roon playback. I think the most interesting part of the whole package is the software flexibility - that's great. But considering the asking price for the microRendu of US$640 (without power supply) for functions I can replicate with the ODROID-C2 for less than $100USD including power supply and some decent USB cables to boot, I see no reason to bother giving it a try. In fact, given how fast the SBC sector develops, one might even try the ~$10USD NanoPi Neo as reported on Computer Audiophile (no video capability like the microRendu) and be reasonably confident that you're not missing anything - kudos to Chris for posting information on the device! I think it's wiser to take the savings and go support one's favourite artists by buying more music.

By the way, I've seen the recent microRendu + iFi power supply measurements done by Amir. My suspicion since I don't have the device to test is that there was a ground loop in the measurement set-up given the very strong 60Hz hum using the iFi SMPS. Elimination of the ground loop would likely have shown no noise floor difference between the various conditions. This of course speaks to the importance of managing one's wiring and grounding which can be tricky in some situations and might need some ingenuity for where and how the components are plugged in (especially in a complex multichannel system in my experience when there are many amps, unbalanced RCA cables, etc...) This is why I prefer balanced XLR gear and cabling through my system when I can. You'll generally know if this is a problem because of audible humming through the system. In any event, I'm rather confident that in another year, the microRendu will be deprecated for much nicer looking boxes with claims of even better sound :-).

On "Everything Matters", Product vs. Process-based Solutions

To end off, let's just spend some time thinking about the old audiophile adage that "everything matters". On the most basic level, of course! Interconnect cables matter to transmit signal. Power cords matter to power your devices. And of course a well-functioning computer server/transport matters to reliably send data. But the problem presented to audiophiles whenever we come across yet another ad, forum post, or impressed reviewer claiming "better" sounding cables/digital transports/DACs is about the threshold of meaningful, audible differences. With some understanding of how the science works, objective data to consider, and some experience in listening to the various offerings, I do believe the wise hardware audiophile realizes that it's not simply "everything matters", but rather "everything can matter, and not to an infinite degree".

At the most basic level, we have to realize that there are both product-based solutions and process-based solutions to improving sound. It's easy for reviewers and forum posters to make recommendations and talk about product-based solutions; just buy a new, supposedly better product to replace your old one and you'll be happy! It is harder to consider process-based solutions like taking a good look at room construction, equipment placement, acoustic optimizations (eg. sound absorption and diffusion), or room correction (like EQ or DSP-based room correction). These solutions take time, study, and typically measurements to verify an improvement; and they're harder to recommend without knowledge of a particular soundroom and system. I would argue that by the time one has spent thousands on a nice sound system, it is going to be these process-based solutions that will get you further and develop the satisfaction in the "hardware audiophile" hobby. Instead of just taking out a credit card, you'll start engaging in a critical evaluation of the hurdles to overcome, start problem-solving issues, be creative with everything from speaker placement to furniture selection to tweaking with an EQ or calibrating a DSP target.

I truly hope that wisdom builds with the years of experience as we engage in this hobby (and in life in general my friends!). And this wisdom comes with it a certain degree of knowledge, sophistication (like being about to choose product-based vs. process-based ways of improving fidelity), confidence and even certainty in what is true; not easily swayed by every new model, hyped-up product, or unusual claim. Excitability and susceptibility to hype are perhaps valuable characteristics for advertising departments and successful salespersons paid on commission to move product; not the mindset of maturing hobbyists, serious journalists, or thoughtful reviewers.

Remember that stereo audio is very much a mature technology these days. Likewise, computer technology is well understood (it is after all engineered over many generations!) especially for comparatively low bitrate audio data transport. Claims of significantly improved quality whether it be software techniques like MQA, specialty cables, unusual room treatments (like these), or "fantastic" new digital transports really do need to be examined rationally in a sober, objective light. Contentious eye-brow raising claims about sound quality are often so because they suggest that the item runs contrary to accepted science. In situations like these, I think it's totally reasonable to be suspicious and request objective results from makers of these products as to how they propose whatever claimed effect can be audible. Barring the availability of such results as has been the case with hyped up products like the UpTone Regen (notice all the empty, impressionistic chatter by the designer) or microRendu, I do believe audiophiles should be asking questions and expecting real answers before signing up quickly into the hype.


** Note that there was one exception worth mentioning for completeness: the time that I purposely pushed the i7 and nVidia GPU to create a hostile environment with massive heat, acoustic and electrical noise. Sure, it's evidence that noise (EMI/RF) can seep through to affect the DAC, but this is quite atypical operation for audio playback. I'm thinking the only time this may be an issue would be while using HQPlayer and running all kinds of DSP processing while off-loading to an nVidia GPU... Even this should not be as extreme as what I tested! I'll look into this later perhaps.

*** For those curious about the innards of the microRendu, check out the pictures and comment from AbeCollins on AudioAsylum. A SolidRun System on Module (FreeScale Cortex A9) with a small custom carrier board.

-----------------

Greetings from Beijing!

A scene from Beihai (北海).
From here, I've got a couple more weeks in Southeast Asia... Then back to the comforts of home, and away from the heat and crowds back in "Beautiful British Columbia" :-).

Hope you're all having a great time and enjoying the music...


20 comments:

  1. Barring Room Correction or any other fancy processing I guess the $5/£4 Raspberry Pi Zero could be used as well. Or the former could be done on a central server.

    The thing with software solutions is they take a lot of tweaking and tinkering, and sometimes they foul up, or things go awry, and it they are complex, and support can be so, so, judging from threads on various forums, including Computer Audiophile.

    As for software audio processing, HQPlayer is intriguing partly off loading processing to the GPU. Surely Room Correction software tapping the GPUs must be the way forward, just based on the processing capabilities and speed of GPU development, generation to generation. A reasonable GPU today has the processing capability of 24,000 DEQX room correction pre amps. So software is the key.

    ReplyDelete
    Replies
    1. Thanks Gadget!
      Yes, without question, the parallel processing of GPU's would open up fantastic avenues! The only limitation I see is "for what?" when it comes to playback...

      Delete
    2. As in "not necessary" or "overkill" or "only useable for upsampling" or "only necessary for real time filters"? I guess faster processing, faster convolving, longer, more accurate filters, multiple channel correction, simultaneous multi room room correction? On the other hand, are you saying the current state of the art is sufficient, and the problem is solved, aka field is closed, so why bother?

      Delete
    3. Imagine using smaller, cheaper units, with strong GPUs and less powerful CPUs taking on room correction tasks.

      Delete
    4. I read somewhere Accourate's CPU filtering for phase correction creates audio delay of 1.5 seconds. With GPU that delay could be handled.

      I guess it could also be handled by software syncing, say through JRiver (of which I'm not a user), but I guess you know whether that is possible? I assume it's not, as it's been raised as an issue.

      Delete
    5. So is syncing of playback room corrected audio with live broadcast possible? Will both streams be delayed so it's a non issue?

      Going beyond playback, and go to play. Imagine someone playing Pianoteq with room correction, having the grand piano present instantaneously as the keys are played through a truly powerful full range system. That requires faster processing of the room correction. GPU tech is increasing speed by about 30% each year. Memory tech on the cards is expanding at break neck speed.

      At one time room correction wasn't possible at all. Or with floating point. It is now. Tech changes the scope of what's possible. I guess conventional wisdom relates differently to that.

      Delete
  2. Hi Archimago,

    I have followed your blog for long and I really appreciate your posts because of your pragmatic and ”non-audiophoolic” approach to the topic. One thing that I have missed there is that you haven’t ever mentioned Daphile (AFAIK). As Daphile is kind of my “child” I would really appreciate if you would give it a chance, try it and maybe even review it on your blog - doesn’t everyone want to have some attention for their kids ;-).

    In case you aren’t familiar with Daphile at all it’s a music listening focused and minimal sized Linux distro integrating LMS and squeezelite (plus some additional features). I have tried to make it as easy as possible to install, use and maintain even for non-technical person (even at the expense of flexibility and expandability). I don’t argue that it would sound any better than any other system capable of bit-perfect playback because my strong view (as M.Sc. of EE) is that as long as computer is able to provide continuous bitperfect audio stream to DAC using asynchronous transfer the computer software or hardware don’t have any effect on audio quality (assuming of course that DAC has reasonable protection against EMI/RFI). Purpose of Daphile is to provide the complete core (incl. OS) of the wonderful Squeezebox ecosystem to users as easy as possible.

    Please have a look more on Daphile web site (www.daphile.com) and I’ll be of course happy to answer any questions you might have.

    Cheers
    Kimmo

    ReplyDelete
    Replies
    1. Yes Daphile is easy to install and use...it's interesting to test the Play from RAM drive option and evaluate the sound difference during in playback.
      This features push the playlist in the RAM instead of playing back from the hard drive.

      Delete
    2. Thanks Kimmo, looks like a great distro and appreciate the work folks like yourself have put into these great options for the audiophiles out there!

      I'll look into this suggestion on my list of to-do's :-).

      Best regards...

      Delete
  3. Hey Arch,
    Would your comments be any different if HDMI was included in the discussion? (reference MCH playback on a Windows10 NUC/JRiver to HDMI-in on an Oppo 105)
    Enjoy your trip,
    sk16

    ReplyDelete
    Replies
    1. Thanks for the note SK.

      I think I can borrow my friend's 105 to have a look... Another on the to-do list :-)

      Delete
  4. In my experience I have found noticeable differences between various hardward boards (and I have tried many...). These differences do not get erased using the various USB gizmos (Regen, Jitterbug, Intona are the ones I have tried). I do not know why this is the case...

    I have recently compared the following boards (have tried others in a more distant past): Alix 3D2, Olimex Lime A20, NanoPi Neo, Intel dn2800mt, and a fanless i5 PC. Well, the Alix 3D2 and the Intel DN2800MT have significantly better sound from USB. The difference is immediately noticeable.

    Concerning the USB gizmos (regenerators, isolators, cables), I have found that any effect they may have is marginal, and the difference in SQ may just be attributed to testing conditions and bias (wanting to believe there is an improvement).

    Don't know how this confirms or not your findings.

    ReplyDelete
    Replies
    1. Hopkins. Great that you're experimenting with the different options!

      Wondering what DAC are you using? Basically I was unable to find difference both with my LH Geek Out V2 and my TEAC UD-501. I suspect the key here if there are findings to be discovered may be the DAC and if there truly are susceptibilities to difference between the machines...

      Delete
  5. Excellent article Archimago! Even the few audio tests I ran on comparing bit-perfect music players confirmed to me that as long as one is getting bit-perfect audio, what me worry – it’s as good as it gets and has been for the last decade or so. The pro audio guys figured it out back in 2007 that all pro Digital Audio Workstations (DAWs) sound/sum the same.

    Like you say, digital audio is a +40 year mature engineering discipline that has met its raison d'être of providing ubiquitous bit-perfect sound quality to everyone. In fact, there are several industry standards and tests for the transmission of digital audio that are extremely well understood, to the point that it can be reduced down to a simple pass/fail test called a Mask Test. A mask test defines the allowable shape of the transmitter waveform. The mask test is highly efficient since it quantifies both time and amplitude parameters in one “eye pattern” measurement. As long as one stays within the “mask” during digital transmission, with no bit errors, then the data has been transmitted “bit-prefect. An excellent educational tutorial can be read from Agilent’s application note on digital audio compliance testing.

    I hope folks take a bit of time to read the Agilent application tech note as one will see the real reason why product based digital audio solutions mentioned in your article don’t supply eye pattern tests is simply because they all pass bit-perfect audio – i.e. they pass the mask test. It is as simple as that. And that’s all that matters when it comes to the transmission of digital audio.

    As to the audibility of digital audio artifacts, whatever they may be, I tested how far away from bit-perfect audio I had to go before I noticed an audible change in sound quality. While I used rock music to compare, I had to reduce the bit-depth to about 12 bits (i.e. 72 dBFS) before I noticed a change in sound quality listening at reference program level (e.g. ~83 dB SPL C weighting at the listening position). Note that once below the threshold of audibility, it does no matter what the digital artefact is, because it is inaudible. Folks can try for themselves at Fun with Digital Audio – Bit-perfect Audibility Testing or if using JRiver, one can reduce the bitdepth by using the Bitdepth Simulator in DSP Studio under Parametric EQ. Highly educational and hopefully ear opening for some, as folks will realize that the measured difference in noise or jitter or pick your digital artefact poison between -120 dBFS and -100 dBFS, under normal listening conditions, is simply inaudible.

    So now what? As you say, if one has a reasonable amount of computing power to deliver bit-perfect sound, then what does matter? To me, its loudspeakers and rooms that have real world issues and make the most noticeable audible change when applying a process based solutions. Some of that is getting into product based solutions like the Kii THREE and BeoLab 90 loudspeakers using advanced DSP techniques to more accurately reproduce the music to our ears or shape the music reproduction to ones ears based on preference. To me, this is the future advancement of digital audio and what matters most. Hardware based solutions and solutions like MQA, matter very little to me as these are all solutions solving a problem that has already been well solved +10 years ago.

    Keep up the good writings and enjoy the music!

    Mitch

    ReplyDelete
    Replies
    1. Thanks again Mitch for your comment and certainly exhaustive experience in this area over the years!

      I really like the practical suggestion of trying the "Fun with Digital Audio" article. It's things like this that may surprise the folks who seem to have a notion that hearing is somehow almost capable of infinite resolution.

      Again great job with the DSP e-Book. You might find the recent Stereophile article on their web on MQA and subsequent discussions interesting!

      Delete
  6. Hi, I was wondering if you could help me use RightMark Audio Analyzer. I don't know if this is a bad place to ask. I don't see an email.

    My problem is that test results of USB devices don't seem to register or show up when I do a test. I either get blank test results, or, when using the "Generate/Analyze" buttons, gets identical results with various devices (like dummy data). Any help is appreciated.

    ReplyDelete
    Replies
    1. Hi there SP, can you give me more details on what hardware you're using?

      Are you seeing the calibration signal and able to adjust amplitude to the appropriate target? Feel free to PM me on Computer Audiophile or the Squeezebox forum as well!

      Delete
    2. I'm using a Windows 10 64-bit PC with two USB-powered dac/amps to test. They both show up in the device listing of Rightmark. What are the steps I follow to see test results and compare them? I see the buttons at the bottom: Run tests (Playback/recording, etc) and Generate/Analyze (Generate WAV, Analyze WAV, etc.) When I run the test I get blank results (but do hear the beeps and sine sweeps running). Am I supposed to keep the headphone disconnected? At a certain volume? Do I use "Generate/Analyze" first before running the test? Confused. Thanks. I Did try reading the manual.

      Delete
  7. Hi Archimago,

    As always, you've come up trumps with another thoughtful article. This is a little off topic, but since you mentioned it, ...... South East Asia.

    If you're going to Bangkok (and why not?) there are half a dozen or so real high end HiFi stores in the Siam 'Paragon' Shopping Mall, up on Level 4, I think (Bangkok is full of shopping malls, but Siam Paragon is perhaps the most central and easiest to find - well worth a browse, and I highly recommend the coconut cake in any Starbucks in Thailand). There's a bunch of headphones shops too, but you're on vacation so go for the som tum or Thai massage instead.

    Also, if it's Singapore you're going to, then a visit to the Adelphi Mall on Coleman Street is a good way to spend a rainy afternoon.

    Raspberry Pi? Nah, coconut cake! :-)

    Bob

    ReplyDelete
  8. Very informative, keep posting such good articles, it really helps to know about things. Thanks for sharing useful information.....
    Desktop computers

    ReplyDelete