An Open Letter To Naxos
Dear Jeff,
I think you visited my display last January in the High-Resolution audio TechZone room at the Venetian. It was good to see you. I’m still advocating for honesty, clarity, provenance, and openness in the production and distribution of high-resolution audio. As a member of the CEA audio board, The Recording Academy®, and the Audio Engineering Society, I’ve been involved in the introduction and promotion of high-resolution audio since the beginning. As you know, I’ve produced over 75 new HR recordings using state-of-the-art techniques and released them on DVD-Audio/Video and Blu-ray discs as well as digital downloads through iTrax. My approach captures live performances using PCM at 96 kHz/24-bits…real high-resolution according to my definition.
I received Sarah Sakamoto’s email newsletter of June 2, 2015 with the subject, “Digital Newsletter: HD Audio & Upsampling” and was very interested in the information presented in the newsletter. I applaud your efforts to ensure that all labels understand what high-resolution audio is AND what it isn’t. The use of up-sampling to “juice” a lower resolution recording to higher specifications threatens to diminish the enthusiasm audiophiles have for this emerging market segment and tear at the reputation of any organization that permits “hyperbole” in its marketing. But it doesn’t stop with up-sampling.
In the interest of making sure that accurate information is communicated to all of your distributed labels, I wrote an extensive email to your digital content manager concerning the large number of technical mistakes included in Sarah’s newsletter. I believe that it’s critical to communicate a unified and technically sound message whenever we address companies in the high-resolution audio arena. Here’s a brief summary of the six points that I pointed out to the member of your staff.
1. Under the Digital Audio subdivision – The diagram of how digital sampling works is wrong. There are no “stair steps” in the process.
2. Under the subsection “Further information about Digital Audio”, the statement “For example, a recording captured at 96 khz [spelling [kHz”] or 96,000 slices has 250 times the audio resolution of standard definition audio. More samples equals more resolution.” This statement is not true. The “audio resolution” is not 250 times greater when we sample at 96 kHz rather than 44.1 kHz. It adds another octave and improves noise levels.
3. The subsection “Standard Definition and High Definition Differences” should refer to Standard-Resolution and High-Resolution to be consistent with industry norms. It is correct to establish the Standard Resolution at CD quality and 44.1 kHz.16-bits. But saying that 20 or 24-bits then moves the quality to High-Resolution is incorrect. There are almost zero recordings that measure better than 16-bits of dynamic range. 24-bits is terrific for recording engineers but means almost nothing on the delivery side of the equation. High-resolution audio starts at 88.2 kHz/24-bit. Despite the definition issued by the DEG, CEA, NARAS and the major labels, high-res audio requires content and hardware to use 88.2/96 kHz and 24-bits. Their definition includes everything ever recorded, which obviously makes no sense.
4. In the High Definition subsection – “Any sample rate that is used from 44.1 khz – 192 khz which also contains a 24 bit depth can be used for HD audio distribution.” This is incorrect. You need higher sample rates…minimum of 88.2.
5. I like the colored chart…but it needs a little tweaking. Vinyl LPs do not have the fidelity of CDs. The dynamic range of a piece of vinyl is less than 10-12 bits of equivalent PCM…50-70 dB. Move the sample rate line to the right…48 kHz needs to go in the standard definition category. Your recommended specifications should be the minimum standard.
6. The key concept is the fidelity of the source recording and the delivery specifications. I think you understand this quite well. The Music Scope plot is good but you might want to use a track with high-resolution dynamics. The orange area of the plot shows a very narrow dynamic range.
Your representative responded politely and explained that things needed to be kept simple for the “less audio savvy labels” on the roster.
That would have closed the issue for me until I happened across a piece on The Absolute Sound website in the Download Roundup section dated May 19. (Click here) Andrew Quint’s piece is very positive and closes with a discussion of the 2xHD remastering and the various formats that one can acquire. This led me to “Cantate Domino” product page at ProStudioMasters where I saw all of the options available for this 1976 analog stereo recording priced from $17 – $30.
As a staunch advocate of properly identifying real “high-resolution audio” and knowing that this recording was made 40 years ago with standard-resolution analog tape (and not even professional grade equipment), I decided to purchase the product at 96 kHz/24-bits in AIFF format. Being experienced and knowledgeable about analog recordings from that period, I did a rigorous analysis of the “Cantique de Noel” track and found it doesn’t meet the high-resolution audio specs, which came as no surprise. It’s what we would expect…a good analog recording from 40 years ago. I’d be happy to send you the spectrograph and explain why it should not be considered high-resolution. In short, the dynamic range of the “Cantate Domino” doesn’t even reach CD specs and the frequency response just barely eclipses CDs as well. The guys at 2xHD are misrepresenting what they are capable of doing…even their logo is misleading.
All analog recordings should not be identified as high-resolution because of the shortcomings of the format. The large catalog of analog masters can be masterfully remastered and placed into high-resolution but buckets by 2xHD or HDTT (High Definition Tape Transfers) or by myself…but that doesn’t elevate the fidelity to current high-resolution specifications. Lumping them with new high-resolution tracks confuses people, drives complaints, and will ultimately diminish the market for your real High-Res titles…and mine.
The recently issued newsletter was a step in the right direction in spite of the mistakes contained within. But it missed the whole topic of provenance and analog masters. There needs to be a follow up that accurately describes the transfer of old standard-resolution audio (analog tapes, vinyl LPs, etc) into new higher resolution files. These can be marketed as “the highest quality transfer of the master tapes”. I refer to these as “master source quality”. But there is no additional fidelity in the 192 kHz version over the 96 kHz file.
Naxos has been a leader in the high-resolution market. Your labeling should serve as a model for the rest of the industry. However, you risk additional push back by associating the work of 2xHD or any analog sourced content with high-resolution. As a recognized authority on the topic, I would be happy to contribute my knowledge and experience in the preparation of a follow up document to set these issues straight.
All the best,
Mark Waldrep, Ph.D.
AIX Records
Well written. Keep up the good fight. Don’t allow the marketers to hype B.S. This type of misrepresentation will bring high definition audio into disrepute. Even for labels like Naxos with their lofty reputation, they need to recognize that taking the high road is good for them in the long run.
Maybe it’s time to re-introduce the SPARS code, or some variant of that convention. If the first letter in the code for any given recording is A (as in AAD, or ADD), then the recording should not qualify as a true High-Res title.
If it ain’t at the source in the recording chain you can’t make it so by upsampling. Great article, and one that I am passing on hoping to stop others from being duped out of their hard earned shekels. I haven’t bought one recording from PONO because of this.
While I agree with most of your letter to Naxos, I do see real value in the use of HD instead of Hi-Res by Naxos. I would agree with you if your definition of Hi-Res was being used for media, but we are now at a point where the High-Resolution audio marketing term includes all media ever recorded using any analog capture technology and digital recordings captured at a sample rate of 88.2 kHz or higher and at a stored sample value precision of 24 bits or higher. This makes as much sense as defining high performance vehicles as all human powered bicycles and Grand Prix racing cars with anything else just being called standard performance. For media the Hi-Def designation has become useless.
The term High Definition (HD) is commonly understood to mean “an increase in … resolution over a previously used standard” (source: Wikipedia). The Naxos use of HD is consistent with this definition. Naxos divides recordings for downloading or streaming into 3 categories: Lossy, Lossless, and HD. Lossy is all lossy formats regardless of the original recording format, Lossless is equivalent to CD distributions regardless of the original recording format, and HD is a distribution format where both the source and media exceed CD format specifications in at least one of the frequency and dynamic range specifications (i.e. >22 kHz or > 16 bits). In addition, Naxos provides some information about provenance. For example, when a media file has been produced using 2xHD, that fact is clearly indicated.
The Naxos website “www.classicsonlineHD.com” is the only website I have found that provides true HD (>44.1/16) streaming and extensive metadata search capabilities. Metadata search includes Quality, Performer, Conductor, Composer, Period, Country, Instrument, Year Released, Year Composed, Label, Genre and Title. The HD but not true Hi-Res recording of “Cantate Domino” in a 192/24 package can be streamed from this site, or it can be downloaded for $19.99 (a better price than ProStudiosMasters). Is there any chance of making AIX classical recordings available on this site?
We can’t possibly have High-Resolution Audio AND High-Definition in the audio world. Naxos does a good job…that’s why I’m so surprised that they got suckered down this analog 2xHD nonsense.
The Hi-Res specification for audio capture and editing (hardware and software) is meaningful, but Hi-Res as it is being applied to media distribution is useless for determining the actual quality/provenance of the recording. Its purpose appears to be more to mislead rather than to inform the consumer. Could it be that Naxos is intentionally avoiding use of the term Hi-Res for media distribution due to the misleading use of the term? At this point I would prefer a new term for end-to-end Hi-Res. My suggestions are “True-Res” or “True-Fidelity”. The use of the term high is part of the problem. It inevitably leads to the pursuit of something even higher without regard to whether higher is better.
Good points. Although, I don’t think Naxos’ use of HD is anything more than an innocent mistake. Take a look at the name of this website…I used to call it HD-Audio because I thought it made more sense (and I still believe it does!). But I lost that war just like I’ll probably lose the rest of them.
Yes, I was remembering where you started on the HD vs Hi-Res debate. HD is also in the name of the Naxos classical website (classicsonlineHD). This is yet another reason making it harder for them to drop the use of HD.
I would like to point out again that the Naxos website does support streaming that exceeds the Hi-Res specs. When I streamed the “Cantate Domino” recording, the one that really shouldn’t be considered Hi-Res, at 192/24 as a test, the measured bit rate exceeded 5Mbps.
I prefer HD-Audio but industry-wide they’ve opted for high-resolution. Naxos won’t change their site any more than I would change mine.
I saw a typo right after submitting my last comment. I meant to say “For media the Hi-Res designation has become useless.”
Is the Naxos document publicly available anywhere?
Increasing the sampling rate from 44.1 kHz to 96 kHz increases the time resolution by a factor of 96/44.1 = 2.18. In the frequency domain, that slightly more than doubles the bandwidth – adding the extra octave, as you point out. Moving from 16 to 24 bits increases the amplitude resolution by a factor of 2^8 = 256 – giving us the 20*log(256) = 48.16 dB increase in dynamic range. So, the statement that more samples yields higher resolution is correct – at least so far as the capabilities of the digital system are concerned –, but, of course, they should use the right numbers.
On the one hand, you point out that 16 bits of dynamic range is adequate for the delivery of almost all recordings, while on the other, you place vinyl below CD for having limited dynamic range. 70 dB of dynamic range is probably enough for most recordings, and the bandwidth available on a vinyl LP is greater than that on a CD, so putting them at parity seems an entirely reasonable position.
I applaud your continued efforts at establishing astringent definition for high-resolution audio, but it behooves you to be especially clear in your statements and reasoning, when trying to teach others.
Thanks Andrea…I guess I could make the Naxos document public, but I think I should get permission. Their digital audio manager is knowledgeable but lacks a focus on the topic at hand. Throwing out “250 times” the resolution of CD is simplistic and incorrect. The dynamic range is a tough one…because practically no products or music released these days gets close to needing 16-bits. Btu that doesn’t mean that we should lower our goals.
Mark, I agree with your and Andrea’s comments about the difference between 16 and 24 bits, and the usually insignificant value of 24-bits over 16-bits for distributed media, but I do have a question about a possible playback advantage of 24 over 16 bits. The sound/video processor (Anthem Statement D2v), I use for my listening and video by default converts all digital and analog input to 192/24. All sound processing, including room correction, then takes place at 192/24. If capturing and editing at 24 bits is better for processing with a DAW, then shouldn’t that also be true for the processing in a sound processor?
Conversion to 192 kHz/24-bits is more than enough to take care of the kind of processing that you need to do room correction. The benefits of longer words are clear when you have amplitude changes that are done in the digital domain. I wrote a post a long time ago about how the bits get shifted to the right and left. More bits keeps the integrity of the original word.
Yes, 24 bits, or at least some additional precision, is necessary for the processing, but is there any advantage in the source being 24 bits when the dynamic range of the source material is well below 93 dB. If converting the source from 16 bits to 24 bits immediately prior to processing, results in no truncation errors, etc. in the final processed output, then what’s the benefit of a 24-bit original source? For example, if I record at 44.1/24 to be safe, and then after verifying the dynamic range, save the master source at 44.1/16, how have I lost anything? This just seems to be very similar to lossless compression. Now, if I record at 96/16, and I have everything set right so that 16 bits covers the dynamic range, what is the difference between this 96/16 recording and a 96/24 recording or a conversion to 96/24, and why isn’t the 96/16 file Hi-Res? I just think that way too much importance is being placed on maintaining a 24-bit based dynamic range across the entire production path when almost all sources including classical music require a dynamic range of less than 60 dB. I’m not advocating recording at 16 bits, I just don’t think 24 bits always has to be maintained through the entire production path.Some of your recordings such as Bolero do require more than 16 bits, but more than 99.9% of commercial recordings do not.
You’re right. We need it to be safe during recording. And music rarely get close to using more than 75 dB…but it does happen. So why not maintain 24-bits? The marketing is misleading.