transients. Note by the way that we could hear the degradation of this 100 kHz filter in spite of the fact that other previous filters in the recording/reproduction chain (e.g. the microphone and master tape or vinyl disc) themselves were already rolling off the music signal's response above 20 kHz. Countering this rolloff above 20 kHz was the fact, proven independently by our other research measurements, that the spectral energy content of many musical transient sounds was actually rising above 20 kHz (such as that gentle cymbal kiss, which we measured as peaking at 40 kHz). Of course, if these other links in the chain had had wider bandwidth, then the introduction of our 100 kHz filter would surely have been even more audible as a more severe degradation. Thus, our research experiment demonstrates that extending the bandwidth of the medium to 80 kHz (as with a 24/192 DVD-A), instead of restricting it to being an additional filter in the chain at 20 kHz (as with a 16/44 redbook CD), is sonically important. Note too that, as mastering equipment continues to improve its bandwidth beyond 20 kHz, it becomes even more sonically important that the playback medium bringing this extended bandwidth signal into our listening room have bandwidth that extends far beyond 20 kHz.
Another possible or potential advantage, of the wider bandwidth offered by high resolution audio discs, is that it allows for gentler filters and perhaps more accurate filters. As a general rule, gentler filters can yield superior sonic results, since they impose less degradation upon the time domain waveform of the music signal. This topic is very complex, so we'll only skim the surface here.
In the case of DSD/SACD (or other 1 bit or delta modulation or pulse width modulation system), there is the promise of needing only a gentle filter at the output of the processing, to merely reduce ultrasonic noise and perform simple signal integration over time (some of which is accomplished in any case by our loudspeakers and by our hearing). However, in the case of DSD/SACD in particular, they don't tell you about the other steep filters within the system, which negate the sonic value of the gentle output filter that they do brag about. There actually are very steep filters within the DSD/SACD system, and to make matters worse, some of these are not merely rolloff filters that are relatively benign. For example, the noise shaping of DSD/SACD includes a filter that actually introduces a steep peak (of ultrasonic garbage) at about 53 kHz. Also, the overly aggressive and extended temporal signal averaging of DSD/SACD may itself be regarded as a steep filter that degrades the music signal, degrading it so severely in fact that it obliterates singular musical transient information.
In the case of PCM, extending the potential bandwidth of the system (to 48 or 96 kHz, by upping the sampling frequency to 96 or 192 kHz) could theoretically allow the playback anti-aliasing filter to be far more gentle, as opposed to the steep filter required by redbook CD, which is trying for a full 20 kHz bandwidth with a paltry 44 kHz sampling rate. However, in practice it may be inadvisable to employ such a gentler filter, since this filter's other important task, reconstruction of music's higher frequencies, would then be performed less accurately. This topic is a matter of active current debate (see recent articles by our colleague Keith Howard in Hi-Fi News), so we won't get into the complex pros and cons here.
Finally, let's look at the most sonically crucial issue: reconstruction accuracy. Here the high resolution audio disc formats, especially the PCM of DVD-A, score a decisive sonic advantage over redbook CD. And, with this issue of reconstruction accuracy, DVD-A scores its sonic advantages both by virtue of its greater bit resolution and also by virtue of its higher sampling rate, with both these features working together, hand in hand, to provide you with a vastly superior sonic result.
To understand this, let's first look at how reconstruction accuracy works. The common layman explanations of digital show you a follow-the-dots model. In this model, they show you the dots representing the sampling points of the digital system, and they show you the music waveform superimposed upon the dots. What digital systems do, according to this layman's follow-the-dots model, is simply trace a smooth path between the sampling dots, and presto, the digital system gives you the original music waveform, since the music waveform clearly is represented by the smooth path between the dots, just as the layman's diagram shows. But there's a severe failing in this commonly taught, layman's model. It is valid only up to a certain threshold frequency, which is about 1/5 the bandwidth of the digital system (i.e. only up to 1/10 the sampling rate). Thus, the PCM system of redbook CD operates via this follow-the-dots model only up to about a frequency threshold of 2 kHz. Above 2 kHz, there are not enough sampling dots for this follow-the-dots model to work. Above their frequency threshold (2 kHz for redbook CD), digital systems must employ a different tactic to furnish you with the music signal waveform.
What do digital systems do above their threshold frequency, instead of following the dots? They artificially synthesize the music signal waveform, and the tool they use for doing this is called (aptly) a reconstruction filter. Now, to do its job perfectly, and reconstruct a completely accurate music signal waveform, the reconstruction filter would have to be infinitely steep, have an infinite number of bits of resolution, and incidentally take an infinite amount of time, to perform a perfectly accurate computation of what detailed twists and turns the artificially synthesized music signal should take, for all frequencies above their frequency threshold (2 kHz for redbook CD). But of course it is physically impossible to make a filter that is infinitely steep, or that has an infinite number of bits of resolution (and you don't have sufficient lifetime to wait an infinite amount of time to hear the first note after you put on a digital disc to play).
Thus, the music you hear from a PCM digital system above its threshold frequency (about 2 kHz for redbook CD) is only a synthesized approximation, less accurate, and hence less musically natural, than the music below its threshold frequency, which is derivable via the simple follow-the-dots model. This amusical inaccuracy, above 2 kHz with CD, might explain at least in part why CD sound often has artificial hard glare in the upper frequencies, especially audible in the upper midrange and lower treble (where human hearing is at its most sensitive). This glare, from the approximation of calculating the correct subtle details of upper frequency information, itself also serves to further clog or block subtle musical information above 2 kHz, which explains why a vinyl LP of the same recording can reveal more of music's subtle natural detail, especially in the upper midrange and above. The fact that the reconstruction filter's calculated synthesis of music is only approximate above 2 kHz for CD also explains why different reconstruction filters sound so remarkably different. For example, different CD players or D/A converters sound different in large part because they employ different reconstruction filter algorithms, which perform different approximations to synthesizing the correct music signal waveform above 2 kHz. And D/A converters that offer you a switchable choice of reconstruction filters sound remarkably different at those different filter choices for the same reason.
Incidentally, at upper frequencies the output from DSD/SACD is even more of an approximation, i.e. even less accurate. Because of its particular design, DSD/SACD happens to be better than redbook CD up to about 8 kHz, but above 8 kHz is actually even worse than redbook CD.
Obviously, we could get better accuracy out of a PCM digital system if we could raise the threshold frequency. Then the follow-the-dots model would be valid to a higher frequency, and could furnish an accurate musical signal waveform to a higher frequency. In short, music would sound much better (more naturally accurate) up to a higher frequency. We would not have to introduce and listen to the approximations of the reconstruction filter artificially synthesizing our music, until some much higher frequency. That's precisely what high resolution PCM discs (for example DVD-A) give us. And this in fact is the most important sonic benefit they give us.
With a sampling frequency more than 4 times higher than redbook CD, a 24/192 DVD-A has a threshold frequency more than 4 times higher, i.e. beyond 8 kHz instead of redbook CD's 2 kHz. Thus, the 192 kHz sampling rate furnishes accurate (i.e. natural sounding) music to a frequency more than 4 times higher than redbook CD, without having to invoke the synthesized approximations of the reconstruction filter. Most listeners report that discs with very high sampling rates sound more musically natural than CD, well within the audio spectrum. Now you know the reason why.
Listeners also report that 192 kHz discs sound better than 96 kHz discs, again well within the audio spectrum. And controlled listening tests direct from master hard disc at various sampling rates likewise report that the 192 kHz sampling rate sounds better than the 96 kHz sampling rate. Why should this be, if the higher ultrasonic sampling rate merely provided the benefit of superfluous wider bandwidth in the ultrasonic region that is beyond audibility? The answer is that the higher ultrasonic sampling rate, while providing slightly better time domain reproduction of music's highest frequencies, also provides another, far more important benefit. The 96 kHz sampling rate provides follow-the-dots musical accuracy only to just beyond 4 kHz, whereas the 192 kHz sampling rate provides this musical accuracy to beyond 8 kHz, again without having to invoke the artificial synthesized approximations of the reconstruction filter. The musical spectrum from 2 kHz to 8 kHz is rich in very important musical energy and information, and is also the frequency region in which our human hearing is the most sensitive and discriminating. Thus, it makes perfect sense that the higher sampling rate of 192 kHz, by providing better musical accuracy up to 8 kHz, instead of redbook CD's mere 2 kHz, or up to the 4 kHz provided by 96 kHz sampling, would sound more musical natural, in this crucial spectral portion between 2 kHz and 8 kHz.
Additionally, 192 kHz sampling provides further sonic benefits. Because it reproduces all recorded information more accurately, it not only sounds more musically natural (e.g. sweeter with less glare), but it also reveals more of the subtle information contained in a recording. This subtle recorded information contains important cues about the timbral and textural sounds and noises of musical instruments and voices, so everything simply sounds more transparent and effortlessly real. And this subtle recorded information also contains crucial cues about spatial imaging, so all aspects of spatial imaging (width, depth, ambience, 3D air and space) are revealed with better transparency and richness.
All these crucial sonic benefits, provided in the critical spectral region between 2 kHz and 8 kHz, represent the most important sonic advantage of 192 kHz high resolution discs over discs with lower sampling rates. The extra ultrasonic bandwidth also provided by these discs might or might not provide audible benefits for listeners, but this is far less important than the accuracy improvements well within the audible spectrum that these ultra-wide bandwidth discs provide.
What about the 24 bit amplitude resolution of these high resolution PCM discs? What sonic advantages does it provide, over the 16 bit amplitude resolution of redbook CD? Below 8 kHz (for a 192 kHz sampling disc), the increased 24 bit amplitude resolution teams up with the higher sampling rate, allowing the simple follow-the-dots model to provide even better signal waveform accuracy, since the dots are now positioned with 256 times better accuracy on the amplitude scale. Then, above 8 kHz, the 192 kHz PCM sampling system must still turn to the reconstruction filter, to synthesize a calculated approximation to the music waveform, for the frequencies above 8 kHz. But, thanks to the reconstruction filter having 24 bit resolution signal data to work with, instead of the coarser 16 bit resolution of redbook CD, its calculated approximations are 256 times more accurate, and so its synthesized waveform is 256 times closer to the original music signal. Thus, the recording's high frequencies, from 8 kHz to 20 kHz, can be synthesized far more accurately by the reconstruction filter. And this provides all the aforementioned sonic benefits for the 8-20 kHz region (better musical naturalness, better transparency, more realistic reproduction of subtle detail, and better spatial imaging).
To take full advantage of all this improved accuracy and resolution that 24/192 discs can provide, the disc player must of course be extraordinary. The disc player design must be very carefully executed for maximum accuracy and transparent resolution, and minimum coloration, in every phase and stage, from transport to power supply to output stage. Only a player like the McCormack UDP-1 can fully exploit and reveal to you all the potential sonic benefits that these discs intrinsically contain. The sound of the UDP-1 takes a tremendous leap upward with these high resolution discs, because this player has indeed been carefully designed for maximum revelation at every stage. Other players might sound somewhat better than CD when playing high resolution discs, but they don't take as tremendous a leap in quality as the McCormack player does. Some of the finer sonic details and benefits provided by good high resolution discs simply don't make it through the circuitry of other players. As stunning as the UDP-1 sounds when playing a good CD, it really comes into its own and struts its stuff when playing high resolution discs, and its margin of sonic superiority over other players becomes even more obvious.
A single example will suffice to illustrate this. When we changed from our best demo CDs to some demo 24/192 DVD-As, in evaluating the UDP-1, we heard all of the sonic improvements we had expected to hear from high resolution formats, based on previous listening experiences. These sonic improvements included much better transparency, revelation of subtle timbral detail, musical naturalness, relaxed ease, open airiness, speed, delicacy, and upper frequency extension. In all these areas of sonic improvement the McCormack player sounded superb, and better than we had heard from other players playing DVD-A discs.
But the UDP-1 also gave us a whole new kind of sonic improvement, a totally unexpected surprise. The spatial imaging also became dramatically better, from the high resolution discs, than it had been from any CD. The sense of 3D space and air surrounding each instrument, and the solid palpable body of each instrument, was astounding. Also, the space of the stage acquired literally whole new dimensions of width and depth and vivid reality. We deliberately used DVD-As which were simple 2 channel stereo recordings, rather than surround recordings, so we would be comparing apples to apples in our sonic comparison to 2 channel CDs. But, heard through the UDP-1, the sense of 3D space was so vivid and rich, from just this 2 channel stereo 24/192 DVD-A, that it sounded almost like a what a well made surround sound recording can achieve (by putting spatial ambience into the surround channels, thereby helping to enrich and vividly define the stage space up front). It was shocking, and a welcome surprise, to hear such a huge improvement in spatial imaging from 2 channel stereo, thanks to the high resolution 24/192 DVD-A format and thanks to the McCormack's revelatory powers. The sonic cues are very subtle, that allow the human ear/brain to hear and construct such a rich, believable 3D space from just a 2 channel recording. So this dramatic spatial improvement proves that the UDP-1 must be doing a superb job of reproducing and revealing the subtlest information from recordings, both musical information and spatial information.
Incidentally, when high resolution audio discs contain more than 2 channels (e.g. for full surround sound), there is not enough bandwidth to process a 192 kHz sampling rate, so these discs are presently confined to a 96 kHz sampling rate (the new HDMI digital interface is reportedly capable of handling 8 channels at 24/192, so with that and blue laser a better sounding future is near). Although a 96 kHz sampling rate provides merely 4 kHz as a threshold frequency, instead of the 8 kHz that a 192 kHz sampling rate provides, these 24/96 discs can still sound far superior to 16/44 redbook CD. That's because the reconstruction filter, when it does take over at 4 kHz (which is merely a doubling of redbook CD's 2 kHz), works with sampling dots that are 256 times more accurate than redbook CD (24 bit resolution instead of 16 bit), so with 24/96 discs the calculated and synthesized music waveform above 4 kHz is 256 times more accurate, more subtly detailed, more musically natural than redbook CD.
Imagine now that that richly holographic 3D stage space, so realistically portrayed by the UDP-1 from merely two channel recordings, as described above, gets wrapped all around you. This is a nutshell is what the UDP-1 can give you, from well made surround recordings. The space that was so vividly tangible and believable up front on stage now becomes the space in which you are immersed.
Thanks to the UDP-1's whole new higher level of transparency, taking the transparency state of the art to new levels, the UDP-1 also literally re-defines the state of the art for surround sound, especially in spatial portrayal. The sonic cues that define the acoustics of a large alternative venue constitute very subtle recorded information. So the UDP-1's superior transparency reveals more of this subtle hall acoustics information, thereby portraying the large alternative acoustic space around you more realistically and believably.
The UDP-1's superior transparency also reveals more of the subtle natural timbres of each musical instrument and each human voice, so all of the music and voices around you sound more real, more three dimensional, and more palpably solid - in short, more believable, as if they were really there with you, sharing the large acoustic space of the alternative venue that the UDP-1 portrays so superlatively.
In sum, for well recorded surround sound, the McCormack UDP-1 makes the music more real, the voices (both singers and dialogue) more real, and the space more real. When surround spatial imaging gets this superb, a special complementary synergistic effect occurs, which is almost magical. The believability of the richly defined space all around you bestows added 3D solidity and believability upon performers sharing that space with you, and, conversely, when the timbres and nuances from the performers themselves are so realistically limed, their added believability makes the space around them, the large portrayed space itself that they share with you, more believable.
Many surround recordings, especially of pop and jazz, are unfortunately poorly recorded, essentially in closely miked multi-mono, which limits the performers to being located at the loudspeaker positions in your listening room (and even traps them inside the loudspeaker boxes). The UDP-1 accurately portrays this multi-mono imaging, so it does not create space where none is captured in the recording. Even on these recordings, however, where there is no large space to
(Continued on page 129)