Digital System Wars, Part 2
This update covers three areas. First is a report on the most valid comparison and test we have yet encountered, to reveal the merits and demerits of DVD-A vs. DSD/SACD. Second is a further report on another DSD/SACD presentation we heard, with corroborating listening observations. Third is further probing analysis of the competing systems, amplifying upon material we have previously published. Before proceeding, we suggest that you read the following background material: the analysis of DSD in IAR issue # 76-79 (part of volume 7), and several previously written articles on this website, including the reviews of the dcs Purcell, the background analysis for the Audio Aero Capitole review, and the original Digital System Wars article. Below we also refer you to explanations and analysis published in IAR issues # 55, 58, and 61-62.
Turtle / Avalon Demo Comparison
We were privileged to be given a special private presentation, to evaluate the most valid demonstration we have yet encountered, for impartially comparing and revealing the true intrinsic sonics of 24/192 PCM (a DVD-A format) vs. DSD/SACD. This comparison and demonstration were initiated, engineered, and presented by Turtle Records (turtlerecords.com) and Avalon Acoustics (avalonacoustics.com). Their goal was simply to fully reveal and accurately compare these two competing formats, each system performing at its absolute best.
What Turtle Records did, according to our understanding, was to simply take a single feed (two channel) from a set of mikes and recording console, and then insert that same single feed simultaneously into the electronic professional master digital encoders of the two competing recording systems, thereupon preserving those digital encodings on a master recorder hard disk (as is common for master studio recording). What I then heard in the presentation was the very same digital signal from this very same master recording hard disk, played back through the professional master digital decoders of the two competing systems. Thus, I was hearing the master generation recording, which had been through only one encoding and decoding, using only the professional mastering electronics certified for these two competing systems.
Avalon Acoustics contributed their largest, most expensive speaker system to this demo, the Sentinel Reference. This speaker was awesomely revealing, very natural, and wide ranging, so it was an excellent choice for revealing differences, and without bias.
The two competing recordings could easily be synchronized, coming off the master hard disk, so that I could at will instantly switch between the 24/192 DVD-A vs. the DSD/SACD version of the same musical passage from the same mike feed.
This demo was the most accurate, meaningful comparison of the two competing systems we have yet encountered, and for several reasons. First, there were the fewest possible links in the chain, to possibly veil, color, or corrupt the comparison. Second, the electronics used were the official mastering electronics of each system, so we were hearing and comparing the maximum mastering quality capabilities of the two systems pitted against each other. Third, the signal path was parallel and equal (and optimal) for the two competing systems. Fourth, no questionable additional coding or processing or corruptions (e.g. watermarks) were introduced. Each system, in its purest, most professional form, was given the same mike feed and asked to do its best with it.
The average consumer cannot fare nearly as well when he attempts to compare the two competing formats via consumer media and playback devices. Essentially, there are too many variables in the chain to make a valid comparison likely, and the complexity of the chain is likely to introduce veiling or corruption of the sonic contrasts. For example, if a record label releases the same recording on both formats, chances are great that their master recording was done only on one format, and therefore the release in the other format is not pure nor truly representative, since it has already been corrupted by the vices of the format that the master recording was done in (essentially, on that other format version you'd be hearing not one format but rather both formats in series, with all their sonic vices, colorations, and limitations superimposed). Or, for example, one format of the dual release might contain watermarks executed with greater sonic intrusion than on the other format. Or, post processing steps in preparing the consumer version might have been executed differently for the two formats, or in such a way that one format responds better than another, or might simply be veiling the sound such that the true sonic differences between the two formats are less apparent or have been skewed.
Which brings up an important warning. Never compare the sound of the CD layer on an SACD with the DSD layer. Industry insiders tell us that the DSD/SACD consumer mastering facilities deliberately corrupt the sound of the 16/44.1 CD layer, say by truncating bits from the submitted master tape to make it sound like a crude 12 bit recording instead of its true 16 bit self -- the reason of course being to make the new DSD/SACD format sound artificially better than the classic CD format, so that you'll rush out to buy all the titles being reissued on DSD/SACD, even though you already own them on CD. Record labels were shocked to find that the CD layer put on the SACD by the mastering facility sounded much worse than their own straight CD (and much worse than the master tape they had submitted) of this same material.
Finally, you're not yet hearing the best sound possible from even the consumer versions of these new super format recordings, since we know for a fact that consumer DVD-A and SACD players are not yet as good as they could be. We know this because they sound worse playing ordinary CDs than our current state of the art CD players do. Thus, they must be doing some things still wrong (e.g. high jitter and noise), which of course would degrade all digital recordings they play, be those recordings CDs or one of the new super formats. We also know that the classic CD format itself, introduced in 1984 as perfect sound forever that therefore could not be improved upon, nevertheless has continued to continuously improve over the past 18 years, and that the very latest CD players can obtain superior sound from even the earliest CDs, superior even to CD players of two years ago (similarly, today's phono cartridges can obtain sound from even early vinyl that is far better than a few years ago).
Humans are blessed (and cursed) with an open ended learning curve. Super format consumer players will continue to evolve, just as surely as CD players and phono cartridges have, and next year they will play today's recordings much better than the best you can hear today.
The Turtle / Avalon demo was a glimpse into that wonderful future, since it revealed the potential of these two competing super formats far better than any comparison of consumer players with consumer recordings could.
The sound of these Turtle recordings was superb. It was extraordinarily revealing, clean, and wide ranging. The most interesting portion for us was a track featuring a solo violin. A single instrument making a complex sound, a sound that we're very familiar with. Incidentally, this recorded selection did not include examples of the difficult material that pushes DSD's buttons and makes it crash and burn, such as vocal sibilants and cymbals. So both systems were represented on their best behavior.
We thought going in that the sonic differences would be subtle. After all, these are both mastering systems, whose sole purpose is to document music accurately. And both these systems are the latest, best state of the art efforts at newly accurate super digital formats for super accuracy. If both are super accurate, then they must logically sound very similar.
But the sonic differences were not at all subtle. There were huge differences. It was literally night and day. This was shocking.
And it also meant big, important news.
Because if two systems sound different, then one at most can be right. And if one of these two systems were indeed perchance very accurate, then this huge sonic difference logically meant that the other system must be very inaccurate, very wrong.
That's big news. You need to be warned about systems that are very inaccurate. Every recording released in this inaccurate format should carry a parental warning label: "May mislead and subvert the minds and ears of all who listen." And the industry needs to be warned, lest they use this very inaccurate format to archive master recordings for posterity.
Back to the solo violin. Its sound was very different on the two competing super formats, as we instantly switched back and forth. This meant that at least one of the two competing systems was wrong. Perhaps both could be somewhat wrong, but in different ways (after all, no system is yet perfect).
Ideally, this A-B demo between the two competing systems would have been an A-B-R, with us sitting in the recording studio, hearing the mike feed from the live performing violin as a reference R, and comparing this reference to the two competing systems A and B, to see how each system differed, and which was more accurate overall. But, since our space-time worm didn't put us at the time and place of the original recording, this ephemeral experiment was lost forever. Fortunately, we could rely on the next best thing. Having been to many live concerts, we know intimately what a live solo violin should sound like. And, having recorded many concerts (with Neumann condenser mikes), we also know what a mike feed of a live violin should sound like.
By now, some of you are probably tired of reading about scientific validity, and are itching to read the results of the comparison. So we won't tease you any longer.
It's really very simple to summarize. It turned out that one of the competing digital systems sounded very accurate to the sound of a live violin, or live mike feed of one, whereas the other system sounded very different from the first system, and also very different from what a real violin sounds like. The envelope, please. The 24/192 DVD-A format was very right, and DSD/SACD was very wrong.
Let's discuss some details. First, the midranges. A live violin has a richly complex sound in the midranges. There's the complex sound of gut and rosin stroking steel, plus the complex sound of the sounding board resonating in sympathy with the many string vibrations. These complex sounds have a slight bite (after all, steel is being attacked), and even the wooden sounding board has what one could call a bright orangy timbre (echoing its color), due to the thinness of the wood. Also, when a steel string is first attacked by the bow, there's a striking transient sound which requires exact temporal coherence of the reproducing system if it is to sound real (all parts of this transient attack, and all its overtones and complex timbral noises, have to be in exact synchronism, for the attack to sound like the real thing).
The DVD-A rendition of this solo violin captured and reproduced all these above sonic aspects very well. It was amazingly vivid, just like a real violin playing right in front of me. Indeed, the sound was even slightly grating, just like a live violin right in front of me would sound.
In contrast, the DSD/SACD rendition did very poorly, and was very inaccurate. It did not sound at all like the violin was playing right in front of me. Indeed, it was so very different that it made the violin sound as if it were playing behind a thick velvet curtain, interposed between me and the violin (whereby I would be hearing virtually no direct sound from the violin at all, but only indirect sound that had leaked around the edges of the velvet curtain and had been reflected off the hall walls a few times). The direct sound of the violin was drastically altered, to become very indirect and diffuse, very rounded and gentle, with the natural hardness of attack completely gone, and a wholly different relationship created among the complex sonic components of each note, both harmonically and temporally. Much of the subtle and complex inner detail of a real violin's sound effectively disappeared, perhaps because it was simply averaged out (see discussion below), and/or because it was no longer exposed at the temporal beginning of each note, instead being postponed into the middle of each rounded note, where it became buried and obscured.
In sum, the DSD/SACD midranges provided a very comfortable, relaxing sonic portrait of a violin, easy on the ears, and fine for background cocktail or elevator music. One might even hear a live violin sounding something like this in the midranges if one were to sit way back at the rear of the ground floor (orchestra) section of a plush concert hall, filled with velvet curtains and people in heavy coats or furs, and thusly hear a violin from afar. But of course this violin on this recording was not miked this way, and therefore should not sound this way. It most certainly should not be altered to sound this way by any recording system that purports to be accurate, or to be archival, or indeed to have any use beyond giving listeners a heavy dose of sonic valium.
How did I know that the live violin was miked such that the DVD-A was the more accurate portrayal, not the DSD/SACD? Mightn't the DSD/SACD be correctly reproducing the actual miking of the live violin, whereas the DVD-A was in fact the inaccurate reproduction? These are good questions in principle, and would be good questions in practice if the sound of the two competing formats had been similar, one being (say) slightly brighter or harder than the other. But these questions can be easily answered here because the sonic differences were so huge, and because of the nature of the differences.
Recall that the DSD/SACD rendition sounds as if the violin were playing behind a thick velvet curtain, interposed between the microphone and the violin. If indeed the signal coming in from the microphone actually sounded like this, then there is no possible way on earth that the DVD-A system, working from the same microphone feed, could have fabricated the wealth of detail (and the right kind of detail) needed to lift that heavy velvet curtain, so that it could sound like it does, as if there had been nothing but clear air between the violin and the microphone. In other words, if there really had been a heavy velvet curtain between the violin and the microphone, then the sound would have become rounded and a lot of detail would have been obscured by this curtain, just as the DSD/SACD rendition sounds. But you can't recover information after it has been lost through the velvet curtain, and no system on earth, including the DVD-A system, could possibly create valid musical information after it had already been lost through the velvet curtain. Thus, the sonic portrait that sounds like nothing but clear air between the mike and the violin logically has to be more correct than the portrait that sounds like there was a velvet curtain interposed between the two.
Also, it's worth reiterating that we're familiar with what a live mike feed of a violin sounds like when the violin is not behind a velvet curtain, and that's the sound that the DVD-A rendition furnished. It's highly unlikely that a recording engineer would be so deliberately perverse as to make a reference recording for system evaluation by hanging a heavy velvet curtain between the violin and the microphone. Although, ironically, some recording engineers are now effectively doing that same perverse thing by choosing DSD/SACD to archive and distribute their music.
Let's now discuss the treble regions. Here DSD/SACD fared even worse. The violin's trebles from DVD-A sounded very much as they should from a live violin: fast, extended, clear, pure, articulate, focused, direct, and very coherent, both temporally and harmonically. Through DVD-A, the leading edge of a transient attack was just where it should be, at the leading edge. DVD-A was especially remarkable at sounding phase coherent, such that the spectral extremes of the music were all in the same phase as the rest of the music, pushing together or pulling together. Indeed, DVD-A sounded far better than CD and even better than many electronic audio amplifiers, at presenting treble information with excellent coherence, instead of the rounding, softening, or dulling we hear from components whose phase response rolls off (rotates) in the trebles.
In contrast to DVD-A, the trebles of DSD/SACD were virtually the opposite in their sonic qualities. They were indirect, diffuse, fuzzy, defocused, soft, veiled, slightly dulled, and very incoherent, both temporally and harmonically. Through DSD/SACD, the leading treble edge of a musical transient was not at the leading edge, but instead was buried later within the note, or disappeared altogether into a sea of soft, fuzzy noise. The violin's overtones became silvery sweet and soft, again as if the violin had been miked playing behind a velvet curtain.
The most remarkable strength of DVD-A, phase coherence, was one of the most remarkable weaknesses of DSD/SACD. The phase of treble information was not only somewhat rotated relative to the rest of the music, but actually sounded very much inverted, as if it were pulling or sucking in when the rest of the music was pushing or blowing out.
Phase inversion is highly audible once you have sensitized yourself to tuning in on it, and it is especially noticeable when a third of the music (the trebles) is inverted relative to two thirds of the music. This seeming inversion robs the trebles of all sense of forceful dynamics that real live music has. Because of this, DSD/SACD turns the trebles of music into fuzzy defocused bunny fur. The forceful attacks of the violinist's playing were completely emasculated and de-passionized, meaning that DSD/SACD actually committed two cardinal sins. It not only changed the sound of the violin itself, but also changed the interpretation by the artist, changing a tiger into a harmless bunny (safe for relaxing cocktail music or elevator background music). Consider violinists whose playing style is at opposite ends of the spectrum: the hard charging attacks of Jascha Heifetz or the astringency of Gideon Kremer at one extreme, and the feminine bunny soft silvery gentle stroking of Erica Morini at the other extreme. Well, DSD/SACD takes a recording by Heifetz or Kremer and alters it so much that it sounds as if Erica Morini is performing instead.
There's worse news yet. On many kinds of musical instruments, DSD/SACD actually changes the character of the natural treble sound so severely that it literally is no longer music. What happens is that, whenever these musical instruments create an interesting musical sound in the trebles, with interesting timbres and textures characteristic of that musical instrument, DSD/SACD literally discards that entire musical treble sound -- and in its place DSD/SACD substitutes a burst of white noise!! That's the ultimate chutzpah for any music reproduction system, to throw away valuable musical information and substitute white noise. Even your lowly MP3 players are not quite so rude. And DSD/SACD presents itself as an archival system to be entrusted with master recordings?
(Continued on page 40)