 |
| A little bit of dramatics for effect. ;-) |
Like many audiophiles, I've been well aware of Patricia Barber's recordings over the years. Her discography of "reference"-sounding quality albums have already had decades of influence among hi-fi enthusiasts with increased exposure beginning in the 1990's (her first album was Split in 1989). Personally, I was introduced to her music with Café Blue released in 1994.
I think Barber is an interesting contrast to the other well-known female jazz vocalist we all know well - Diana Krall. Barber's material tends to be more original (less of the standards), more edgy, compared to Krall's accessibility and mainstream appeal.
Already, on this blog, I've mentioned the most recent album Clique! (2021, DR14 hi-res, DR13 MQA-CD) but we're not going to be mainly talking about the music on the album in this post. Rather, I want to explore the opinions and beliefs of Michael Fremer in his article "How Best to Hear Patricia Barber's Clique!" recently posted on Tracking Angle.
As far as I can tell, in that article, Fremer did not have access to the actual "master" hi-res DXD files (24 or 32-bit 352.8kHz) nor other high-resolution PCM version like 24/176.4 derived from the actual master, available for purchase at the usual places such as HDtracks and NativeDSD Music (the album is not native DSD).
Instead, Fremer is basing his listening opinions on this (bold added for emphasis):
Both the CD and SACD editions are also MQA encoded, which means if you transfer the CD files onto a computer and load it into Roon or Audirvana, when you stream it from your drive if your DAC does MQA it will unfold the original 384/24 bit DXD "master file". In other words you get the "master tape"! Not a copy of the master, not a third generation of the master, but the master.
Despite the years of controversy around what MQA is, numerous posts by various authors (including myself), the widespread awareness by many audiophiles (look at the comments on his post) that MQA at best provides a lossy reconstruction of the high-resolution signal, Fremer appears to be in some kind of "fact-free" bubble where MQA decoding actually reconstructs the original "master file" - not some facsimile, or lossy reconstruction, but "the master" as he puts it.
To make matters worse, Fremer isn't talking about 24-bit MQA files, but claims the 16/44.1 "MQA"-CD is capable of delivering that "master file" when decoded. To make this claim is astonishing because it shows his phenomenal lack of understanding as to how digital audio works and the impossibility of the 16/44.1 data to ever fully include all the nuances of an original "384/24" (DXD is actually 352.8kHz) file.
Based on that erroneous belief, he then makes a comparison between the LP with the "MQA edition through a good DAC" and makes interchangeable comments about the MQA decoded sound and "DXD version". What a thoughtless mess!
I'm amazed by this statement (again, bold added for emphasis):
Is this Jim Anderson's best sounding Barber album yet? I don't like to dance on the head of a superlative so let's just say it's a "you are there" recording. If you're used to the vinyl edition and then hear the DXD version, you'll know from the opening bass line of Lee Hazelwood's "This Town" that this version is in another class. Barber's voice sounds "sweet" on the LP. She sounds eerily in your room on the master file. There's a serious uptick in overall transparency and macrodynamic detail and resolution.
So the supposed "DXD version" from 16/44.1 "MQA edition through a good DAC" makes the sound eerily good. There's a "serious uptick" in quality eh? And let's not forget that:
The DXD file sounds better than the disc's DSD layer: cleaner transients, greater transparency, etc.
Wow! So the MQA-CD decoded version is the ultimate experience even over SACD and vinyl! If we take a step back at this point, let's discuss what we know about 16/44.1 MQA-CD based on testing by others and myself over the years:
1. It's not the original "master tape" data that gets reconstituted even if the decoder spits out the same number of bits as what was fed in.
2. There is no ability to reconstruct the full 24-bits (much less 32-bits) and 352.8kHz of the actual master file.
3. With only 16-bits to use, there is not even enough data to reconstruct a good lossy representation beyond 22.05kHz in the frequency domain.
4. At best, MQA-CD is lossless 15-bit audio with noise-shaped dithering and limited quality digital upsampling filters applied assuming the MQA control stream was assigned down in the lowest 16th bit. The system literally has to rob bits from the audio data stream to let the DAC/decoder know it's there including the "authentication" cryptographic signature.
Now you may be asking at this point:
"But Arch, how do you know that the MQA-CD decoded Clique! isn't exactly like the DXD master? Maybe Fremer has untarnishable Golden Ears and his hearing ability is the GOAT!"
And this, my friends, is why we need "evidence" don't we when talking about audio stuff? Far too many people hand-wave about this and that for way too long without any sense of whether they know anything at all about what they're talking about.
Let's have a look at the MQA-CD data and decode the first song "This Town" in Roon - easily done:
The "Authentication" step tells us that the intended final samplerate is indeed DXD 352.8kHz. Using Roon's "Core" decode which reconstitutes the output to 88.2kHz, we capture that PCM stream using VB-Audio Virtual Cable at the appropriate samplerate. As discussed over the years, it is the "Rendering" phase of MQA decoding that further upsamples the Core decode into that final 352.8kHz target if the DAC is capable of it.
So, if we standardize using the 88.2kHz MQA Core output samplerate, we can resample the original DXD Master (which I bought around the time the album was released), a blog reader sent me a copy of the SACD/DSD64 rip of the song to convert to PCM, and I upsampled the MQA-CD data from the SACD's CD layer, while aligning the waveform peak values to create this comparison graph:
Whereas DSD64/SACD is a high-resolution system capable of maintaining excellent correlation with the DXD master out to >30kHz, after which the noise increases beyond the recorded noise floor, notice the subtle tonality change in the MQA-CD versions within the audible spectrum below 20kHz. In particular, the Core decoded MQA version is subtly "less bright" in the 10-20kHz octave. To a large extent, this is likely the result of the digital upsampling used by the MQA decoder as it recreates the 88.2kHz content from 44.1kHz using a more gentle filter than the steep ("ideal" Chord-like, high tap length) setting that I used with the yellow undecoded signal.
This subtle tonal shift exists across the whole album. For example, here's an FFT at 2:52 during one of the louder portions of the song:
What's the take-home message?
This exercise has not taught us anything new about MQA. The results here are the same as what we've been aware of since 2018, and the 4 points I listed above about MQA-CD are consistent with the findings.
Rather, I think we have learned something about Michael Fremer - how he thinks, how well he perceives supposed differences, and how he writes about things he doesn't quite seem to understand:
1. Fremer appears to be gullible and seems to have bought into the claims of the MQA company and Bob Stuart that they are able to encapsulate the "master file" with their technology. We've known for years that this is not possible, but apparently the message did not get through to some of these guys at the main North American audiophile magazines whether Stereophile or TAS when every once awhile they drag out MQA in their reviews. I wonder whether they're just purposely "trolling" these days to get more clicks because they know some audiophiles will get a bit upset about this "revolutionary" nonsense.
2. Fremer does not understand how digital audio works. No self-respecting, knowledgeable, rational audiophile with awareness of PCM encoding would agree to the preposterous claim that a 16/44.1 "MQA-CD" stream is capable of being decoded to a full 24/352.8 DXD master! A music listener who's getting into hi-fi posting on a forum might be forgiven for having not thought this through, but Fremer is a veteran audiophile "journalist". This is simply embarrassing.
3. Fremer can't hear the difference between the <16-bit resolution MQA decode from the true high-resolution quality of the SACD playback which would be closest to the actual DXD "master tape". Furthermore, he believes that the <16-bit PCM version sounds better than vinyl. I actually don't disagree with this latter point, since vinyl resolution is poor, but then again I'm no vinyl evangelist and he's supposed to dislike CD quality! I fear that he has clearly lost his Golden Ears, especially for those higher frequencies above 10kHz, and obviously cannot hear the difference between high-res digital audio and something that's basically CD-quality (not that I expect him to of course ;-).
4. Fremer clearly is exaggerating or misperceiving what he supposedly hears. There's no way for the MQA-decoded version to have "cleaner transients, greater transparency" than the DSD layer in particular, supposing of course we accept that <16-bit PCM is better than vinyl.
Yet again, I honestly wonder about the basic educational level of the "audiophile reviewer" and how well these guys actually hear! I have to laugh while doing some researching into what Michael Fremer wrote in the past about the sound of digital such as this one from 1994 (posted online in 2009): "Ten Years Into The Digital Revolution: A Continuing Disaster in Sound". There has been no "disaster" when we listen to a well-engineered recording like Clique! The real disaster is in the foolish quality of Fremer's articles for many years now when he refuses to understand reasonable principles behind the technologies he talks about.
Is it any wonder why audiophiles are seen as "audiophools" when time and again, the "elder spokesmen" publish articles of this nature? Perhaps it's time for someone close to Mr. Fremer to gently give him a tap on the shoulder and suggest that it's time to retire from high-fidelity digital audio reviewing. I'm sure he'll still be great at vinyl unboxings, humorous videos, music reviews and turntable setups.
--------------------
BTW, here's a picture I posted back in August 2022 from the Pacific Audio Fest in Seattle:
Fremer himself was interviewing Anderson and Schwarz, the audio engineers involved in Clique! It was very clear that they wanted the highest quality they could achieve, hence all recording and processing was done in DXD. In my discussions with them later, it was quite clear that further processing like what was done with MQA would be seen as compromising their project.
That Fremer would now post an article suggesting that audiophiles literally listen to the lowest resolution version of the album as being the "best" sound is even more bizarre given the discussions that ensued over that hour about the high-resolution aspirations for the album!
To be honest, I had the impression that Fremer wasn't actually listening when he was interviewing / moderating anyways. Not surprisingly, he seemed to have his own internal script about the greatness of analogue and vinyl which came out at times throughout the session. I suspect there are multiple levels and forms of "selective hearing" going on in his mind.
[PS: I noticed in the comments for Fremer's article that Ulrike Schwarz commented and Jim Anderson's name was brought up. Let me ignore that for the time being as they might want to think about this discussion. Let's deal with what Fremer should have known... ;-]
Happy listening audiophiles.
As usual, be careful with all the claims and allegations out there typically spoken without evidence. If you have the opportunity, I do recommend listening to the multichannel version of Clique! (such as on the SACD). In fact, I'm listening to it right now as I finish this article. Wonderful sense of spaciousness with palpable impression of Patricia Barber singing, "embodied", in the sound room.
Bonus:
Just thought I'd add a bonus graph before packing up the files...
This is showing the full bandwidth all the way to 176.4kHz which is captured in the DXD (24/352.8) master file and a comparison with DSD64/SACD version converted to PCM 352.8.
As you can see, the SACD does a good job of tracking the DXD master all the way to around 35kHz or so and then the SDM modulator noise takes over and remains high thereafter. This is why DSD64 playback DACs usually employ analogue filters starting >30kHz to keep all that ultrasonic content away from the amplifier.
[To be clear, these days, I'm very much "Post Hi-Res" in my mindset so I'm not suggesting that one can hear the difference 352.8kHz samplerate makes! A well produced 24/96 is more than good enough for me with genuine high dynamic range, high resolution provenance recordings. Otherwise, I'll happily downsample to something like 16/48 for many albums.]
ADDENDUM - Extended Edition! (Feb 2, 2023)
Thanks for the comments folks.
Based on the discussions, here are a few further analyses to consider. First, using Paul K.'s DeltaWave to create the delta file, I ran a comparison between the original DXD 24/352.8 (downsampled to 24/88.2) vs. MQA-CD Core decode (24/88.2) to have a look at the residual "nulled" difference. For consistency with the other graphs above, I imported this into Adobe Audition to examine:
Interesting. Notice I used the log scale to show the difference in the low frequencies, not just the highs and ultrasonics. While obviously the MQA-CD is not the same as the DXD master, for some reason, the right channel shows more variation than the left! I wonder what they did between the DXD and MQA-CD versions to cause this!?
On average, the difference is less than -65dB between the two versions so it should still be subtle.
Here's the audio spectrogram view from about 2:00 to 4:30:
The difference in the right channel is also seen in the undecoded MQA-CD version itself, so it's not just an artifact of the Roon MQA Core Decode process. I assume this isn't some kind of watermarking affecting only the right channel.
Peter Veth talked about the idea that when we use an MQA-decoding DAC, "Every MQA certified DAC will perform 2nd and even 3rd unfolds which will show the full spectrum of the original DXD master. Spectrum analysis on MQA should be done by resampling the final analogue output." Thanks for the suggestion:
As you can see, I'm sending the Clique! MQA-CD data to the Sabaj A20d 2022 DAC for fully decoded playback. It accurately identifies the target 352.8kHz sample rate. While it's playing the song, I recorded the XLR output with the RME ADI-2 Pro FS ADC (I used 384kHz to capture beyond the target).
BTW, notice that the Clique! MQA-CD is a "green LED" album, not "blue LED" MQA Studio so presumably the original artist and engineers did not sign off on this version?
There's your second or third or whatever MQA "unfold" recorded from the analogue output.
On the one hand, it's a pretty good approximation out to ~30kHz for this track I suppose (remember that this is just the average spectrum across the whole song, details obscured). I suspect the Core decode process on the XMOS DAC performs differently from Roon. However, with 16 bits, as I was saying, there's just not enough data to create a lossy reconstruction out to a full 44.1kHz bandwidth as 24-bit MQA might be able to do, much less "the full spectrum of the original DXD master"!
Bottom line guys & gals: The more we look into the MQA-CD version of Clique!, whether software decoded in Roon or using an MQA-capable DAC with full decode, the more it's obvious that MQA-CD does not unfold to "the master".
Audiophiles, don't be fooled by superlatives about how it "takes (you) there", represents "the best sound", "gets (you) closer", is the "exact" studio recording, or suggestions that there's some special "time domain" improvement. MQA has always been a kind of "partially lossy codec" with pseudo-hi-res aspirations done with various not-particularly-accurate DSP tricks like weak filtering.
Dump MQA - there's just no point. Avoid enablers of MQA like Tidal's HiFi Plus tier, turn off software decoding in Roon, and try to avoid hardware that licenses the codec (and makes you pay for it by bumping the cost) to send them a message.
To magazine writers like Michael Fremer, please get your facts straight and speak the truth. The audiophile press looks more ridiculous every month it tries to hype this in the reviews they publish.
And Peter Veth, thanks for the suggestion and helping to prove that the decoded analogue output is clearly not the same as the original DXD master.
The "tonal shift" as shown in the FFT is interesting. The MQA Core Decoded version has what I would refer to as "high frequency roll off", and it is of sufficient magnitude to potentially allow the MQA version to sound little "nicer" in a system that borders on being a little bright. It actually looks pretty much equivalent to what you would expect from turning the treble knob down by -4dB. Do we know what is causing this? Clearly it is invoked by the core decoding, but is this HF cut consistent across all MQA core decodes?
ReplyDeleteHi Bogleto,
DeleteThat HF roll-off is basically a result of the upsampling filter they use. Sort of like the weak filter for something like PonoPlayer:
http://archimago.blogspot.com/2015/08/measurements-ponoplayer-another-mans.html
But not that dramatic.
I can't say for sure how consistent this is. Who knows if different implementations of the decoder, for example Roon vs. Audirvana vs. XMOS DAC decoding will do exactly the same thing. Suspicion is that it'll be very consistent across the platforms, I think!
Thank you, Arch. Yes there is some magic in Barber's recordings, even Tidal basic through Bluetooth to my stereos sounds wonderful.
ReplyDeleteAgree Juhazi,
DeleteAlways nice to really have access to the "master" quality without unnecessary processing of course.
Thanks for saving me from reading that article. Yep, it's a great sounding recording. Yep, following Fremer's advice is a waste of time and money. These days even laughingstocks make bank from their herd of followers. Stupid, provocative, and dangerously untrue pays the bills for these shameless prancers.
ReplyDeleteI wonder if there have been studies in the literature about shameless "influencers" like this, Doug.
DeleteI can imagine that for short periods of time, contentious articles like this might get some clicks (even from folks like myself who would not frequent Tracking Angle). But in the longterm, the entertainment value wears thin and even the regulars might realize that there's no point.
I think he has been fortunate riding the wave of vinyl's resurgence just like many physical assets have grown in value over the years as "collector's items". I still think vinyl's resurgence is very much waning now with audiophiles evaluating the true sound quality of these things - perhaps spurred by revelations like the MoFi "one step" DSD controversy, eating into the cult following of "anti-digital" people. Probably will correlate with the decline in interest in Fremer's influence and opinions.
The "one step" debacle seriously damaged his cred, as he reviewed the vinyl and proclaimed it the ultimate version. Now he will tell you the reasons that it makes sense that an LP mastered from digital is superior.
DeleteNever acknowledges that it's simply euphonic (to him) distortion.
The vinyl lovers never really have a good answer for how the LP can sound better than the digital source played back digitally, other than to reference some form of "magic" added by analog.
Hey Danny,
DeleteAnd that was another completely predictable outcome wasn't it? It only took a company like MoFi to be sneaky about it for years until finally the truth to be revealed that analogue adherents never could tell the difference between an "all analogue" chain compared to one with digital steps!
For hi-res digital folks, it's already intuitive that "of course" a good hi-res recording (that's not overly dynamically compressed, natural sounding) will translate well when cut onto vinyl. The limitations of vinyl resulting in resolution limits (smoothness?), temporal distortions (smoothness?), frequency variations (warm?) had always been the "magic" that defined much of what was heard.
Obviously the vinyl faithful really liked those distortions even if they could never allow themselves to speak about it like that!
Stating "mqa is lossy. You may like it, but it's lossy. Just stating a fact" on facebook leads to deletion of comment and one month suspension.
DeleteYikes jherbert,
DeleteI guess that's a literal example that he/they "can't handle the truth"!
Which Facebook page did this happen on?
*Fremer, phone auto fix not always the best as we see ;)
ReplyDeleteKeep fighting the good fight Arch.
ReplyDeleteAdmittedly I just can't get interested in the MQA debate. I'm good with CD quality and vinyl.
Greetings Vaal,
DeleteI don't blame you man for being uninterested... I suspect the vast majority of music listeners are uninterested as well!
IMO, there's really no "debate" in that when it comes to facts, evidence, logic, science, and just plain rational talking points, it's been a one-sided affair. The only thing the pro-MQA side has provided are opinions, claims, "testimonials", and MQA-sponsored unsubstantiated materials (like questionable Bob Stuart articles).
I hope one day, this whole debacle will be looked back on as a case study in the futility of spending millions of dollars (angel investors, even the music industry) trying to sell something to the public that is technically inferior. It won't work if hobbyists are educated even if the advertising people were able to corner the market on gathering the favour of the press, and have articles in the AES.
Objective truth eventually will win out.
Indeed Milan,
ReplyDeleteFremer has provided years of entertainment... Sometimes good, other times like this rather unfortunate for the audiophile hobby I think. I see that he has a Wiki page:
https://en.wikipedia.org/wiki/Michael_Fremer
At 75+ years old now, I think it's fair to say that his hearing is limited and probably there won't be too many articles left in the tank when it comes to evaluation of high-resolution, high-fidelity gear and music...
My approach to MQA is ridiculously simple: Rather than attempting to understand it's shortcomings, I simply avoid it like the plague. I'd rather devote my processing resources learning something that will improve my life or my understanding of the universe. I already know everything I need to know about MQA: It's a worthless gimmick that actually degrades what it claims to improve.
ReplyDeleteIndeed Phoenix,
DeleteI seriously hope most music listeners and audiophiles will just avoid it, avoid reviewers who hype it, and products that advertise it.
That and so long as there are those who speak truth about nonsense like this without fear when needed, we'll all be better for it, and likewise I think the hobby healthier for it.
Hi Phoenix, maybe this recent review in Stereophile of The CH Precision C1.2 D/A Controller will enlighten you a bit about the concept of Time-Smearing and MQA https://www.stereophile.com/content/ch-precision-c12-da-processor
DeleteI don't think that article enlightens anything. ;-)
DeleteThe idea of short filters (as indicated by short impulse responses) being correlated to reducing "time-smearing" is another silly idea Meridian and Bob Stuart introduced back in 2009 or so. It does no such thing. All that changes is latency which may or may not be significant depending on what you're doing.
Oh yes hearing loss could be a thing with all these 'old' folks. I make A/B level match test with some of very pricey high end dacs. And I can see how somehow these devices could be called as having more details etc.
ReplyDeleteThe only thing that I heard was more energy on top (probably distoriton) as symbols were boosted to the point of being harsh less clear. On low levels it was hard to spot. Everything changed when you wanted to listen much louder. Than highs were just too much.
1. What program do you use for downsampling? I use audacity myself. Used it a few times for this, but I don't dare sample down all my high rez files to free up space on my hard drive.
ReplyDelete2. Does high-res even make sense if our hearing is at best up to 14 kHz?
Hi Stephan,
DeleteI've been using iZotope RX for years now for resampling which has been my "gold standard" for quality. I really like its options around filter steepness (strongest looks like the one I applied for the undecoded MQA above) and variations along with control for cutoff frequency and also pre-ringing adjustments for intermediate phase options (not just linear) if I need - mainly for listening tests. Not cheap but I saw it as a lifetime purchase a few years back. ;-)
Likewise, iZotope's MBIT+ algorithm for dithering to 16-bit is excellent.
No, I don't think hi-res makes sense for most people (even if one is younger and fortunate enough to have hearing to 18kHz). As a perfectionist audiophile, I'm happy to own some of my favourite recordings as 24/96, but that's the most I need. 16/48 has been my "standard" otherwise for the last few years. For the record, these days at age 51, my limit is around 15-16kHz, depending on the day. ;-)
I apply this same logic to music streaming services; I have no need for 24/96 or 24/192 streams. If a streaming service can get me 16/44.1 or 48kHz lossless, and also lossy multichannel/Atmos at a good bitrate like say 768kbps E-AC3 as per Apple Music, I'd be plenty happy. For example, if Tidal were to dump MQA for just a single lossless (+ lossy multichannel) tier like that at a competitive price, I'd be happy to support them especially since I use Roon!
MQA is not something I get excited about and haven't pursued it. Patricia Barber is someone I have and followed for years. I have the digital download Clique DXD 24/352.8 and its a lovely, clean, accurate recording, but you know what, its in the camp of a so called audiophile recording. The actual music is not exciting and somewhat boring. I never play it out of preference over her early stuff, her previous album is similar. I guess she's played out with new ideas but she sure has always made quality recordings something others should aspire to. The lowest being DR10 with many at DR15. Her live at the Green Mill recordings worth a listen.
ReplyDeleteThanks Robocop,
DeleteI'll check out the live Green Mill recordings! I was in Chicago a number of years back and should have made a trip out there one of the evenings.
Yeah, as usual, we'll all have different opinions and preferences on the music itself. For sure, the quality of the sound is definitely that of an "audiophile recording" - I guess that's why Fremer rated it as "11" for Sound (for the lossy, lower resolution, decoded MQA-CD sadly). ;-)
Interesting comparison of different versions of the album here
ReplyDeletehttps://magicvinyldigital.net/2021/08/06/patricia-barber-clique-review-dxd-2-0-dxd-5-1/#Part4
Wow, heck of a lot of work! Looks like that guy's the biggest Clique! fan ever to do something like that. ;-)
DeleteLooking at the graphs, I wonder if Audirvana might decode MQA Core a little differently than Roon. Hard to know since I can't tell exactly how the data capture was done on that page.
Regardless, the result still shows a deviation between the MQA decode with the DXD below 20kHz - clearly the decoded MQA-CD is still not "the master".
Interesting, thanks for the link...
DeleteJust had a look at the MJ Thriller 2022 remaster article. Would be interesting to hear the Sony 360 Reality Audio multichannel version!
Try the Apple music Dolby Atmos MCh of Thriller. Sounds great using my Apple TV box connected to an AVR via HDMI.
DeleteWell magicvinyldigital is not immune to woo. He writes nonsense like this: "An alternative to SACD is to have a converter that allows you to convert the PCM 24/192 to DSD256 to find a similar sound" I don't think he quite knows what he's talking about.
DeleteThanks for the Apple Music suggestion Jacob, didn't know they also had the Atmos MCh version out there.
DeleteHey Steven, yeah, don't know why I would want to take the 24/192 --> DSD256 for sound similar to DSD64/SACD... Just go direct from 24/192 --> DSD64 using a decent 5th order modulator and voilà! As good as SACD! ;-)
Instructive article as always! I also like Patricia Barber’s recordings. Too bad they are not available in Atmos (at least not from Amazon Music).
ReplyDeleteThe MQA battle seems to be reviving again, as can be read on various forums, being pushed on Stereophile for this outrageously expensive DAC…Pity!
By the way, I was intrigued when seeing the spectrum comparison in the article: what could be in this 20-40KHz section that I can’t hear but seems not to be noise?
Since I don’t have access to that specific file, I tried something wiith a 2L 24/192 download of a string quartet. I hi-pass filtered it to keep the >20KHz content, then reduced that in frequency to bring it down to hearing range. I got a sort of cricket-size string quartet that strictly followed the original piece, showing that indeed there was a lot of ultrasonics generated by the string instruments, not mere noise.
We humans (especially us, older ones) are missing a lot… ;-)
Good eyes Gilles,
DeleteI don't know what's in his spectrogram either. The FFT certainly doesn't show much up there. Wondering if it's the color scaling used although that would be really weird. I'll have a peek at my conversion data and what it looks like in Audition later maybe.
Just a small nitpick, but you first quote him saying that DXD (which actually is decoded MQA) sounds better than DSD layer, and then you conclude in 3. that he can't hear the difference between them. Clearly he thinks that he can ⃰ . The most that you can say is that he can't judge what's closer to the master.
ReplyDelete*) Of course without some ABX test we can't say if he can hear any difference, but somehow I have the impression it's not an option :)
Hi Dan,
DeletePoint 3 is to basically suggest that since Fremer heard the SACD version, that's basically the DXD hi-res version (or at least very close), yet he was not able to recognize that as the superior version compared to the MQA-CD decode!
Indeed, whether he actually is able to hear the difference or simply "thinks" he could hear a change is up for debate. A nice ABX test would help. But I doubt he'll be up for the test! ;-)
Mikey F. is undoubtedly clueless about digital, but to be fair about " the impossibility of the 16/44.1 data to ever fully include all the nuances of an original "384/24" (DXD is actually 352.8kHz) file", those missing nuances are in all likelihood inaudible anyway.
ReplyDeleteYes, I don't doubt that either, Steven. Which is why in point 3, I added that little "not that I expect him to of course" comment with smiley face.
DeleteIntellectually, we can appreciate that 16/44.1 is simply unable to have the resolution of 24/352.8... And that should be obvious for any experienced audiophile writer who understands the way the technology works. No way we could compress the data to incorporate all that detail "losslessly", much less still maintain compatibility with standard CD playback!
That Fremer would seriously write in a way that suggests he actually believes this can be done after 50+ years in audio is IMO shocking! Almost as shocking as someone in NASA believing the sun revolved around the Earth, or me trying to sell you guys Nordost cables with a straight face. ;-)
Also, you could plot the difference in dB vs frequency between the reference (the DXD file or SACD) and either of the MQA versions, using a frequency scan of the whole or part of a track, using, say, a Blackmann-Harris window and high FFT sampling. Adobe Audition has a Frequency Analysis tool for generating such data. The plot visually indicates whether the difference is likely to be audible (as even 1 dB in the most sensitive aural range would be , and I suspect this is here).
ReplyDelete(actually as these differences are fairly high frequency, a low FFT size would work too)
DeleteYeah Steven that can be done. Like a null test between the MQA-CD +/- decode and DXD original can be done either as a spectrograph or FFT could be interesting.
DeleteHello Archimago, you surprise me that you only show measurements of the Roon Core MQA decoder. You are aware that this MQA core decoding is limited to 24/96 1st MQA unfolds, just like the Tidal and Audirvana core decoders are. Every MQA certfied DAC will perform 2nd and even 3rd unfolds which will show the full spectrum of the original DXD master. Spectrum analysis on MQA should be done by resampling the final analogue output.
ReplyDeleteGreetings Peter, nice seeing you around.
Delete1. There is no "3rd unfold" as far as I can tell. Show me what that is in an MQA-certified DAC, and what it does?
2. This is MQA-CD, the 16-bit version. As you can see, there is nothing meaningful in the decoded data even from 22.05 to 44.1kHz! So where do you think material is going to come from for anything up to 176.4kHz?
3. "Spectrum analysis on MQA should be done by resampling the final analogue output." Please, by all means show me the analogue output from a DAC that handles MQA-CD decoding and show me the high frequency content.
Thanks for the suggestion Peter,
DeleteAs you can see, the analogue output from an MQA DAC does not unfold to the full DXD master in the latest Addendum.
Peter, again, there is no "third unfold". There's a first unfold to either 88 or 96k. At that point the file is fully uncompressed. What MQA calls the second and third unfold is simply upsampling to 4X or 8X rates from 2X rates. Their "unfolding" terminology is simply deceptive. Period.
DeleteThe fact that you still don't understand this after years of daily posting about MQA is astounding. If you can't read the technical material and understand it, you shouldn't be posting on the subject.
Quote from Stereophile article MQA Tested - Into the fold part 2, writtten by Jim Austin: "Audio origami comprises two distinct processes; MQA calls them folds. The first fold, which they call encapsulation, folds data from four times the baseband rate (176.4 or 192kHz sampling), or higher, down to half those rates. Very high-rate files—think DXD—may require two encapsulation folds. " https://www.stereophile.com/content/mqa-tested-part-2-fold
DeleteThe second fold, called type-L, folds data from 2x to 1x—from 96 to 48kHz, or from 88.2 to 44.1kHz. This is the range where most of the ultrasonic musical information is, so it's important to keep these data intact.
Thanks Archimago, for publishing the measurements of the analogue output of the Sabaj A20D DAC. Although the reconstruction does not perfectly matches the DXD frequency curve, it clearly show that there is a frequency signal present up to 120 kHz, which 'should not' be present if there would be only 1 unfold. Upsampling a normal CD will show a cut-off at 22 kHz, but the MQA-CD of a DXD master unfolds 3 times. I have to admit that I don't like the 'bump' between the 30 - 55 kHz region though.
DeleteA you can see Peter,
DeleteAll that stuff out to 120kHz is actually artifact of the decoding process, not accurate content correlated with the master DXD material.
The Core decode tries to re-create content up to 44.1kHz. Stuff from 44.1-88.2kHz is just the distorted "image" of the lower material, and then from 88.2-132.3kHz is a second "image".
This is all the result of the fact that they used weak, low quality digital filtering in the "rendering" process which correlates with your concept of the 2nd and "3rd" unfolds. It's actually nothing more than just allowing distortions to exist in the ultrasonic frequencies, like how NOS DACs don't even bother filtering so there's all this stuff up there.
The MQA decode system is only at best interesting in the "Core" decode with 24-bit material with the lossy reconstruction to 44.1/48kHz. With 16-bit stuff like this, as you can see, it's not that good.
Regarding the comment from Stereophile:
Delete"Audio origami comprises two distinct processes; MQA calls them folds. The first fold, which they call encapsulation, folds data from four times the baseband rate (176.4 or 192kHz sampling), or higher, down to half those rates. Very high-rate files—think DXD—may require two encapsulation folds. "
https://www.stereophile.com/content/mqa-tested-part-2-fold
The second fold, called type-L, folds data from 2x to 1x—from 96 to 48kHz, or from 88.2 to 44.1kHz. This is the range where most of the ultrasonic musical information is, so it's important to keep these data intact.
Yes, it would be nice to keep some of that ultrasonic information intact for hi-res material. Again, realize that with 24-bit MQA, they're able to reconstruct some of that in a lossy fashion up to 44.1/48kHz in the Core decode but of course at the expense of the audible frequencies with ~16-bit resolution. There's no "free lunch".
I think Jim Austin is another audiophile magazine writer who needs to double check what he's repeating from the MQA company, and convey claims truthfully to audiophiles!
I disagree, but that will not surprise you. MQA unfolds are well described and we've seen many examples of them at 2L and other labels.
DeletePeter, think freshly pressed orange juice and juice made from concentrate. The latter will resemble the former, but it will never be identical. Your choice.
DeleteJacoacci, Quote: " FLAC is a lossless file format, a container for audio data. MQA is an advanced method for coding audio contents. MQA is normally delivered (losslessly) in a FLAC container from the music label. PCM is another type of audio that can be delivered by FLAC. Suggesting FLAC is better than MQA is like saying ‘bottles are better than wine’!"
DeleteIn that case PCM would be the (lossless) freshly squeezed juice and MQA the (lossy) concentrate. Both are transported in a (lossless) FLAC bottle.
DeletePeter, I'm impressed by your ability to counter any argument with a quote that is superficially impressive but in its essence total nonsense. Are you by any chance a troll?
Only a troll would aks such a question dear Jacobacci. If everyone who disagrees with you opinion needs a troll qualification, be my guest. No.. for me, MQA is a true gamechanger and I am learning a lot of the science behind it. With digital audio it is all about time and there are many factors in the audio chain which are able to improve the time resolution. I recently purchased the Innuos Phoenix USB reclocker and it's impact is huge! Just great this hobby :-)
DeleteThis is the kind of information from the MQA AES papers which are fascinating to read.It shows how 'oversized' a 24 bit FLAC container can be compared to the musical content inside it: " An analysis of 100,000 commercial 24-bit recordings with sampling rates between 88.2 and 192kHz, Note that very few have noise at or below the 16th bit and the median is at 11.5 bits. The noise-floor defines the bottom end of the dynamic range." https://bobtalks.co.uk/wp-content/uploads/2021/05/moisefloors.png
DeleteHey there Peter,
DeleteI agree that hi-res audio including 24-bit recordings are "oversized". Many hi-res recordings are clearly not needing the extra resolution. And indeed, MQA is a way to compress that down. (Or just downsample as I do with most of this stuff.)
The problem with MQA remains though:
1. We have the bandwidth to stream 24-bits losslessly without problem these days (Amazon, Apple Music, Qobuz). No need for the consumer to buy MQA-capable equipment. No need for streaming sites to re-encode with MQA to "authenticate" the master data. Maybe in the early 2000-2010's MQA would have been beneficial, but not now. These days also with multichannel content becoming more popular, MQA has no special niche in audio streaming.
2. As witnessed with the objective data not just here but over the years, MQA clearly is not capable of reproducing the "exact" master in a lossless fashion. So while it could still sound good, and "perceptibly" like the original due to limitations of the recording, it's still not the untouched, exact source which perfectionist audiophiles would often prefer when playing the music on very high quality gear capable of true >16-bit, >44.1kHz reproduction.
As you said above:
"I have to admit that I don't like the 'bump' between the 30 - 55 kHz region though.
That's what MQA-CD looks like, and that's what happened to Clique!, a recording that was meant to be a modern uncompromised high-resolution album. This is simply the nature of MQA and how it works, especially at 16-bits resolution regardless of whatever superlatives the company might claim.
Hey Peter, are you brand new at this? Otherwise why do you think the "FLAC container" matters at all, and why are you surprised the learn that many 'hi rez' DSD and PCM offerings in fact contain content that would be perfectly comfortable in a 'lower rez; format? This is very old news to most of us
DeleteThis comment has been removed by the author.
DeleteHi Archimago, for me it's indeed all about the "perceptional sound quality" aspect of MQA and aothough I am aware that each .mqa.flac file is intrinsically different compared to the original PCM of FLAC master, it does not bother me at all that somewhere down the MQA encoding process, bits are thrown away, especially if we know that so may of the 16- or 24 bits do not contain any musical information at all. It fascinates me more and more that re-sampling with modified B-spline algorithm offers new possibilities like MQA describes. Another aspect which I was unaware of, is that dithering methods are used for many years to enhance the dynamic range of a plain 16 bit CD recording. This is for a non-educated newby like me just furn to discover and learn and it shows me that 24 bit or even 32 bit recordings are 'nice to use' in the studio, but not really requiered at home. Solving the damage caused by filter artefacts is, however, much more important and audible. But besides MQA, I also learned a lot about other aspects in digital audio which play an important role on the final sound quality. Jitter, powersupply, clock accuracy, induced noise, network influences like latency etc. it all adds up and if these factors of influence all have been improved and I still hear a difference between CD, HD and MQA, then I just feel a bit surprised and blessed at the same time, since somehow, my hearing is sensitive to all this and improving the percepyual resolution is the goal of every audiophile I guess. We listen to music, not to a file :-)
DeletePeter , 'FLAC master' is a pointless term. FLAC is simply one of several (FLAC, ALAP, APE, MLP) lossless compression schemes for a PCM master file. FLAC has no 'audio quality' of its own, it simply compresses the size of whatever is fed to it and decompresses when played. Moreover there is zero evidence that MQA 'solves' any filtering artefacts (nor is it guaranteed that you could hear such artefacts in the first place -- and if so, it should bother you no more than throwing bits away does). Whatever possible audible hits may occur, sighted comparison is not the way 'hear' them in a verifiable way. You may, in fact, not be hearing a difference at all between CD and HD (assuming the same mastering -- if the mastering EQ is different , you are hearing THAT, not the 'higher resolution'.). This all seems to be parts of audio science you have missed.
Delete(sighed *and* not level matched)
DeleteStevenS, I know the difference between PCM, WAV, CDA, FLAC and even compared FLAC uncompressed with WAV (and preferred WAV a few years ago) FLAC and ALAC have become the standard for lossless streaming and most highres download stores offer the FLAC version. I tested conversion of .mqa.flac to .mqa.wav and the diffierence are inaudible to my ears. Processing power of DAC chips improved much over the last 6 years. Before I understood anything of MQA or what happens in a DAC, I experimented with off-the-fly upsampling of 16/44.1 wav to 32/176.8 and my ears liked it, but intuitively I suspected that the DAC had less problems with such large files. I also compared Jriver, Jplay and playback from a hibernated laptop. All audible... so I was frustrated. These were my first experiments and then Tidal introduced MQA and that was a relief for me and still is. With streaming audio, there are many other factors of influence which influence the sound quality and resolution of your system. Avoiding RFI interference, assure clean power, reduce noise, improve USB output etc. all factors of influence and audible, just like MQA is.
DeletePeter, 1st unfold should show signal above 20kHz, which it clearly does not. What makes you believe 2nd or 3rd unfolds will show signal 1st unfold does not carry?
ReplyDeleteTo be sure that full MQA unfolding is achieved, it is worth to know this. I've seen measurements which show the frequency spectrum of 2nd unfolds, for examples this one: https://www.hifinews.com/content/laura-metcalf-matei-varga-first-day-mqa
ReplyDeleteYes Peter,
DeleteThat is what a 24/48 MQA file decode looks like. Notice that the content >48kHz is a mirror image, at lower levels of the stuff <48kHz. Guess why ;-). Are you sure that graphic shows accurate, actual content from the original 192kHz source?
Again, we're talking about 16/44.1 MQA-CD here, not even the 24-bit MQA so it's even more limited!
Peter simply doesn't understand what he's looking at, yet feels he should comment all over the internet on the technical aspects of MQA, and making false "corrections" to people like Arch, who actually do understand what they are seeing.
DeleteSorry Danny, I am not claiming that I fully understand how MQA works, but I do study their scientific papers and articles thoroughly. I encourage you to download these five 'open access' AES papers to understand more. https://www.aes.org/journal/online/JAES_V67/5/
DeleteInteresting, looks like they updated the paper from 2014 on "A Hierarchical Approach for Archiving and Distribution" with changes to the references. Moved away from animal "neuroscience" references for example?
DeleteAnyhow, doesn't change basically what the technique is doing...
I'd like to crowdfund a hearing and lie detector test for Mr Fremer. Actually, I'd pay for it myself.
ReplyDeleteLOL BDC,
DeleteI'm sure every man has his price so if the guaranteed take home pot is big enough for Fremer, it'll add to his retirement stash.
But seriously, I think we've had enough of his claims and jokes for more than a generation; he has offered enough hype, and sold enough goods for the Audiophile Industry. Time to pack it up and do some vinyl unboxing videos. Unless he thinks he's especially genetically endowed (he might), I'll take the odds that there aren't enough hair cells in the ears for high fidelity hardware reviews going forward.
While I can "respect the elders", I think it's very uncool for 75 year olds to be prancing around as if they have the abilities of 20 year olds.
Hi StevenS, yes, compared to other people on this blog, I am relatively new. It all started when I heard MQA for the 1st time about 7-8 years ago I guess and I since them I want to understand how it is possible that this end-to-end technology works. The more I read about it, the more I learn to think both in- and outside the PCM 'box'so to say. I've encountered serious sound quality problems with digital audio ever since the CD was introduced. MQA technology makes it clear that increasing resolution and reduce digital harshnes can be done by targetting the musical content of the PCM container and correct for the blur which is losslessly stored on the digital master. If I would not hear an improvement, I would not be interested at all in MQA.
ReplyDeleteI'm sure YOU hear an improvement. Others don't, or even think MQA sounds worse. If MQA did what it claims, virtually everyone would hear an improvement. But they don't.
DeleteBottom line, it's a slightly lossy compression system with proprietary filtering.
The compression is somewhat innovative, but the filtering is nothing out of the ordinary. Either you like it or you don't.
But the whole blur and deblurring claims are nonsense, as Archi and others have show.
And again, please stop referring to upsampling as "third unfold" and "fourth unfold". They don't exist. MQA made up marketing terms to hide that they really just upsampling everything after the 2X unfold.
Hi Danny, MQA is indeed downsampling 3x or even 4x the original PCM file, but alongside this, the HF frequencies are folded as well. The reconstruction you see in the analogue output is proof of that. IThe MQA-CD can easily be compared with the analogue output of a non-MQA DAC and if there is a non-MQA-CD 16 nit version available, you can play this on an MQA DAC with MQA On or OFF ( Mytek offers this ) and also compare. I am convinced that their technology works as they describe in the patents and papers, although I also realize very well that they are hding some valuable information to keep for themselves. This is usual business ethics, the development of MQA took at least a decade of research. But let's just enjoy what we prefer I would say. At the end we have the wealth to choose.
DeleteHi Peter, you find me confused. "MQA technology makes it clear that increasing resolution and reduce digital harshnes can be done by targetting the musical content of the PCM container and correct for the blur which is losslessly stored on the digital master."
Delete1.) There is "blur" losslessly stored on the digital master
2.) MQA corrects for that blur by targeting the musical content
3.) Thereby increases resolution and reduces "digital harshness"
I have questions:
- MQA is supposed to be as close to the original master as possible. Why does it correct for anything that is part of the orinal master then?
- What is "blur" supposed to mean?
- What is "digital harshness"?
- If MQA "targets" musical content, doesn't it change that content?
Hi Jherbert, these are excellent questions - thanks!
DeleteIndeed, what is that 'blur' what causes it and why is this an artefact and how is it even possible to deliver a better quality than the original master? MQA wants to reproduce the original sound in the studio itself by getting rid of the time-smear introduced by A/D and D/A filters on the mastertape. I am constantly on the lookout to discover more how this is being done. A while ago, someone tried to encode music in MQA via an online service, but also manipulated the original music content by hiding high-level signals inside it. This made a mess of the MQA ecoding and was used to 'expose' MQA is bad, but proved to be complety wrong. But afterwards, MQA replied to the incident on their website and this provided interesting new informatioln ( well.. for me at least ..) It explains a bit more in detail how MQA encoding works:
Quote "MQA is not a codec in the conventional sense. It takes account of the source (A/D and mastering) and playback (D/A converter). The conceptual target is analogue to analogue, with a temporal blur equivalent to a few meters of air and a noisefloor target of atmospheric ‘absolute zero’."
https://bobtalks.co.uk/a-deeper-look/appendix-4-mqa-encoding/#
But I guess it's time for me to stop responding on everything in Archimago's blog, since I have been accused for 'trolling' . My intentions are not as such, but people might feel annoyed anyways.
So thanks Archimago and evryone here for allowing me to say what I wanted to say. We alle are in the 1st place music lovers and there are already too many wars outside in the real world. Wish everyone a nice weekend!
Peter, MQA never precisely defines what "blur" is, and don't allow anyone to actually test what the encoder and decoder do. It's all clouded and obfuscated by claims that can't be clearly understood or tested by others.
DeleteIt's therefore impossible to test their claims.
They also continually refer to the MQA delivery file as "lossless" which is misleading. It's a lossless compression of the lossy MQA master source.
And again, if it's "deblurring" was so obviously better, virtually everyone would hear it and it would have taken the market by storm. Neither have happened.
But of course, no worries. MQA has virtually no market. It exists almost exclusively on the Tidal premium tier, which has a tiny (a fraction of 1%) of the market. No one else of any significance has adopted it, after about 8 years in the real world. The company loses money every year, and at some point in the not distant future, Bob's friends and family will tire of subsidizing it with more investment.
Then it will go belly up, and the Peter's of the world will be left with MQA DACs, but no MQA (unless they bought MQA CDs....)
I have no concern at all with regard to their busienss model and if MQA is making profit or loss. Everyone knows that music streaming services are not profitable as well, even Spotify isn't So I'm happy with the luxury of choice and listen to CD, HD, DSD, DXD and MQA. Both Highreasaudio and Prostudiomasters offer thousands of MQA downloads and the most recent list with MQA albums on Tidal is 264200 MQA albums and EP's, of which 157x 24/352.8, 5114x 24/192, 14442x 24/96, 1578x 24/88.2, 14443x 24/48, 29447x 24/44.1, 1714x 16/48, 19811x 16/44.1 MQA Studio (blue) and MQA (green) SONY, WARNER, UNIVERSAL, and all their sublabels, EMI, ECM, Columbia, RCA, Arista, Parlophone and literally hundreds of other labels ( all in the list) Each TOP 100 single release is > 90% available in MQA . So no worries at all, MQA is here to stay :-)
DeleteI don't think I'd be so optimistic Peter,
DeleteThese are simply digital files that can be deleted and replaced by standard PCM at the stroke of a few keys or a batch process.
And Tidal has <5% of the streaming market, the only one that serves MQA. Not exactly an entrenched format...
"75+ years old"
ReplyDeleteBwahahahaha... I'm dying. Arch, you buried the lede bro.
On the other hand, it does suddenly make sense now. Almost anyone any amount younger would understand the technologies well enough to recognize they were making a fool of themself. You have to be old enough have made up your mind when CD's were first introduced _and_ be so old that any further evidence wasn't going to change it, which by now (2023) is very old, indeed. If you read Fremer with the frame of reference that he hates red book compact discs, everything else he writes naturally follows. It's still wrong, of course, but at least it's predictable.
Danny,
ReplyDelete# MQA Studio Master downloads on Highresaudio: 2690 https://www.highresaudio.com/en/search/?format=mqa&sort=-releaseDate
# MQA Studio Master downloads on Prostudiomasters: 6600
https://www.prostudiomasters.com/featured/format/mqa
But meanwhile, over hundreds of thousand albums have been MQA encoded and will continue to be released.
But let's stop this little 'format war' and enjoy MUSIC - have a nice weekend :-)
Hey Peter,
DeleteYeah, there are thousands of titles in MQA already. But even 6600 titles on ProStudioMasters isn't that impressive as file downloads when there were 5000 physical HDCD titles produced at one point back in the early 2000's!
Formats come and go. Some are winners and some are losers... By the end of this year, we'll be 9 years since MQA's announcement which is a long time already!
Ukraine vs. Russia is a real war with very serious consequences to millions. All that's happening here is a bit of "price discovery" as audiophile customers assess the value and quality of what we're being offered since mainstream audiophile magazine writers seem to have avoided independent testing and publishing facts.
I think many of us have discovered concerns about what MQA does to the audio signal. Sure, I can enjoy music encoded in MQA; just like I can enjoy high bitrate MP3 or AAC or Bluetooth LDAC encoding... I just wouldn't recommend MQA or those other encoding systems when an audiophile wants the absolute best playback quality listening in their treated, low ambient noise room with 5+ figure speakers and electronics though. ;-)
And since there are multiple alternative non-MQA hi-res streaming services, I don't think audiophiles who might want to listen to their streams with the room/equipment above should use Tidal and spend money on MQA decoding.
More a "tempest in the audiophile teapot" among those interested in the best-quality audio niche than an actual war. Most music listeners probably don't care (hence Spotify is still way more popular among my kids and their friends). I think MQA made a serious miscalculation in targeting audiophile consumers and using the audiophile press in their push for this. If they had just been more honest/modest and sold it as a higher-quality codec than the typical AAC/MP3/OggVorbis stream back in 2014, they might have had a better head start in the streaming world recognizing at some point full hi-res lossless would be coming anyways. Oh well.
Indeed, enjoy the music.
Hi Archimago, if I did not hear am improvement even if MQA would show more proof, then I would not be interested in it. We've read the reviews from mastering engineers who compared their own masters with the MQA version on Stereophile, Absolute Sound and more. All that counts is that it works fine for me and others and the higher the resolution of the system you are listening to, the easier it gets to hear the difference, but that's a personal obeservation and if others don't share this experience and prefer the original instead of the MQA version, that's just fine.
ReplyDeleteListen without prejudice is all I say to my friends who never heard it. Just fun!
I agree Peter,
DeleteUltimately we all have to just have a listen and decide with our own ears/brains, in our rooms/headphones and systems. Some of us might have preferences one way or another... Total freedom to choose.
Yes Allan,
ReplyDelete75+ revolutions around the sun is a good amount of mileage. I try not to emphasize that at the start simply because it might seem rude but nonetheless, it is extremely important as you say when it comes to the mindset and some expectations that might come of it.
His mind seems to have crystallized around an opinion of Red Book a long time ago. And "core" beliefs like that will have consequences when it comes to believing in other things, and his personal philosophy around audio tech!