Stacking the deck. Expecting split hairs on the order of selectable digital filters—their differences can be so marginal as to be irrelevant—I picked from our collection ancillaries of the highest resolution. That meant a full Allnic cable loom, Vibex DC/AC filters, Wyred4Sound's best linestage version and Cube Audio's magnificent Nenuphar widebanders. Since those speakers prefer FirstWatt's SIT-1 monos, those were my amps of choice. Switching the Denafrips Terminator between coax, AES/EBU and I²S made for instantaneous comparisons.

For music, I continued the theme of a stacked hi-rez deck with albums like Dhafer Youssef's 24/96 Sounds of Mirrors and Le Trio Joubran's 24/88 The Long March bought as Qobuz downloads. If your eyes have rolled up into your forehead by now with a big mental "puleeze", welcome to perfectionist hifi. It demands ideal conditions to come off, not MP3 dumbed down to 20% of a CD's original data density. Once I'd learnt where to hear differences in my best-case scenario if any were to be had, my game plan was to transition to standard Redbook files.

Actually, it took very little time to arrive at a preferential hierarchy: I²S was best followed by AES/EBU. S/PDIF via BNC came last. The offset was archeological. I'll explain. Have you ever watched a reportage on an archeological dig? Workers equipped with just brushes carefully peel away fine layers of dirt. Layer by layer, they lay bare a buried statue. What starts out as a vague amorphous shape gains more and more distinctiveness. Removal of the encakement retrieves ever greater relief and surface textures. Compared to I²S, the two other transmission regimes didn't remove as much. Their brushes were coarser. Their workers gave up sooner. Dimensionally, the sound was flatter, just not in terms of reduced soundstage depth. That is about the relative distances between virtual sound sources. This type of flatness had to do with individual sounds relative to themselves, not others. Front to back, how well had they been excavated and stripped of concealing surface obscuration? This game was all about dimensionality. It was about how well sounds stepped out from that more generalized field—our musical equivalent of soil—to register as fully individuated and distinct. If sounds were carved from wood, most remain attached to the plank. The fronts stand out but the backs remain connected. You can't walk around to inspect them from behind. Here I²S disconnected the most. Its sounds freed into the best semblance of 3D imagery. They sounded most real

In a fine sense, this also involved an aspect of airiness or illumination. That was greatest with the AudioQuest Coffee HDMI. Cable quality and transmission protocol could thus pull in opposite directions. What in AES/EBU I soon considered a lesser protocol than I²S could edge it out if the latter ran an inferior cable. With the Coffee patch and its battery bias on the dielectric, there was no doubt that for example Dhafer's vocals had the greatest realism, finesse, surface textures and spatial keenness. Whilst subtle in the grand scheme of room disturbances and speaker nonlinearites, that was plainly demonstrable, hence meaningful. It wasn't like the vast majority of digital filter changes I'd made over the years with sundry DACs. Those rolled up my eyes. Relative to possible differences between DACs per se, I'd still peg those as potentially larger. Hence I wouldn't suggest to pursue a particular converter just because it has I²S. But given matched conditions—same sender, same receiver, equivalent cable quality to focus down on different transmission protocols—I certainly give I²S the nod. Going forward, it's my new default choice between the Jay's Audio CD transport or Soundaware USB bridge as senders; and the Denafrips as receiver. Considering that the AudioQuest cable is rather more affordable than my S/PDIF alternates was a nice bonus.

How about 'normal' CD fare of standard 16/44.1 resolution, perhaps not even recorded that well? Likely because much of it starts out with rather less archeological potential as it were—and conversely, because the premium stuff doesn't suffer glare to begin with—I now heard the core improvements as less glare/hardness and greater smoothness/suppleness. I was surprised by how demonstrable it remained. Only the focus, of what fell away to be improved, had shifted from finer stuff like greater spatial relief and surface texturing to coarser matters like hardness and glare. That points at high-end hifi's intrinsically hierarchical pursuit. One first attends primary matters. Once those are sorted, one polished up the secondary tier. Only when that's in order does the tertiary tier emerge and so forth. Gross errors completely overlay and mask finer deviations. It makes no sense to worry about the low-level stuff if the basics haven't been handled already. On exactly what tier does today's I²S belong? How many levels of audiophile hell or heaven are there to begin with? I haven't the faintest. It's enough to acknowledge the basic multi-tiered setup, then add that in my opinion, power delivery like cords and conditioners plus resonance control aka purpose-engineered racks must come before one worries about I²S versus AES/EBU. Ditto for sorting USB/RJ45 pollution from PC audio. Address that first, then tweak up what happens past optimized USB or networked music delivery. I'll conclude with suspecting that due to its ultra-speed 18Gbps video specification, top-quality HDMI cable might actually be superior to hifi's coax/balanced audio alternatives. If so, the cause for I²S-via-HDMI's improvement could go beyond just the I²S protocol of separate data, word and bit clock lines. Responsibility for the sonic gains might be shared by a better cable concept per se. As a listener of course, why and how don't matter. If it sounds better, it sounds better. And I²S over HDMI did, period. Puleeze? No, thank you very much to Denafrips, Jay's Audio and Soundaware for including it so I can exploit and enjoy it from now on. In this racket, every little bit helps!