Music in the Round #82
Following the B&Os' departure, I reassembled my system: three Bowers & Wilkins Diamond 802 D3 speakers for the left, center, and right channels, each driven by its own Classé Sigma Mono monoblock; two B&W Diamond 804 D3 speakers, for the left and right surround channels; and two JL Audio Fathom f113 subwoofers. Then I decided to listen to it only in stereo for the next week or so.
At first, it was a downer. I played recordings that had impressed me through the B&Os, including Mighty Sam McClain's recording of Carlene Carter's "Too Proud," from the compilation BluesQuest (SACD/CD, AudioQuest AQM-1052). McClain's voice at center stage was stable but lacked physical authority, and the instrumental support was a bit lightweight. That forced me to reposition the L/R speakers, including adjustment of their toe-in, tilt, and floor spikes. I also reran and implemented Dirac's Live Room Correction software. VoilÖ! Sam McClain was again Mighty.
Nonetheless, as I continued to play my collection of usual suspects, it became apparent that something was missing. The tight focus in the soundstage that had been so impressive with the BeoLab 90s in their Narrow mode was gone. Sure, the center fill provided by the left and right B&W 802 D3s was good enough to convince me that the center speaker was on (it wasn't), and the soundstage depth was excellent. Soundstages were a bit wider than with the BeoLabs in Narrow mode, and were free of the fuzzy aura of the B&Os' Wide mode. Nonetheless, my two-channel reference system didn't alter my conviction that the digital-loudspeaker technology behind the BeoLab 90 is truly revolutionary.
That wasn't the end of the story. Let's grant the superiority of the BeoLabs in Narrow mode with recordings of smaller groups of performers. Comparing two-channel recordings of larger ensembles played through the BeoLabs with the two- and multichannel versions of the same recordings through the 802 D3s, everything was different. A great example was La Tarantella: Antidotum Tarantulae, with Christina Pluhar conducting the vocal and instrumental ensemble L'Arpeggiata (SACD/CD, Alpha SA503), a recording that had greatly impressed me in stereo through the BeoLabs. The central voices aren't as laser-focused through the two front B&Ws as through the two BeoLabs, but the wider soundstage is partial compensation. The multichannel track is unusual in having the basic instrumental and vocal feeds in the L/R channels, with only ambience in the center and surround channels. Even so, there is a vast difference in sound compared to two-channel, and, as you might expect, I prefer the multichannel version. When I listen to the multichannel track, the entire sound is much more relaxed and open. That relaxation was at the expense of the intensity of the two BeoLabs' focus, but it raises the issue of what one desires from music reproduction.
The BeoLab 90s offered spectacular stereo sound, but despite the fact that these speakers minimize early reflections, they produced in my room soundstages with definite side boundaries. While the BeoLabs' stereo soundstages were generousand, almost miraculously, seemed to adjust to properly accommodate every recordingthey could not erase my awareness of unseen lateral borders and convince me that I was sitting in the performance venue itself, as is consistent with my idea of perfect playback.
On the other hand, the multichannel mix of La Tarantella can convey the illusion of being at the performance and of being within the same acoustic space as the performers. Even without punctate surround cues, the reverberant field surrounding the performers is unconstrained by lateral limits. Large orchestral forces, such as the Netherlands Radio Philharmonic Orchestra performing Prokofiev's Symphony 3 under James Gaffigan (SACD/CD, Challenge Classics CC72584), made the difference easily apparent. Rachel Podger's Guardian Angel, a recital of works for unaccompanied violin by J.S. Bach, Biber, Matteis, Pisendel, and Tartini (SACD/CD, Channel Classics CCS SA35513), is my most prized example of the benefits of recording a solo voice or instrument in multichannel in a sympathetic acoustic. In two-channel, Podger's violin is in my room, at the center of an ambient field that spans from BeoLab 90 to BeoLab 90 or from 802 D3 to 802 D3. In multichannel she seems no less present, but now I have joined her in the lovely little church in Haarlem, Holland, where this disc was recorded, in sessions that I was privileged to attend. This is the sort of reproduction of a real acoustic space that two-channel stereo can't providebut multichannel can.
Though my temporary return to StereoLand was made exciting courtesy the BeoLab 90s, I can't return to listening in only two channels. However, even at the risk of my financial future, I would leap at the chance to hear what five BeoLab 90sor any smaller version that B&O might come up withcould do with multichannel recordings.
Tools from another task
As downloadable multichannel recordings at higher and higher data and sampling rates become available, to say nothing of up- and downsampling by the user, we increasingly tax our playback hardware. I'm not talking about DACs, whose capabilities are clearly documented. I'm talking about my disappointments with playersdevices that organize, display, and send audio signals to a DAC, and that do just fine with high-resolution files of two channels but quickly poop out when asked to play six in hi-rez. Scrutinized, it becomes clear that they're run by wimpy CPUs chosen for their low power usage and thus low generation of heat, which permits them to run in fanless silence. You can buy silent, Intel i7-class CPUs, but only a very few are offered as fully configured audio players, and even these are often configured for predefined software and environments and thus won't support multichannel playback.
One can hit the limits even with a relatively potent and fully packaged player such as my i7-based Baetis XR3. For example, to use Dirac Live EQ, I often downsample hi-rez filesand downsampling a DSD256 recording to 24-bit/192kHz PCM rather than to 24/176.4 PCM, or a 24/352.8 PCM file to 24/176.4 instead of 24/192, results in a substantial increase in CPU load.
So, I can see, waiting for me on the horizon, a more potent processor. There's the option of a powerful but noisy computer placed in a different room (as is my NAS, on which my music files are stored), but that complicates thingsI'm uncomfortable using my smartphone or tablet to control playback. Linking a mouse, a keyboard, and a large screen to a remote PC is not a task I welcome. The majority of silent PCs are offered as industrial models, business servers, gaming PCs, and/or as DIY projects. They require that users acquire and install the necessary hardware and software, and the latter makes even greater demands. The few that are specialized for audio applications often seem to be more expensive than average.
So when I saw the reasonably priced and attractively packaged Compulab Airtop, I figured it was worth a try (footnote 1). The Airtop-D i7 is a desktop CPU that, for $1423, sports Intel's fifth-generation Core i7-5775C processor, 16GB of RAM, a 256GB mSATA solid-state drive for the operating system and programs, and on-board Intel Pro Graphics 6200 graphics card. It runs cool, is dead silent (no fan), and it's compact: 12" high by 4" wide by 10" deep. It's also solidly built (12.75 lbs), and its cast-aluminum case and heatsinks pass the knuckle-rap test. If you're using the www.signalyst.com/consumer.html">HQPlayer app, you should consider adding the optional NVidia GeForce GTX 950 Graphics Processing Unit, which raises the price to $1694. Because HQPlayer supports offloading audio processing to the NVidia GPU, it can provide a real boost in sound quality.
It was a snap to upload a copy of JRiver Media Center and install Dirac Live, as well as import the EQ filters and exaSound ASIO drivers from the Baetis. I also had the Airtop scan a couple of the music-file directories from my remote NAS and build a quick library. Within minutes, music was playing just as it always has. Regardless of the degree of up- or downsampling, or transcoding from DSF to PCM, there were no problems, and the Airtop seemed faster at loading the files for playback. Examination of the workloads in each core revealed that the Airtop's newer-generation Intel CPU (i7-5775C) apportioned tasks more evenly than does the Baetis's older CPU (i7-4770S). That might predict greater workload capacity, but since my exaSound e28 can play multichannel files of up to only DSD256, I couldn't say how either machine would fare with upsampling to DSD512.
Since the Airtop was selected as a basic silent PC, I also wanted to see what it could achieve if configured with the files onboard. The Airtop supports up to five additional SATA 3.0 drives, and can even organize them into a redundant array of independent disks (RAID), to ensure survival of all data should any drive fail. This is how most of us configure our NASes, but remember that a RAID's data redundancy is no alternative to independent backup of valuable files. To allow me to assess the Airtop's performance with onboard storage, Samsung sent me an 850 EVO ($1499.99)a high-performance, 4TB solid-state drive (SSD). (Smaller SSDs are cheaper: 1TB EVO drives have a street price around $300, and 500GB are well under $200. Depending on the size of your library, any four or five of these drives in a RAID5 configuration should get you high-speed performance as well as the safety of data redundancy.)
With the same files on the internal Samsung SSD, performance with JRiver was lightning fast. Huge fileseg, the 24/354.8 FLAC file of Prokofiev's Symphonies 3 and 4 with Gaffigan and the Netherlands RPO (Challenge Classics CC72584, download from www.spiritofturtle.com); and a DSD256 recording-session file of Mahler's Symphony 3, with Iván Fischer and the Budapest Festival Orchestra (Channel Classics, gift from the producers)loaded as fast as a 16/44.1 CD. Such minimal buffering made possible listening comparisons that were almost fast enough to be A/B comparisons.
The response time with HQPlayer was almost startling; it benefited from the quick file access from the SSD, but also from coprocessing by the NVidia GPU board. Overall, the onboard Samsung SSD didn't seem to affect the sound quality, but it did transform my experience. I had become inured to the buffering delays that seemed inevitable with hi-rez multichannel files; now that I've experienced such speedy response, I'm not enthusiastic about going back.
For those of us willing to assemble the required hardware and software, a silent PC such as the Airtop is an attractive option. If you use HQPlayer, the optional NVidia GPU has value. If you don't, consider that it occupies the Airtop's only accessible PCIe slotone you might want to use for a specialized output board. As for the speedy, silent Samsung SSD, it can't be faulted as a desirable option.
Footnote 1: Compulab, Israel. Tel: (800) 832-9824, (972) 4-829-0168. Web: airtop-pc.com.