Music in the Round

I'm sure that the larger manufacturers were displaying new audio/video receivers and preamplifier-processors at the Las Vegas Convention Center last January, during the 2014 Consumer Electronics Show—but except for dinner, I didn't get out of the Mirage, where I slept, or the Venetian Tower Suites, where the high-end audio displays were concentrated. Still, there were new developments to report; see Stereophile's almost live on-line coverage.

Most interesting was the demonstration of a new approach to room equalization, presented by the canny folks from Dirac Research. Using a 7.2-channel system, Dirac demonstrated their Active Room Treatment System, in which each speaker is also used as an active compensation for the room modes in its vicinity. This harks back to an idea, by Nelson Pass, long ago marketed as the Phantom Acoustics Shadow active low-frequency control. Stereophile reviewed it in December 1989, and Keith Howard included it in his discussion of room correction in January 2008. I reviewed a related device from Bag End, the E-Trap, in July 2008. The Dirac system adds no hardware to a multichannel system, but instead measures the local modal behavior for each speaker and, in software, creates compensatory filters. The result should be a global minimization of modes that will be equally effective for two- and multichannel sources.

The playback of only the correction signals was analogous to examining your vacuum-cleaner bag and being pleased that its contents were no longer in your room. The Unity is not yet in production, but it's tantalizing. Dirac's Live Room Correction System (see below) was featured in the Theta Digital room at CES—this speaker/room EQ system is included in the latest version of Theta's Casablanca IV preamplifier/processor.

Dirac Research Live Room Correction Suite
For more than a year now I've been hearing the siren call of high-definition file downloads, and have finally surrendered. The problem was choosing a setup with enough quality and capacity to be suitable for multichannel music. It was also necessary to somehow incorporate room correction into the system, and while I'm pretty confident in dealing with audio hardware, audio software is not one of my strong suits. It's clear that there are many ways to implement room equalization in a file-based system, ranging from the packages and plug-ins used by professional musicians and studios to the facility built into Windows Media Player.

There are also external devices, such as the Trinnov Optimizer (which I wrote about in my May 2010 and September 2013 columns), that one can insert in the signal path. I believe that I'm on the right track now, and my experience with Dirac Research's Live Room Correction Suite has given me that confidence.

I'd known about Dirac's Live Room Correction Suite (LRCS) for a while. John Atkinson and I were impressed by a presentation Dirac gave us two or three years ago, when they explained their basic research and demonstrated their mixed application of IIR and FIR digital filters. Since then, LRCS has been offered in only one consumer device, the Datasat RS20i audio processor, although both Theta Digital and Emotiva have announced implementations. In theory, one might feed the output of a streaming device into one of these via HDMI. On the other hand, it's unlikely that any such hardware audio devices will have the processing power of a full-blown computer.

Timidly, I downloaded LRCS from Dirac, and installed Dirac's Live Calibration Tool (DLCT) and Dirac Audio Processor (DAP) on the C.A.P.S. V3 Zuma PC that I use as a music server with JRiver Media Center 18. Once installed on a PC, to achieve the desired results, the DLCT measures the room and speakers, establishes target curves, and develops digital filters. The DAP becomes the virtual output device for the server, and uses the user-selected filters to apply the corrections to the datastream, which is then sent to the output device. While it's clear that the DLCT and DAP operations are linked, they are performed independently. In practice, they can be run on different hardware, and the filter files can be transferred from one to the other.

DLCT is operated through a clear and informative graphic interface; it's easy to follow, and the help info at the right margin makes the manual supplemental. First, you specify the output configuration (in my case, 5.1 channels) and the desired filter resolution (44.1, 48, 88.2, or 96kHz). Then you select the output device (fig.1). This was the only bump in the road—my exaSound e28 multichannel DAC's ASIO driver is not yet compatible with the DLCT. Both exaSound and Dirac have promised a solution, but for now, the simple workaround was to use another output. I chose HDMI, which I piped through an Oppo BDP-105 and on through the rest of my system.

After the output device had been selected, DLCT asks for a microphone input and its matching calibration file. I had success with both the Omnimic2 and the miniDSP UMIK-1, but almost any USB or analog mike would work. This led me to the critical step of adjusting the mike gain and the overall output level for the test signals, as well as for each channel (fig.2). Click on each channel, and DLCT plays the test signal while displaying the detected signal level on a bar-line meter, the optimal level range indicated in color. Adjustments here assure you that each channel is correctly identified and that the stimulus and microphone gain will return useful measurements.

Things get interesting on the fourth menu page. First you click on the seating/listening arrangement (chair, sofa, auditorium): a picture appears with the recommended mike positions indicated (fig.3). You can view the picture from any of three perspectives, to better appreciate the mike positions in three dimensions. Nine positions were recommended for my sofa-centered setup. After positioning the mike at the main listening position, you click Proceed and stand back. DLCT tests each speaker individually, and then retest the first speaker to identify any clock drift between playback and recording devices. It chews on the data a bit, then prompts you to move the mike to the next recommended position. Repeat until done.

DLCT then displays the frequency-response and impulse measurements for each speaker, along with a superimposed target curve. Both the target curve and the frequency-band correction limits can be modified by clicking and dragging. Then, click to calculate the filters; in seconds, DLCT displays the predicted corrected frequency (fig.4) and impulse (fig.5) responses. Finally, save the project and the target curves by name for DAP, and to use those measured results to develop alternative correction filters. DAP can hold four filter sets in memory for easy comparison, and swap any of them with any others you've saved, so feel free to develop experimental sets.

I had no trouble starting DAP and selecting it as JRiver's output device. The main DAP screen has status indicators for connection and clipping at the top, along with a filter bypass switch, four filter trays in the middle, and, at the bottom, a dropdown to select its output device (again see fig.1). I selected the exaSound's eight-channel ASIO driver as DAP's output target, although HDMI was still available. To select my filters, I clicked on each filter tray to bring up a list. Behind the main DAP screen are two others: one lets you adjust the overall throughput gain, and the other shows that filter's channel levels and delays, both of which are adjustable. DAP defaults to –8dB gain to prevent clipping, but you can raise that if needed, and if no clipping results. After this, the user no longer needs to see any of these screens—DAP operates in the background. All you deal with is your normal music-streaming display.

The proof of all this pudding was in the listening, and it was delicious. Most easily obvious was the clarity of the low frequencies, which was accomplished with no loss of power. In fact, bass impact was enhanced in all channels, including the subwoofer. This was particularly evident with B.B. King and Eric Clapton's Riding with the King (from DVD-A, Reprise 47612-9), when I compared the original disc with the file played straight and with DAP correction. The bass slam was the same from the disc and the straight file, but it had more weight with DAP correction, even though there was less low-frequency ambience in the room.

The bass had much more of the impact I associate with a live performance than what I hear from any domestic system. Admittedly, I began with DAP's default target curve, which raises the sub-100Hz range by about 2dB, then slopes down to about –4dB at 15kHz (again see fig.4), but after some fiddling, I had settled on this curve as subjectively optimal. Moving on to my favorite Willie Nelson file, ripped from Night and Day (DVD-A, Free Falls/Surrounded-By SBE-1001-9), the bass was similarly improved. Moreover, with this very immersive surround mix, all of the performers were more discretely positioned around me, and yet the ambience of the performance space was smoothly continuous between them. Individual instruments, easily discriminated among because of their distribution, had a more lifelike presence. It was addictive.

Dirac's Live Room Correction Suite has hooked me. Yes, it's limited to 24-bit/96kHz PCM resolution (24/192 is promised), but no matter. More experiments, comparisons, and independent measurements will follow—but for now, Dirac LRCS has made me confident that file-based multichannel playback can sound as good as any from physical discs—and maybe even better.