Digital Processor Reviews

From either server, via either LAN or USB, I was thrilled with what I heard. The s88 sounded just right from the first notes, and that impression endured as I immersed myself in a wide range of music over several weeks.

I often begin my listening with something favored and familiar; in this case it was Arvo Pärt's Fratres. I have 10 recordings of it, ranging from the classic ECM recording with Keith Jarrett (piano) and Gidon Kremer (violin) to a percussion solo or a saxophone quartet plus violin. Of my two current favorites, the first is the one by violinist Victoria Mullova (Onyx 4201, CD). Mullova, up front, is all hope and awe accompanied by Paavo Järvi and the Estonian NSO in a spacious, calming acoustic that's very well conveyed on this standard CD.

This acoustic contrasted sharply with the one on my second favorite, with cellist Maya Fridman (multichannel DXD, The Invisible Link, TRPTK TTK-0001, download from trptk.com). Via the s88, Fridman's intimate, intense cello was centered in a large, well-defined space. The sudden entry of pianist Daniâl Kool is almost shocking because he is so up-front and close-miked. The recording puts us there with them; the engineering is perfectly matched to this intensely personal performance. The s88 lays out the contrasts between these two performance and recording aesthetics while making each of them easy to savor on their own terms.

Solo piano recordings, in stereo or multichannel, are also revealing of a DAC's ability to communicate the music. Every piano has its own sound; each pianist shapes that sound in different ways; each recording places the piano differently, relative to me (and the microphones) and to the performance space—which itself has distinctive characteristics.

A recent release of Thalberg transcriptions by Paul Wee in multichannel (BIS BIS-2515, SACD) comes on the heels of a stereo set by Marc-André Hamelin (Hyperion CDA68320, CD), which I have been enjoying. There is no overlap in the specific transcriptions, but they're all similar in style, drawn from famous operas and lieder and transcribed by Thalberg for bravura performance. Hamelin's selections are more bravura, and his instrument is close enough to let us hear its body and the zing from the strings as they are struck. Wee's selections are more seductively ingratiating, his instrument more distant yet more present. The distance, though, gives it less of an edge.

Neither is better than the other, but they are different, and the exaSound exposes those differences to the point that you can start to understand the engineer's thinking and the way that each perspective suits the performance.

Among my guilty pleasures are modern arrangements of baroque and early classical-era music. A spectacular example is "Gypsy Baroque," a torrent of charming melodies performed by an ensemble of bowed, strummed, blown, and hammered instruments (Il Suonar Parlante Orchestra, Vittorio Ghielmi, Alpha 392, CD). Telemann, Vivaldi, and Tartini are interspersed with gypsy tunes, arranged to maximize instrumental color.

All of this comes smiling through the s88, both the bloom of the ensemble and the flavors of the distinct voices. The last track, my favorite, is a mash-up of a sîrba (a traditional Romanian dance melody) and the last movement of Mozart's Violin Concerto in A major K.219! It includes a passage that sounds like Stéphane Grapelli. It's a rip, from beginning to end.

How did the s88 fare with a full orchestra? In multichannel, it was great, with a wonderful recording of Beethoven's Concerto for Violin, Cello, and Piano in C major, Op.56 (the Triple Concerto). I've always preferred performances of this piece with established string trios instead of three famous soloists; in true trio performances, the dynamic is more concerto grosso than competition. The most recent recording, featuring the Van Baerle Trio and the Hague Residentie Orkest conducted by Jan Willem de Vriend (Challenge CC72801, DSD64 download from Native DSD), is thrilling. The trio plays as one, and the s88 balances and contrasts it with the weight, depth, and authority of the orchestra.

The final test was to determine how the s88 handled the human voice. I bypassed my old favorites, Emma Kirkby and Marianne Beate Kielland, in favor of a new favorite: Go to Qobuz and stream "Sister Rosetta Goes Before Us," sung by Alison Krauss on her album with Robert Plant, Raising Sand (Rounder 11661-9075-2, CD). The s88 doesn't obscure the background noise, but it does differentiate the bloated and indistinct bass and—oh my—it does present Krauss's voice and violin as naked and pure as if she were here.

A transient issue
About halfway through the review process, I noticed a sound when I changed audio format (from DSD to PCM or vice versa) or digital input (from USB to LAN or vice versa). Under normal circumstances, it was little more than a tick from the main speakers and a bip from the subs. I only heard it from the balanced inputs—it was absent when I hooked things up with RCA. It wasn't much, but it could be annoying when playing a playlist containing mixed-format files.

Not to worry: exaSound's George Klissarov suspected a fault with the review unit and, in a swift round trip to Toronto, an out-of-spec resistor was discovered and replaced. The tick disappeared. Klissarov told me that there "will be a permanent change of manufacturing and testing procedure" to prevent recurrence of the issue.

A couple of comparisons
The s88 is nominally equivalent to exaSound's e38 Mark II plus the exaSound Sigma Server, although that combination has fewer USB "A" connectors, comes with either XLR or RCA outputs (not both), and dedicates its one USB "B" connector to pairing the Sigma Server to the DAC. I set up the s88 and the e38+Sigma as separate, grouped zones in Roon in the Delta Server so that I could A/B them as they played in synchrony.

I was surprised to hear so little difference between them, but perhaps I should not have been. The s88 is evolved from the e38 Mark II, a recent and excellent DAC. There was equal detail in the extreme treble, but the s88 seemed more natural in reproducing a violin's pure single-string note while the earlier DAC was a little bit drier. There was also a distinction in the midrange that helped delineate human voices and instruments in reverberant spaces. Finally,

I sensed marginally more tightness and impact in the bass. Without the instant A/B comparison, I would not have managed to delineate these small differences but they had their effect nonetheless: Cumulatively, they drew me into the music more from the first time and every time after.

Next, I compared the s88 with the Okto dac8 PRO, connecting the s88 and the Okto by USB to the Baetis server and (again) linking them as "zones," this time in JRiver. Again, the differences were subtle. Perhaps the Okto was a bit tighter in the bass and thinner in the vocal range than the s88. I do not know which presentation is truer to the source. Each DAC has advantages. The Okto (Euros 989) is limited to 24/192 PCM and DSD128, which eliminates some options for upsampling that are available on the s88. But the Okto's price advantage could tip the scales in that direction. It may lack Ethernet, but it has a bigger volume control knob!

Conclusions
The exaSound s88 streaming DAC performs its tasks superbly; in fact, it exceeds the performance of any DAC that I have used. I would describe its sound as transparent rather than detailed, dynamically responsive rather than lively, and honest in how it presents voices.

Readers with two-channel systems may wish to wait for an equivalent stereo device; there probably will be one, given the company's history. But others might see this as an opportunity to build a multichannel system or to DIY their own stereo system with active loudspeakers, DSP, and crossovers in the host server. For that, you need multiple DAC channels even for a two-channel system.

For some who are already committed to multichannel, the s88, with its superb DAC, convenient streaming and that oh-so-welcome volume control, may be the realization of their dreams. It is of mine.