This substitution of white noise for music is more obvious on some musical instruments than others. For example, the delicate shimmerings of a gently struck cymbal should have a brassy metallic timbre (and they do on both CD and DVD-A). But DSD/SACD discards this natural metallic timbral sound, and in its place inserts a burst of white noise. Once you hear DSD/SACD pull this cheap parlor trick on the most obvious musical instruments, you'll become sensitized to the phenomenon, and you'll start noticing that it occurs to varying degrees on almost all musical instruments and vocals. And once you start hearing it, it will start bothering the heck out of you on virtually all DSD/SACD recordings.
As you may recall from our previously published observations, DSD/SACD does something even worse when music becomes rich in treble energy, as it does on vocal sibilants and when a cymbal is struck more forcefully. At these times DSD/SACD seems to crash and burn, probably from overload distortion in its necessarily aggressive averaging and noise shaping algorithms. The white noise, that was substituted for true musical information in the trebles, itself changes character, and instead of being a burst of clean white noise it sounds dirty and garbled and smeared, as if the white noise were undergoing modulation distortion. This makes the whole musical sound (the vocal sibilant or cymbal strike) sound distorted, dirty, grundgy, smeared, and garbled - since the midranges are now accompanied by distortion rather than by a burst of clean white noise in the trebles.
Both systems were very good in bass reproduction, so far as we could tell. The program material we heard in this demo was not geared toward comprehensive assessment of bass capabilities. In any case, the sonic differences are so huge throughout the midranges and trebles that they will be the deciding factor for every listener, and no one is going to choose one system over the other based on bass capability.
Was the DVD-A reproduction perfect? Not quite. Our chief observation was that there was a little glare, mostly in the upper midrange. However, this could actually be yet another feather in DVD-A's cap. How, you ask?
Here's how. We know independently that most solid state devices tend to produce sonic glare, especially in the upper midrange, and that IC chips are usually the worst offenders. We also know independently something about how this glare is caused, since we also know how it can be cured. In research conducted by our sister company, TRT (Tomorrow's Research Today), we found that a micro-thin coating (of a specially engineered material) applied to the surface of solid state devices could substantially cure this sonic glare, in all analog and A-D and D-A circuits. We also found that applying this micro-thin coating to purely digital chips in a purely digital circuit was important for curing glare that was audible when the signal going through this digital chip was eventually changed to analog and heard.
What's going on here? The theory is that this micro-thin coating substantially eliminates multiple internal reflections from the electromagnetic wave carrying the music signal, these multiple reflections being the culprit that causes the audible glare. The same principle is already widely recognized and used for camera lenses, where a micro-thin coating substantially eliminates multiple internal reflections of the very same electromagnetic waves (which we call light at those frequencies), these multiple reflections being the culprit that causes optical glare. Anyone may obtain this micro-thin coating material from TRT to use on all the solid state devices and chips in his audio chain (until a separate website is set up, email them at email@example.com).
The central lesson here is that even purely digital chips contribute sonic glare, and that it is the chips themselves which are adding the glare to the sound. The sonic glare is not the intrinsic fault of the digital coding system or digital format, but rather arises in the many solid state chips which these digital coding systems must use. And we know this because we can cure this sonic glare by treating the solid state chips themselves, in accordance with established anti-reflective coating theory.
This means that any glare which we heard from these digital systems is due at least in part to the many chips they employ, not to the digital format itself. We know independently that the DVD-A format was excellent at accurately revealing the sound of the violin, in its many sonic facets. And we also know independently that the DVD-A encoding and decoding electronics employ many solid state devices in the signal path, devices that we independently know cause upper midrange glare (unless treated with a suitable anti-reflective coating). Put these independent pieces of knowledge together, and it follows that the DVD-A format actually deserves credit, credit for being so accurately revealing that it not only accurately reveals the manifold details of a violin's sound, but also accurately reveals the upper midrange glare of the many solid state devices in its signal path.
What about DSD/SACD? It has no glare, since it rounds, softens, and liquefies all the music signal it processes. But we can surmise that there are very roughly as many solid state devices in the DSD/SACD signal path as there are in the DVD-A signal path, causing glare for DSD/SACD just as surely as they do for DVD-A. This means that it is actually a demerit in DSD/SACD's cap that it does not reveal the glare that its own chips are producing. The DSD/SACD format is intrinsically so inaccurate that it is hiding the true sound of its own circuitry, even as it also hides and alters the sound of the music signal it processes.
The above sonic observations are all about sonic accuracy and truth. Having said everything above about objective accuracy, we're happy to admit that the midranges of DSD/SACD sound very euphonic and subjectively appealing. DSD/SACD drastically changes the sound of the music signal, wreaking a transformation similar to what some single ended triode tube amps wreak. The resulting sound is much softer and more indirect. The midranges are totally melted, from the chiseled angular solidity of the original transients, into a liquefied syrup. The resulting liquid is very easy on the ears, and is great for that relaxing cocktail or elevator background music. If that's what you want, fine, that's your privilege. But it's a far, far cry from being accurate.
By robbing the music of its natural attack, directness, and presence, DSD/SACD effectively puts you far back in the concert hall, taking what is typically a close miked recording and utterly transforming it into a semblance of what you might hear from the back of a concert hall from this same instrument. This can actually be euphonically beneficial on many commercial grade recordings. For example, Columbia was notorious for miking strings and violins too closely, and then compounding the felony by boosting the upper midrange and lower treble even further with outrageous amounts of EQ. Put on a Columbia recording of Isaac Stern, and he's uncomfortably, unrealistically right in your face. The huge transformation wrought by DSD/SACD's inaccuracy puts Isaac Stern at a reasonable distance, and tames the overly aggressive sound Columbia produced by its much too close miking and its EQ boost, by softening and liquefying the sound. Note that, when you move backward in a real concert hall, the sound of the live music up front does get progressively softer, more liquid, more indirect, more diffuse, and less coherent. That's because, as you move backward, you're hearing less and less direct sound from the musical instruments themselves on stage, and instead you're hearing more and more indirect, diffuse, incoherent sound that has been reflected many times off the hall walls, and that also has been softened and liquefied by having its trebles and the coherent leading edges of its transients obliterated, having been soaked up by all the soft stuff in the hall (seat cushions, curtains, people's clothing, etc.).
The good side of this coin is that DSD/SACD's severe inaccuracy can be euphonically beneficial on wretched recordings, making them more euphonic and easier to take. Essentially, DSD/SACD is still exercising its penchant for changing the music signal into relaxing background cocktail music, but in this case the results can be euphonically beneficial because the original recording was too aggressive. It's almost as if Sony invented DSD/SACD after it bought Columbia, seeking a way to yet again re-release those recordings (Columbia was the king repackager of software), but this time with more palatable sonics. The bad side of this coin is that DSD/SACD does this same huge transformation with all music and all recordings, even accurate recordings that should not be tampered with for optimum sound and enjoyment (at least by those who cherish the real sound of live music).
But, even if you subjectively like and want the sonic transformation wrought by DSD/SACD, you should still challenge DSD/SACD with these basic questions: Is it the proper role of a master recording and archiving system and format to be making these huge alterations in the music signal? Shouldn't this be accomplished elsewhere in the audio chain, say by an outboard box or euphonizing circuit? And shouldn't this be accomplished in the consumer's own playback chain, rather than in the recording chain, so that at least each consumer has a democratic choice of whether he wants to switch in this euphonizing circuit for himself or not, and for each recording he listens to or not? The pernicious aspect of DSD/SACD is that its inaccuracies were to be dictatorially shoved down the throat of everyone who wanted to hear a particular piece of music, and worse yet that its inaccuracies were to be used to archive recordings for posterity.
In the trebles, DSD/SACD's huge transformation has similar effects as in the midrange, and to that extent this transformation can be subjectively euphonic even while it is objectively very inaccurate. Trebles become much softer, less focused, more indirect, more diffuse, less coherent, and much more gentle. That can offset the too aggressive trebles of many Columbia recordings, and of the many recordings that are made with the mikes too close to the instruments, or overhead over an orchestra's string sections (violins throw their brightest sound upward, so they sound much brighter when miked from overhead than they do if you hear them playing from your seat down in the orchestra).
However, DSD/SACD has other problems in the trebles, which detract from its euphonic transformation. As discussed above, DSD/SACD transforms some musical treble information into bursts of white noise. This does not happen when you listen to live music from a distance in a concert hall, so this inaccuracy sounds like an artifice, not like a music recording transformed into an alternative listening perspective. Likewise, when DSD/SACD crashes and burns on some strong treble energy sounds, like sibilants and cymbals, its distortion and smearing sounds very foreign and artificial, and is not at all a euphonic transformation.
Closely Miked Guitar and Singer Demo
A second demo we evaluated provided surprisingly powerful confirmation of our observations and conclusions above about DSD/SACD. This too was a playing of a master tape (dub this time) through professional electronics. This time only the DSD/SACD recording was played; it was simple demo of DSD/SACD quality, not an A-B comparison. But this time the recording engineer was right there, conducting the demo, so he could tell us all about his miking.
The recording played was very simple music, a guitar and female vocalist singing. It sounded very pretty. The guitar in particular was notable for sounding very liquid, very gentle, very rounded, and somewhat indirect and diffuse. It sounded just like a real guitar might sound if heard or miked from 15 feet or more away in a typical room or hall. It sounded similar to what you would hear an acoustic guitar sounding like if you were sitting in a typical audience seat, about 20 feet away, at a typical church or small hall. All very nice and pretty, very relaxing and easy to listen to, very enjoyable subjectively, and certainly not challenging to one's ears.
But of course all this still leaves one crucial question unanswered. How objectively accurate was this recording? How objectively accurate was DSD/SACD in capturing and reproducing the music signal that the recording microphone fed it?
To determine this, all I needed to do was ask the recording engineer two simple questions. How far did you set up the recording mike from the guitar? And what kind of mike was it?
If the recording engineer had said that he positioned his mike about 15 feet away from the guitar, then DSD/SACD would be doing an accurate job of reproducing the sound of a guitar heard or miked at that distance. If the recording engineer had said that he used a ribbon mike (which typically rolls off above 15 kc and typically rounds, softens, and liquefies trebles), then we would have made some allowances in miking distance for the use of this mike.
But the recording engineer said that he used a high quality condenser mike for the guitar. Since condenser mikes are the world standard in picking up treble detail, they should accurately reveal the guitar's sound at the true miking distance. So, what about the $64 question? How far was this recording mike from the guitar?
The DSD/SACD reproduction sounded like a guitar would if heard or miked about 15 feet away. How close was this to the actual miking distance? How accurate was DSD/SACD in reproducing the actual mike signal fed to it?
OK, it's your turn. Take a guess. How close do you think the actual miking distance was? Remember, the closer the actual miking distance, i.e. the more it differs from the apparent 15 feet, the worse is the accuracy of DSD/SACD.
So what's your guess? Was the actual miking distance 15 feet? Or perhaps 10 feet? If you guess less than 10 feet, remember that you're giving a vote of no confidence in the accuracy of DSD/SACD. So what do you say? Not 15 feet? Not 10 feet? Less than 10 feet? Please, have a little faith!
We were expecting the recording engineer to say about 10 to 15 feet. It sounded like 15 feet, but we know recording engineers typically try to get closer than this. But we also wanted to have some faith in this highly regarded, highly touted DSD/SACD system. So we were prepared to hear 10 feet.
We nearly fell off our chair when the recording engineer gave us his answer. The actual miking distance for this guitar recording was 8 inches.
That one number says it all in a nutshell about the accuracy of DSD/SACD.
There should be a little engraved plaque on each box of DSD/SACD mastering equipment, saying: "Warning: This box may be hazardous to the health of your music. Put in music miked at 8 inches, and out comes music miked at 15 feet." A euphonic transformation is one thing (such as the liquidity bestowed by an SET tube amp), but this change wrought by DSD/SACD went far beyond what any euphonic tube electronics does. The sound of a musical instrument at 15 feet is just too different from its sound at 8 inches. This is not merely euphonic transformation of a signal. This is distortion of a signal. This is massive hemorrhaging loss of information. All thanks to the DSD/SACD digital format.
We were flabbergasted. This guitar was miked at 8 inches by a revealing condenser mike. Do you know what a real guitar actually sounds like at 8 inches? We do (we have a guitar). You hear an amazing wealth of complex musical and timbral information when you're this intimately close. You can easily hear the strong, sharp attack and release of each pluck. You can hear the buzzing noises each string makes as it is released from the pluck. The guitar has presence, vividness, tactile bite, and twang (with an emphasis on the "t").
Yet all these vivid natural intimate sounds of a guitar, heard or miked up close, and especially obvious when miked as absurdly close as 8 inches, were nowhere to be heard on this DSD/SACD recording. Instead, the guitar's sound was rounded, dulled, liquefied, and smoothed down. That smoothing down of the sonic portrait does naturally happen when you hear a guitar at a distance, such as 15 feet, which is why the recording sounded as though it had been made at that distance.
How on earth could any recording system throw away so much information from the mike feed? Where had all the sounds from the intimate 8 inch miking gone? We think we know where they went to, and how and why DSD/SACD makes all this musical information disappear. Explanations in section below (hint: excessive averaging smoothes down individually different transient musical details and obliterates them, revealing only the rounded, smoothed down overall average trend of the music waveform).
The results of this demo corroborate our findings from the A-B demo above. In both cases, the evidence clearly indicates that DSD/SACD is very inaccurate. And both independent (and very different) experiments indicate the same kind of inaccuracy, which clearly implies that this kind and severe degree of inaccuracy is indeed a property of the DSD/SACD format.
It's also important to note that, in both experiments, DSD/SACD did two general things wrong. First, it wrought a huge transformation on the sound. Some listeners might make apologies for this, pointing out that the transformation is euphonic. On the other hand, DSD/SACD's transformation is so huge in degree that it goes beyond good taste for many listeners. And of course many would rightly argue that any transformation or inaccuracy in a recording system and an archival system is utterly wrong as a matter of principle.
The second thing DSD/SACD did wrong was to lose musical information. This is actually a more serious charge, a more devastating indictment of DSD/SACD. Many transformations can be undone by applying a suitable inverse transformation. But lost information cannot be undone. Information that is lost can never be regained.
In the first demo, we heard that DSD/SACD made the violin sound as though it had been playing behind a velvet curtain, by losing all the information that would be lost through a typical velvet curtain. We also heard vivid proof of this information loss in the direct comparison to the same violin miking as reproduced via DVD-A. On DVD-A there was so much more real musical violin information, coming through clear air to the same microphone, that it was obvious that all this information was really being picked up by the mike, and that DSD/SACD was losing all this huge amount of information from the same mike feed.
In the second demo, we heard inferential proof of the same thing, thus corroborating the information loss heard in the direct comparison of the first demo. In the second demo, we heard DSD/SACD throw away huge amounts of rich detailed intimate information that a condenser mike would pick up from a guitar only 8 inches away, indeed such a huge amount of intimate information that it made the guitar sound as though it had been miked from 15 feet away.
(Continued on page 41)