MUSINGS: What Is The Value of High Resolution Audio (HRA)?

enough?

I promised not to talk about Pono anymore, so I won't to a significant degree :-). This article is one about high-resolution audio in general.

It has been interesting seeing the audiophile press's responses to articles such as:
Gizmodo's "Don't Buy What Neil Young Is Selling"
Pitchfork's "The Myth and the Reality of the $43 Download"

with these:

Analog Planet's (Michael Fremer) "Gizmodo Won't Post My Comment So I'm Posting It Here"
AudioStream's (Michael Lavorgna) "Is High Resolution Audio Elitist?"

As I had laid out in this blog a number of moons ago (March 2014 to be exact) in "High Resolution Audio (HRA) Expectations (A Critical Review)...", there are many challenges to overcome in order to perceive an audible difference between a true high-resolution recording and the same track down-sampled to standard CD (16/44) resolution. This challenge is significant and technically difficult; culminating in the ultimate question of whether one's own hearing mechanism is even capable of the feat. (Remember folks, aging is good for wine, not so much for hearing acuity!)

I think there are a number of issues here and it's important to treat each separately without getting overly simplistic into a single declaration of whether HRA is "good/needed" or "bad/worthless". Furthermore gentlemen, there's no need for name-calling or ad hominem attacks.

For your consideration, here's how I see it:

1. Is the high resolution 24/96+ PCM format better (more accurate)?
Of course! Assuming we have a good quality recording, high-resolution formats afford better objective dynamic range for the music and accurately records more of the spectrum than 44kHz sampling rate. We know that LPs retain more than 22kHz of sonic information (the resolution of course is limited in dynamic range and distortions are higher with LPs). High sample rates will get us away from any concerns around ringing due to filters functioning near the audio spectrum which Fremer refers to. But folks, let's not overplay this either because "issues" like pre-ringing from digital filters are of questionable audibility and papers like this one from the Meridian folks (see this thread by page 5 for details and criticisms) show that even if we ignore all the concerns raised, aggregate correct responses was only 56.25% (160 trials) - just above their statistical significance level.

Higher 24-bit resolution is obviously beneficial in the studio to allow for more accurate digital processing. These days I'm doing digital room correction on playback with convolution filters so I think it's nice to have 24-bit files for that extra bit of accuracy during playback.

As a perfectionist audiophile, I don't think there's anything wrong with wanting the most accurate version of the album if available as a high-resolution file representing the "studio master". However, we have to realize that the perfectionist audiophile's desires are not the same as most mainstream music lovers. I don't know if anyone has done a demographic survey, but I'm sure it's a pretty small sliver of the music-buying public that would even care.

2. Is High Resolution Audio (HRA) audible?
For the vast majority of people, I believe the answer is clearly NO.

Differences are at best subtle. Neil Young's musician buddies (video) are clearly over-dramatic about what they heard or he did something to the car audio to accentuate the difference between MP3/lossless/hi-res in my opinion. CD vs. HRA is not analogous to the visible difference between DVD 480P and Blu-Ray 1080P obviously, otherwise we wouldn't be arguing about this. The 24-bit audio test performed here last year as well as the 44kHz vs. 88kHz sampling rate discrimination test from this study for example are consistent with this conclusion (for the 44 vs. 88kHz paper, the abstract was vague and I think misleading, look at the overall results and you see only 3/16 listeners able to get significant results but selected the wrong answer consistently; 13/16 trained listeners scored non-significantly). Other studies in the literature over the years have not been able to show significant effects at all (such as this 2005 study with 24/192).

Despite the above, who knows, maybe there are lucky (and more than likely young!) folks who have awesome auditory acuity. The DAC has to be good enough. Amps, speakers, headphones will need to be up to the task of reproducing the dynamic nuances and high frequency response in a reasonably flat fashion without exciting too much intermodulation distortion or high frequency ringing (like with the tweeter in the old Meridian DSP8000 active speaker measurements). Plus you need a quiet sound room. (Remember folks, research studies are generally conducted in controlled ideal environments using equipment with known objective capabilities. I have a sneaky suspicion that showing off hi-res sound quality in a car ain't gonna cut it!)

[For completeness, I've included Appendix A below for those who want to think about the Oohashi "Hypersonic Effect" referred to in Fremer's article.]

As demonstrated by the Fremer and Lavorgna articles, the audiophile press really likes to claim that the Meyer-Moran study from 2007 has been "debunked". One really should not expect large magnitude differences anyway based on existing literature, so the Meyer-Moran negative study is absolutely to be expected. For those who are unfamiliar, this study basically is a blinded ABX test to see if members (presumably audio enthusiasts) of the Boston Audio Society, around 60 of them, can tell the difference between "high resolution" SACDs and DVD-As played back directly versus going through a 16/44 A/D/A chain (basically "dumbing" the signal down to CD resolution). You can look at the equipment and music used here. The result was that respondents detected the 16/44 "loop" playback accurately 49.8% of the time - purely chance. No evidence the women or younger folks were any better either. To make a long story short, I think it's fair to criticize the study for using questionable high-resolution recordings (like old classical albums, or questionable quality analogue recordings, I've listed many SACDs already which are likely just upsampled PCM here). But remember these recordings were available and sold as high-resolution and would be typical audiophile fare back in 2007. There were at least six albums from Chesky, a couple from Telarc, Steely Dan's Two Against Nature DVD-A, the Dark Side Of The Moon remastered SACD from 2003, and Patricia Barber's Nightclub SACD from MFSL. Sure, we can discount the study as "flawed" because the music wasn't good enough or of high enough quality, but then we would have to contend with the following question...

3. Do we actually have many albums worthy of High Resolution Audio?
As I indicated last week, this is the BIG problem - literally the elephant in the room. I can understand why the industry may not want to talk about it! Because to change it will require massive effort and criticism of what has been "business as usual" for decades in terms of digital production standards since the mid-1990's. And it also means taking a hard look at whether there is truly any benefit issuing old recordings in the analogue era for high-resolution reissues.

Other than new audiophile all-digital recordings (mainly classical, jazz, vocals from specialized sources like Channel Classics, 2L, AIX, etc.), or maybe remasters from high-quality sources from Audio Fidelity/MFSL/SHM-SACD, etc... the vast majority of music does not require the resolution of HRA whatsoever. Heck, most of the top-40 pop/rock tunes don't even challenge high bitrate MP3. Mark Waldrep (Dr. AIX) has been warning us about this for years in his blog - I believe he's right. Highly dynamically compressed music with high inherent noise floors and unnatural recordings that were done without intent to preserve the full frequency spectrum or decent dynamic range does not sound any better in 24/96+. In fact, to truly take advantage of the resolution available, one must ensure full resolution through the whole production chain from recording to mixing to ensuring digital processing is done with adequate precision, to the final mastering step. I think Waldrep is right that the high resolution era (that is, capable of utilizing both >16-bit resolution, and >44/48kHz sample rate) truly began after the availability of high-quality digital recording gear (mid to late-90's?). All the music before then in all likelihood may only benefit from higher sampling rate, but 16-bits is all that's needed. As an example, although we can use 96kHz+ sampling on our favourite analogue recordings to preserve as much of the extended frequency as reasonable, I think most of us realize that as much as one may "love" yet-another-remaster of Miles Davis' Kind Of Blue, Jazz At The Pawnshop, Dave Brubeck, Living Stereo classics or the Rudy Van Gelder discography, the dynamic range for these recordings are limited and 16-bits would be more than enough to capture everything down to the tape noise.

It's easier isn't it to pretend that all music can be digitized in high-resolution and sold in this "new and improved" format? The music industry of course would loathe to not be able to sell yet another re-issue (this time with the HRA sticker on the cover) and have us all buy another copy of something we already have...

4. How much 'should' this cost?
Ah, the billion dollar question! The other day, I was in BestBuy and noted that the high-resolution Blu-Ray copy of Gone Girl goes for $25 and the "standard resolution" DVD was $20 (enjoyable movie BTW especially if you're a David Fincher fan). So, unless one is still stuck with a small (<30") TV or incapable of viewing higher resolution video, having access to a high-resolution movie (which by the way also has lossless surround soundtrack to boot) costs 25% more. Considering that HRA isn't as easy to differentiate from a CD compared to 480P from 1080P video, should we even be charged a 25% premium? Consider that a CD can be bought these days for $10, how much do you think we should be charged for a digital music download when all it is is a data copy sent down a utility which I personally pay for (ie. my internet provider's monthly cost)? We do not get a plastic case, printed cover/booklet, or a piece of polycarbonate in hand even. Furthermore, the CD can be resold! Can I resell a 24/96 music download?

Ultimately the market will sniff out what the cost should be based on demand... Personally, I have purchased many SACDs and DVD-As over the years at significant premiums over the CD. Only a couple music Blu-Rays so far. Truth be told, I mostly buy those with a 5.1 surround mix or maybe a 3.0 mix (like the Analogue Productions Nat King Cole SACDs) and there being a "collectability" component to the purchase which a download will never have. If I were to throw out a number, I think $13 for a high-resolution 24/96+ FLAC download is OK with me (assuming the CD costs $10) knowing that it's a more specialized item and there's a cost premium to that. Short of some kind of inflationary tailspin, I can't see spending $20+/album as reasonable for a standard digital download. I personally cannot see the jump from 24/96 to 24/192 as representing any value so would not likely pay more than maybe a dollar or two. I'd love to hear what others think should be the target price.

High Resolution Elitism!?

In the "Is High Resolution Audio Elitist" post by Michael Lavorgna, he suggests "you have to wonder if some of the stronger negative takes on what is a plea for better sound quality are rooted in an emotional response as opposed to a technical one". Indeed, there has been a strong negative emotional response hasn't there? But really, is this surprising? Considering the unrealistic hype around HRA (like the 'artist coming out of car' video above), and misleading diagrams (the underwater qualitative diagram, Sony's stair-step digital waveform), it's no wonder folks who know a thing or two about digital audio cannot help but feel perturbed by the claims. Mr. Lavorgna, it's not "envy" that is the primary emotion; it's actually disgust.

I find it fascinating that Mr. Lavorgna will even bring up the price issue or "elitism" with HRA. The truth is that even relatively inexpensive devices these days like say a $200 24/192 DAC off eBay from China can easily achieve high-resolution playback (even this oldie). USB-stick DACs like the AudioEngine D3 are also fine (up to 24/96 in this case). What's the big deal?! From a hardware perspective, $$$ should really be spent on quality speakers, a decent sound room, and room treatments for the best "bang for the buck" in general and especially for high resolution music reproduction. In fact, I think it's important to scrutinize expensive audiophile gear and ask for objective evaluation - even the highly touted PS Audio PerfectWave DirectStream DAC doesn't have an impressive measured noise floor that can benefit from >17-bits resolution (it sounds good at the local dealer BTW, but the accuracy is measurably limited). Definitely avoid weird expensive tube DACs like the Allnic and Lector Strumenti. Objective reviews with measurements become even more important when we get into high-resolution audio. Furthermore, just because some piece of equipment is expensive doesn't mean it's desirable. And just because some people criticize the cost of the hardware and software doesn't mean there's an underlying unfulfilled desire because they're lacking financial resources available to the "elite" of this world! Sometimes the asking price is just obviously ridiculous for what one gets. (And "sometimes a cigar is just a cigar.")

In Summary...
Ultimately, I hope folks don't get too side tracked by things like the hardware, whether any particular company/spokesperson is worth backing (ie. Pono/Young, HDTracks/Chesky, Sony Walkman, Astell&Kern, etc...), or even what encoding technique (DSD, PCM-FLAC, PCM-MQA/Meridian/Stuart, etc...). HRA has been around for more than a decade with SACD and DVD-A; it's not really that new or sexy for many of us listening to this stuff for awhile. Better recordings are really what the world needs, not bigger file sizes. Recordings truly worthy of 24-bits and >44kHz because care and judicious processing were applied to maintain nuances and realism. Whether we can hear the difference with HRA or not I think we have to leave to each audiophile to decide through experience, intellectual consideration, or likely some combination of the two. In time, we will see just what "value" digital high resolution recordings hold in the marketplace from a cost perspective and whether lossless/high-resolution store fronts are able to succeed in the face of competition like the streaming services and traditional physical media. If companies, consumers, reviewers, and the press unite to advocate for and get us better sounding albums that can actually benefit from high resolution instead of the crappy, loud, typical mastering "quality" we've been subjected to in the last 2 decades, we all win. This, in my opinion, is the evangelical "mission" which audiophiles and music lovers should be pursuing.

Isn't it ironic that over time, for the most part, hardware like DACs objectively improve and become cheaper, yet it seems like the mainstream music software side just gets further away from high fidelity and realistic sounding recordings?

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

Appendix A: The Oohashi "Hypersonic Effect"

I'm sure some folks will raise the spectre of the Oohashi (J Neurophysiol, 2000) study as evidence that ultrasonic frequencies make a difference - the "hypersonic effect" (a presentation was made by the same group in 2002). Michael Fremer already has in his spiel. (I see he just posted up another piece using this paper as the main point.)

In a nutshell, this study showed that there was enhanced alpha-frequency EEG occipital-parietal power when test subjects were played music with extended frequency response (up to 50kHz; the 20-50kHz ultrasonic contribution peaking at -30dB compared to the audible spectrum normalized at 0dB), and also PET scanning demonstrated increased deep brain activity (increased rCBF in the midbrain and lateral left thalamus) with full frequency music which was not noted when a low-pass filtered version of the sound was presented. I actually like this study and it's one of those nuggets in the audiophile psyche that stands out as a fascinating talking/thinking point! Remember though that functional neuroimaging is a hot topic these days, and there are many reports out there of questionable significance. Suppose we accept fully the methodology, there still remains the question of what it all means... For example, the recording equipment is a "high-speed one-bit coding signal processor operating at 1.92 MHz" - I'm not clear how good this is compared to modern ADC/DACs (it's supposed to have flat frequency response over 100kHz, so did they use noise shaping, if so, doesn't that introduce ultrasonic noise? The sample rate is obviously lower than SACD/DSD64 at 2.8MHz so noise shaping could be more intense!). The music chosen was the "Gambang Kuta" from Bali (have a listen here) - lots of high frequency content; can the results be generalized to Western music? We don't know if the test subjects even like this music so it's worth considering if subjective pleasure would change these results (what if most of the test subjects thought the "music" sounded like fingernails on chalkboard!?). Oohashi also designed the Pioneer-made "super-tweeter" that's supposedly flat to 100kHz - practically, how many reviewers/writers/listeners have speakers/headphones capable of this?

As far as I am aware, nobody has replicated the PET results. The same Oohashi group has now published again in 2014 (Fukushima, PLOS, 2014) using DSD128 (1-bit, 5.6MHz) and the TAD PT-R9 ribbon super-tweeter. They used the same Bali music. This time they feel there is both a positive hypersonic effect (with high frequencies >32kHz added to the audible component) and a negative hypersonic effect (for lower ultrasonic frequencies up to 32kHz) determined by whether the alpha-EEG power increases or decreases! There's even a vague reference to whether there's safety issues with these high-frequencies. Ultimately this appears to be even more confusing. Since they didn't ask about perceived subjective sound quality, we don't seem to know which (positive or negative) hypersonic effect sounded better in this study! If the negative hypersonic effect is "bad" for perceived quality, does that mean it's better to low pass down to 20kHz than retain all the frequencies up to 32kHz? But if the music recording has lots of frequencies >32kHz, then it's better to retain that since the brain then experiences a positive hypersonic effect? Even if these neurophysiological effects were real and replicable, should we even care if there's no conscious awareness? (It's worth mentioning again that DSD generally uses noise shaping and adds to the ultrasonic noise, so in the 2014 study, what happened to the noise that usually accompanies DSD128 playback starting around 50kHz?! In 2014, 24/192 PCM would have been better with less quantization noise I think.)

Bottom line: Before we accept theories around the importance of high frequencies affecting central nervous system functioning, realize that the data is limited and significance unclear. I think it's highly speculative to link these studies with the idea that they argue for high-resolution audio.

One more thing while we're speculating here... I wonder why there's no discussion about beta EEG frequencies? Cortical processing and alertness is correlated with the higher frequency beta activity. Studies of temporal processing for example will use beta-band activity for auditory guidance (like this one). Furthermore, alpha tends to be concentrated occipitally which is of course the major visual processing cortical area (usually when eyes closed, at rest)... Since we're thinking about music as perhaps being emotionally and cognitively engaging, we really should be looking for frontal and temporal activity I suspect although sensory modalities can cross-affect processing (example).

[Of interest, in the Oohashi 2000 paper, they also claimed that the 26 young (18-31, 42% female!) Japanese subjects who took the "psychological experiment" portion scored significantly and presumably in favour of the full-frequency test sample compared to presumably a 22kHz low-pass version (I'm unclear because they also used a 26kHz cut-off version for one of the EEG tests). I would have loved to see a more rigorous examination of the "psychological experiment" portion since they claim that the respondents were able to hear the full-spectrum sample as "softer, more reverberant, with a better balance of instruments, more comfortable to the ears, and richer in nuance". Considering all the negative or barely significant studies in the literature over the years, even this alone would be worth discussing irrespective of EEG or PET data!]