Digital Processor Reviews

A decade or two ago, I stumbled on a surprising demo room at an audio show. I don't recall most of the equipment, but I do remember a pair of Paradigm Studio 20 speakers at one end, their crossover entrails dangling free, connected to the rest of the system by a multiplicity of wires. At the other end, among the usual electronics, was a PC whose screen was a crazy quilt of graphs and menus that constantly twinkled in response to the ministrations of DEQX's Kim Ryrie. He seemed totally absorbed, but looked up and proudly offered to show me what he was doing. When I told him that I was familiar with the Paradigms, he played some music that sounded just fine. Then he clicked his mouse. The sound was transformed from the familiar to the fabulous. I was dumbfounded. "What have you done?"

Ryrie had switched from a normal preamp-amp-speaker circuit to DEQX's crossover/speaker processing in an active biamp setup that bypassed the speaker's own crossover. He then described the limited ability of high-level passive crossovers to smooth the drivers' in-band response and to integrate the drivers' outputs in the crossover region. These problems are further compounded, he said, by the constraints of parts cost and of physically fitting the components into a speaker cabinet of practical size.

It all made sense, and, based on several comparison demonstrations, the results were entirely convincing. I would have jumped on this then, but was deterred by the projected cost—much higher, in today's dollars, than the $4995 price of the subject of this review—the intimidating complexity of the process, and the need to brutally eviscerate my darling speakers. But, most of all, the era of general acceptance of computerized devices in audio systems had not yet dawned.

Since then, Ryrie and I have continued to bump into each other at audio shows. I've learned that DEQX no longer demands the removal of the speakers' passive crossovers to do its magic (although it's still advantageous to do so), and that, yes, it can also do room correction. Each time, DEQX's demos were convincing enough to test my resistance. John Atkinson, in fact, described himself as "gobsmacked" in response to the transformation wrought by DEQX on the performance of a pair of modest speakers.

Now I have one of DEQX's latest products, the PreMate, which embodies the core abilities of DEQX's DSP engine to measure and correct a speaker's frequency response, to construct active crossovers for the integration of multidriver systems, for bass management, and to implement room correction. Combining that with a simple but flexible input/output control and remote control, the PreMate can function as a control preamplifier for a stereo system.

Description
The PreMate's front panel is unremarkable, with only six small buttons and associated LEDs. On the left is the Standby/On button. On the right are a pair of Volume Up/Down buttons, and next to them are three buttons for selecting one (or none) of three user-programmed configurations. There is no input selector; that is well handled by the remote control, or by setting up input priorities so that the PreMate will choose the highest-priority source that is active. So, in a sense, manual input selection may not be needed.

The rear panel is more interesting. The connections are clustered by general function, with RS-232 input/output and control USB ports on the lower left. Above them and to their immediate right are the five digital inputs: TosLink, S/PDIF (BNC), S/PDIF (RCA), AES/EBU, and USB audio. The middle portion has an XLR microphone input—a calibrated Dayton mike is included in the price—and two pairs of analog inputs (XLR and RCA). Above these is the S/PDIF (BNC) output. Toward the right, the PreMate has four pairs of analog outputs (two XLR, two RCA), to serve the main stereo channels and a pair of subwoofers. Alternatively, these can feed the amps to drive the high- and low-frequency drivers of an actively biamplified system. Finally, there are the AC receptacle and main power switch.

In the November 2005 issue, I reviewed NHT's late, lamented Xd active 2.1-channel speaker system, which was a piece of cake to set up because it came with factory measurements of the speakers. However, when I participated in a DEQX setup demonstration at the Manhattan home of former Stereophile contributor Jonathan Scull, it was run by DEQX from Australia, via Skype and VPN. That's when I appreciated the real differences between it and the mostly automatic domestic EQ systems in use today. The fundamental difference is that DEQX first measures each speaker system from a distance only great enough to permit driver/source integration. This makes it possible to approximate anechoic measurements, since the arrival time for the direct sound from the speakers is predetermined by that distance, and easily distinguished from even early reflections of that sound.

I see this as critical. All automatic EQ systems conflate speaker correction with room correction; measurements taken from the main listening position and its immediate neighborhood can't truly distinguish what the speaker is doing from the room's influence. Yes, one can window the time-domain responses, but with the small dimensions of most listening rooms, there is uncertainty, especially in the lower frequencies. Is that a speaker resonance or an early reflection? By excluding, as much as possible, the influence of the room, these early measurements are used to generate filters that will smooth the frequency response, align the time responses of the individual drivers, and, if the option is chosen, create an optimal crossover for active amplification. Indeed, as I've come to appreciate, the magic of DEQX is its ability to make the speaker more perfect than is otherwise possible within the constraints of money, size, and passive circuitry.

It is only after this optimization of the speaker's performance that you separately approach the influence of room acoustics with the DEQX PreMate. Of course, at this time, you and the PreMate know what the speaker sounds like, and when you look at the in-room response, you know how the room's acoustics have influenced it. This will inform your choices about which frequency and magnitude limits can be used for correction without overcorrecting. After all, you want your speakers to sound as they did when you fell in love with them, and you want to hear them untainted by the more pernicious effects of room acoustics.

Speaker Measurements and Calibration
Physical setup of the DEQX is pretty simple. I connected the PreMate to the outputs of a Sony XA-5400ES SACD/CD player and the inputs of a Parasound Halo A 31 three-channel power amp, both via XLR. I also connected the USB outputs of my laptop computer to both the control and audio USB inputs. Finally, I plugged in an Earthworks M23 mike. I loaded the DEQX-Calibration Windows app from the CD-R that is provided with the PreMate and, as I began the "project," chose this mike's calibration curve from the provided library. The software prompted me for information about the room, the equipment in use, and its configuration. The PreMate can handle two full-range speakers, with or without mono or stereo subwoofers; or a single pair of biamped speakers. Mostly, I used it only as an equalizer, without biamping or deriving a sub channel.

I did two calibrations, the first with my B&W 800 Diamond speakers, under guidance, via Skype/VPN, of DEQX down under. After that I flew solo, calibrating a pair of Bryston Middle-T speakers. Late in the game, I ran a series with the Brystons crossed over to a JL Audio Fathom f113 mono sub. After I'd earned my stripes with DEQX, that was fairly easy—but what I describe below is mostly based on the Brystons without a sub.

For the measurements, DEQX advises moving the speakers far away from any large reflecting surfaces and, ideally, elevating them so that their midrange drivers are equidistant from floor and ceiling. The former was a bit of a tussle, but since these speakers are relatively large, I didn't even try to elevate them. The mike was pointed directly at the tweeter of the right speaker from a distance of 1–2m. I opened the DEQX Measurement Wizard and, after setting the number of test sweeps (I chose nine) and a marginally tolerable signal level, I clicked Run. DEQX proceeded to run the tests, analyze the data, and generate a target frequency-response curve. You can then accept or decline the result. The right speaker showed a surprisingly flat response (which, of course, I accepted), but the first run of the left speaker revealed an anomaly at around 150Hz that was due to the proximity of a piece of furniture. This I declined, so that I could reposition the speaker. I reran the test, and the result was a smooth response without the kink.

Next, the DEQX Calibration Wizard steps you through the calibration and correction of the speaker's frequency response and the phase alignment of the output. The first step is to window the time-domain data so that the correction is applied only to the speaker's direct output. This is done by moving a cursor line on a graph of the measured impulse response so that it cuts off just before the easily discernible first reflection. Ideally, this cursor should be about 6 milliseconds from the initial response, to permit anechoic correction down to 200Hz. If it is significantly less than 6ms, it can be corrected by repeating the measurements and 1) placing the mike closer, 2) absorbing the floor bounce by covering the floor between the speaker and mike with pillows, or 3) elevating the speaker and mike so that they're midway between floor and ceiling. Fortunately for me, the delay was about 5.5ms, which is close enough.

Then you examine the resulting frequency response and the target curve, and choose a correction window with the DEQX default set at 200Hz–20kHz. I also added a second calibration with the window extended up to 25kHz (max is 40kHz), because the Brystons have a small peak at >20kHz that I wanted to mitigate; and down to 150Hz, to get a smoother transition to the bass. There are options for maximum group delay and tolerance (in ms), as well as FR tolerance (in dB). I went with DEQX's defaults of 3ms group delay, 0.5ms tolerance, and 0.2dB FR tolerance, which seemed, to me, more-than-acceptable outcomes. Click to create a correction curve, go get a soda from the kitchen, and by the time you've returned the work is done. The PreMate can store four different profiles; I put the two calibrations into Profiles 1 and 2, leaving Profile 3 for experimenting with different room correction, and Profile 0 for bypass. Of course, DEQX lets you store many corrections and projects on your computer for easy download to the PreMate.