Dissecting Dynamic Range: Does It Matter?
The high-resolution challenge page had been pinged almost 1500 times and over 400 people have contacted me about downloading the files. Don’t worry if you haven’t yet downloaded the files, there’s still plenty of time to get in on the fun (yes, this is meant to be fun not a rigorous test or thorough survey!). I would encourage all participants to explore the option of using the ABX plugin for Foobar to do the evaluation. It elevates the rigor of the test a notch or two.
So what should we be listening for when experiencing these files? Is it the inclusion of the ultrasonic octave from 20-40 kHz that a real high-resolution file can capture? It’s highly doubtful that you or even a teenager would be able to identify any differences based on partials present in the highest range of the audio band — and beyond. I know my own hearing tops out at 15-17 kHz. And a really good set of audiophile speakers is unlikely to reproduce ultrasonic frequencies. Some headphones or speakers with supertweeters push the upper limit to around 40 kHz but if we can’t hear it, why bother? So why all the fuss about higher sample rates? The audiophile lexicon uses words like “air”, “transparency”, and “clarity” to describe recordings and systems that are exceptionally good at handling high frequency content. The wind chimes in the Laurence Juber “Mosaic” track (Tune_1_A) is a good example. Listen carefully at around 1:40. The inharmonic partials being produced by the wind chimes (made for a the tuning pins of a junked piano) are crystalline and clear — not mushy and crushed as they would be on vinyl or analog tape.
I’ve said before that increasing the sample rate of a PCM digital recording to 96 kHz more than covers all of the musical components coming from any instrument (especially acoustic instruments). It also pushes any of the dreaded digital “pre or post ringing” to near the Nyquist frequency (with 96 kHz that point is at 48 kHz – nobody is hearing that!) and insures that any “time smear” is relegated to the trash bin. There are very real technical reasons why moving the sample rate to 96 kHz makes sense. Going further is counterproductive and a waste of resources. We get 100% fidelity with a 96 kHz/24-bit PCM recording and that’s what we all want.
However, increased dynamic range in a recording is a very good thing and offers dramatically enhanced listening experiences. Moving from 16-bits to 24-bits increases the potential dynamic range of a recording from about 93 dB to over 120 dB (many authors get these numbers wrong because they don’t factor in dither which adds noise to the theoretical number of 96 dB). An article from about a month ago by my friend Jerry Del Colliano at AudiophileReview tweaked my interest when I read it and when I read the comments this morning. You can read it yourself by clicking here. Jerry is a publisher and not an audio engineer and thus can be excused for drifting into areas that he might best avoid. The focus of the article is that vinyl cannot match the dynamic range of digital — and that’s absolutely true. Only a hardcore digital denier would dispute that fact (surprisingly there are lot of comments from people that take that position). To believe that vinyl LPs are capable of capturing and reproducing the sound of live performance is crazy because the dynamic range of the very best vinyl LPs (direct to disc LPs) is about 60-72 dB. But Jerry a number factual errors in the piece.
“Perhaps your ears know best, but analog master tape, like what many of the most important recordings in the history of music were recorded on, can capture the 120 dB-ish range of an actual musical event.” This statement is not true. The best that analog tape — the multitrack 2″ machines that were used to make many of the classical albums from our past — can do with regards to dynamic range is about 72 dB NOT 120 dB! With each pass over the heads the oxide gets scraped away and is diminished, the high frequencies are attenuated, and the dynamic range reduced. That tape is then mixed to another analog tape — this time a two-channel master. The dynamic range is lowered again by 3 dB with each analog generation. The final mastered version is another copy so maintaining a dynamic range of 60 dB would be considered very good. But the mastering engineer uses dynamics compression and amplitude normalization to increase the apparent volume — at the cost of whatever dynamic range is left at his or her stage. The DR of an original analog master — probably the safety copy from which vinyl LPs were made — may have as little as 15-30 dB of dynamic range. In PCM digital terms this is less than 6 bits! That’s why I hammer at the myth of high-resolution regularly. The difference between what is possible and what is real is rather large.
And it doesn’t really matter whether you consume these recordings from vinyl LPs or CDs. The dynamic range of an actual performance isn’t maintained from recording sessions to distribution format. Yes, the potential of PCM digital to maintain every dynamic nuance present during the original performance is there. But it is almost never realized. The music business doesn’t want their releases to have real world dynamics. Jerry’s argument falls apart when you realize the realities of the music business model. He writes, “Obviously, you can’t get access to the VERY fragile and environmentally sensitive analog master tape of a recording, but what you buy music on current HD formats that are damn-near bit-perfect to that master tape. For $20 you get a meaningful recording with full 120 dB plus dynamics, no meaningful distortion from the playback source. The last line — where he says we’re getting “full 120 dB plus dynamics” — is false. But he’s right that the “high-resolution transfers” HDtracks and others offer represent the very best available digital version of whatever “master” tape was used during the digitization process. Could there be better versions in previous non-digital formats? Absolutely. Is it worth the elevated price to get a “high-resolution” download or stream — probably not. One reader buys old CDs — trusty 44.1 kHz/16-bit PCM — for $1 each at used record stores and loads them into his music server. The fidelity of the CDs are virtually indistinguishable from the high-resolution downloads at 25 times the cost.
What most writers — especially ones that have no experience engineering a record — get wrong is that the pathway from analog “masters” to “high-resolution” digital file doesn’t improve the fidelity of the original master — if done well, it maintains whatever fidelity (including the narrow dynamic range and limited frequency response) was present in the original master. This is the mantra of MQA supporters and the goal of its inventors — to minimize any sonic degradation due to the transfer from analog to digital. 96 kHz/24-bits can do this without any loss and without the overcharges associated with MQA’s licensing agreement. The 11,000 MQA albums on Tidal didn’t come from high-resolution sources so why all of the praise for a method that makes claims to produce high-resolution streams? It’s crazy. It’s why many of the high-end makers of DACs have decided to pass on MQA. They’re right.
Listen to the high-resolution challenge files with dynamic range in mind. The files from my catalog are unique in that they actually do maintain all of the dynamics associated with the original performance. When you experience the 16-bit file, the transients are quieter, the impact of a snare drum lessened, and the slap of Laurence playing his signature Martin guitar softer. That’s what dynamic range can really do and why real high-resolution audio has such great potential.
Mark,
I would really hope the outcome of all this is people voicing concern and questioning the shoddy standards dumped on us by the major players in the music recording world and to push anyone recording digital music to aim higher. Even playing back a well recorded ‘HD’ recording on less than audiophile gear is going to benefit the listener as you explain by all the unwanted artefacts being pushed out of the way in the higher frequencies. It would seem that 96/24 just creates a better recording for people like me to buy regardless of whether or not we can detect the ultrasonics or not.
John, there is little incentive for the major labels and artists to adopt high-resolution standards. There is no additional revenue for them and a risk of having tracks that don’t “punch” as hard as the other heavily mastered tracks. We should just accept that we can get very good digital copies of our analog favorites (inexpensively from CDs) and the same fidelity from streaming services without a closed lossy system like MQA.
Mark,
“It’s why many of the high-end makers of DACs have decided to pass on MQA. They’re right.”
Also EXOGAL has recently dropped futher work on MQA for their DACs.
It was the right decision.
Regards
Mark, I have a question: What does the DR-meter on the dr.loudness-war.info site actually measure? I had a look at Laurence Juber’s “Guitar Noir” album and it only has an average value of 15. This is a lot by modern (loudness war) standards but shouldn’t it be much more if 24 bits are fully used? I own a CD by a German producer named Pole (CD 1) and it has a higher dynamic range and it’s electronic music (dub-techno-something), and the “Drums & Bells” CD by Brad Dutz and Chris Wabich and some other CDs have also higher DR-ratings. Am I missing something here?
I just visited the dr.loudness-war site and was pleased to see their opening remark – “Quite possibly one of the greatest recordings I have ever heard. The dynamic range on this is incredible and it sounds phenomenal.” I haven’t used their recommended tool to do a dynamic analysis but Adobe Audition shows a LUFS of -18.95, with a dynamic range of over 90 dB. You would be very hard pressed to find another album with this level of dynamics. The album you mention with higher DR score was done using purely electronic production methods…that means absolute digital silence vs their mastered peak. Not really a fair comparison.
Re. how DR value is computed, details are here: http://www.dynamicrange.de/sites/default/files/Measuring%20DR%20ENv3.pdf . Simplifying:
1. track is split into blocks of 3 seconds
2. 20% of the blocks with the highest average value (RMS) are picked
3. DR value is a difference between the highest sample and the average value (RMS) of those 20% blocks
So it is more like a crest factor rather than dynamic range.
Re. 16 vs 24 bit, I’m of the opinion that the only difference on the playback side is lower noise floor and nothing else. As per https://www.head-fi.org/threads/24bit-vs-16bit-the-myth-exploded.415361/
Thanks Daniel. I just read the article on 16 vs 24 bits and would agree with a lot of it. But his explanation has a number of incorrect statements, which I will discuss further in a new post. If we accepted the author’s reasoning, there would be no reason to go beyond 8 bits (or even less) because virtually all commercial music releases have less than 48 dB of dynamic range. When he writes, “So, 24bit does add more ‘resolution’ compared to 16bit but this added resolution doesn’t mean higher quality, it just means we can encode a larger dynamic range.” A larger dynamic range — matching the capabilities of human hearing without compromise — DOES MEAN HIGHER QUALITY! This sort of thinking would have us limit the fidelity of our systems to the individual listening constraints of our systems. I prefer to match the real world and let listeners strive to play it back — without compromise. However, it is true that recordings and systems with high-resolution, 24-bit capability are very rare. The run of the mill 24-bit downloads you get online don’t need 24-bits.
Well, all I can say is that for me “Mosaic” in 8 bits sounds exactly like 24 bits + background noise 🙂
I’m not sure I understand your fidelity assessment of “Mosaic”. The dynamic range of this track exceeds that of a CD, vinyl LP, and analog tape. 8-bits is like a heavily mastered commercial track.
I was just repeating what I said earlier but in different words, that to me the only effect of reduced bit-depth is background noise. “Mosaic” was just an example. I took 24 bit version, converted it to 8 bit with dithering and no additional processing and listened to both versions with instant switching. As I said, to me they sounded the same only that 8 bit version had additional noise. No surprises here, really. Maybe this noise is what others describe as quieter transients and lessened snare drum, I won’t argue that.
Also, at my usual listening volume levels I stop hearing things at around -72, -78 dB ( https://www.youtube.com/watch?v=e6Be-bCfkKw ). With over 15 dB to spare I don’t really worry about 16 bit not being enough for me.
Btw. “Reply” button doesn’t show any more at this level of comment thread. I’m not sure if that is intentional or some bug in the blog code.
I understand now…thanks for the explanation.
Agreed. MQA is an elaborate way to fold a high rez file into a lower rez bucket when high rez buckets are widely available. I don’t get it. When the audio writers claim that the process results in better sound than the master itself because it magically restores stuff that should have been in the master file and wasn’t my BS detector goes full scale. It’s similar to the analog mavens at the same publications saying the an LP taken from a digital master sounds better than the digital master. They insisted to me that inserting a cutting lathe, phono cartridge and a phono preamp into the digital playback chain increases fidelity. I forget the logic. Something about music having an affinity for spinning motors if memory serves. The fact that so many audio writers can’t see this for the nonsense it is makes much of the whole high end audio ideal seem suspect to me. If the goal is really to make it sound good to us then lets stop all the pretentious talk about purity and bring back tone controls, loudness contours, and EQ. My luxury AVR does all that to good effect. (and it sounds great playing your multichannel high rez files)
The audiophile press don’t have a choice when it comes to supporting the companies that advertise in their pages. And they feel an obligation to be kind to their revenue sources. I guess if I were also dependent on advertising, I might tilt that way too. Thankfully, I don’t have paid advertisers.
I completely agree with you. None of the formats make full use of their dynamic range due to choices in the studio. An example of this is that I found many albums on dr.loudness-war.info that have higher DR for the vinyl version than the CD or hi-res version! Clearly a choice by the producer since vinyl has the lowest potential DR. Why they do this I don’t know, but it may explain why some people think that vinyl sounds better despite all the other problems with it. It seems like some producers intentionally hobble their digital releases. Or maybe they just unknowingly raise the DR in the vinyl mastering process.
I’m hoping someone from the music industry can explain how/why this happens in the studio–vinyl releases with higher DR than digital releases.
As a former mastering engineer (Allman Brothers, Bad Company, Kiss etc), I can tell you why the records are created louder and louder. It’s to compete with the ever louder records from the other labels. It has been shown that louder records sell more. It’s that simple. When a label exec listens to everything else in the marketplace and their record sound less louder, they aren’t happy. I was asked 5 times to make the Bad Company “Merchants of Cool” record louder…and I had Paul Rogers (the lead singer) in the studio! He approved the first round.
Yes, I’m aware of the loudness war.
What I’m asking is, why would the vinyl version have better dynamic range than the digital version of THE SAME ALBUM? It’s clearly a choice, but why?
It turns out that vinyl LP mastering often uses a different master AND includes a different process. So there is not way to adequately compare a commercially released vinyl LP and a CD of the same album — they likely went through entirely different production paths. If I were to take one of my high-resolution, unmastered files and continue the down conversion with a transfer to vinyl, it would not have “better dynamic range” than the original or even the CD. There are commercial realities that make the ones you experience different.
I’ve been following the newsletters here for a while. Also was at the very beginning of the MQA inception. Ive read a lot of bashing about the format and I’m confused. Correct me if I’m wrong, but wasn’t the reason MQA was even devised in the first place was to give streaming services the ability of streaming higher than CD quality music to its listeners without using anymore internet bandwidth than streaming CD quality. I’ve been a subscriber to Tidal since it started, and was there when they started offering MQA quality music for the same cost as their CD quality service. So if I’m right this format was never created to replace High-Res downloads, but to give a format that would allow better than CD quality music to be streamed without any added cost to the end user. Streamed being the key word here. Now as to where they get their files, that I have no idea, but neither do I have the guarantee that all albums available for purchase and download at the other High-Res download sites are actually 100% from High-Res masters. I just don’t understand where the issue is anyway. Tidal never charged me a dime more for the added benifit of MQA. Also they make no claims as to whether it is from High-Res files or CD quality files, just that it’s MQA. It’s been suggested that other sources of High-Res music are just CD’s ripped to a higher bit depth to give the impression of higher than CD quality. MQA on the other hand, and only from what I’ve read from numerous other sources are folding their music from High-Res sources. Anyway as I see it, this hype about MQA is a non-issue. I think we’re getting so caught up in the quality of the music, we are loosing site of the reason it was invented in the first place, not as a High-Res replacement format, but as a convenience format for streaming services.
Tony, your comment warrants an entire article. But here’s the quick response…Bob Stuart and his company Meridian developed the MLP algorithm for losslessly compressing 6 channels of 96 kHz/24-bit into less than 10 MBps for the DVD-Audio format. When he licensed it to Dolby Labs, it became True HD and is commonly used in Blu-ray discs. But he kept the streaming rights, which became MQA. The pitch was initially as a revolutionary method for folding “high-resolution audio” into CD-sized bandwidth — perfect for streaming. He had enough foresight to see that streaming was quickly going to replace download and that he could pitch the services and labels on a clever way of delivering “high-resolution audio” on demand. He was not alone in this but managed to convince some well heeled investors to back his new venture. The MQA method was offered to the record labels as a way to keep their precious masters and resell new versions of albums that were already digitized at 192 kHz/24-bits and available to licensors like HDtracks and HighResAudio.
My arguments against MQA are based on a number of factors — and none is the cop out used by advocates that “it just sounds better!” But the primary reason — that many fail to recognize — is that the content being offered as “high-resolution” MQA streams didn’t originate as high-res audio. So how can their clever “origami” algorithm produce any meaningful results? Obviously, it can’t. The technical reasons why MQA fails are spelled out on a number of sites. It’s lossy, closed, the non-MQA file sounds worse than normal Redbook etc. But the real news is in the motivation behind the effect to make MQA ubiquitous — to the exclusion of normal, open, high-res PCM. The worldwide head of marketing told an audience a couple of weeks ago that their deal with labels requires all items in the catalogs to come out as MQA. This produces a small cha-chinge in the MQA wallet every time an MQA files is played and can potentially reverse the rather dire financial position of Meridian and MQA. Not surprisingly, the entire effort is about money and less about audio fidelity.
Once again you have nailed the argument on the head. Cut through the techno babble, so pervasive in the hi Rez marketplace. Most notably with journalist .make a over exaggerate problem, simple.
Hi Mark
I read a really interesting article quite some years ago in the now defunct Audio magazine which the authors through a combination of analyzing the spectrum of music and doing real world blind listening tests came to the conclusion which supports your position and that is ; that all that we need to reproduce true fidelity is a sample rate of somewhere around 82Khz and a little over an 18 bit sample rate. From my memory it seemed to be quite a comprehensive and rigorous sampling and testing regime.
The article was publish over two issues and was quite comprehensive in its detail, unfortunately I have since lost my issues in a fire , I had kept the article to show to digital naysayers of the day !.
Perhaps one of your readers has a copy of the article , I would love to re-read it . I believe it would have been published in or around 1998 or 1999.
I know Dan Lavry has written extensively about the minimum needs of PCM to handle all of our fidelity needs. For me 96 kHz/24-bits works perfectly.
Mark wrote: “When you experience the 16-bit file, the transients are quieter, the impact of a snare drum lessened, and the slap of Laurence playing his signature Martin guitar softer. That’s what dynamic range can really do and why real high-resolution audio has such great potential.”
How can this be? We’re talking the difference between ~ -96.3dB (or -91.7dB with dither?) and ~ -144.5dB (-137.6 dB with dither?); even through closed back headphones that’s a tall order, and through loudspeakers in your average ambient environment? Also, isn’t adding dither supposed to increase the perceived dynamic range?
Dave, you have listed the theoretical dynamic range limits of 16 and 24-bits not the actual SPL output from your speakers according to how loud your amplifiers are turned up. The difference between one file at full 24-bits and a downconverted version is the relative difference between the peaks and background AND this can be perceived even without turning up the volume. That’s why I mentioned to listen to the transients and slaps…they sound different.
Mark, your reply doesn’t make sense to me, maybe you could show this graphically at some stage. As far as I see it, if the peak was at -5dB the background would either be at -92dB (for 16 bit) or -138dB (for 24 bit). The relative difference for 16 bit would be 87dB and for 24 bit would be 133 dB. However, this is meaningless since the the real noise floor for speaker listening environment would be -30dB (say), giving a real delta dB of 25dB for both bit depths. Even through headphones, the analogue electronics would most likely have a S/N ratio of ~ -100dB (say), so again the delta would be the same at around 95dB.
So, how would it be at all possible to differentiate between the two bit depths in such a situation?
Dave, you’re working from the top down. Rather think from the bottom up. The transient peaks of the 24-bit samples produce more instantaneous amplitude than the 16-bit versions.
“More instantaneous” meaning greater rate of change, i.e., sharper peaks? I think this need further (graphical) explanation.
I agree that it would be nice to get more than the 5-6 dB dynamic range of pop music, but I’m not sure there is a need to reproduce SPL’s that take us well into hearing damage risk territory, which actually starts being a risk factor (depending on length of exposure) as SPL passes 85 dB. According to the website ‘Dangerous Decibels’, for each 3dB increase, the permissible exposure is cut in half, with the before-damage exposure at 120 dB SPL being about seven seconds according to their math.
Pam Mason is an audiologist that has given presentations at the last two Capital Audiofests; her presentation is worth checking out for those who can attend the show.
Too much SPL is indeed very dangerous. However, instantaneous transients of high volume occur all the time in life and should be modeled in our systems.
Some people might mistakenly think that LPs have better dynamics… because most of digitalized music dynamics is lost in overcompression. So in theory digital formats are superior, but in practice I very often find that digitalized vinyl rips have better sound quality because of overcompression of dynamics in digital formats.
And I have stopped mostly buying any music at all… If I find something in good quality, sure, but I am less and less inclined to look anymore, too much work for very rare rewards.
I would really like to hear your opinion on the Tonian Labs CDs “Drums & Bells” and the new one “Hang Around”. They were allegedly recorded in 16 bit and some people on different forums claim that these are the most natural sounding recordings they have ever heard regardless of the Format, and that it puts every hires to shame.
I haven’t heard these tracks but can imagine they sound amazing. It’s less about format and more about recording technique IMHO.