Skip to Content

Stereophile 評鑑Lavry Gold(MidiMall在庫展示中)

Lavry Engineering DA2002 D/A processor

John Atkinson, August, 2004





Sony Classical's head of engineering, David Smith, is a man whose opinions on sound quality I have come to respect. So when David e-mailed me a year or so back, enthusing over a new DAC he'd heard, I paid attention. When Lavry Engineering contacted me about reviewing their DA2002, I didn't need much persuading.

Lavry...
I was familiar both with Lavry Engineering's pro-audio components, under its older name of dB Technologies, and with designer Dan Lavry, from his often outspoken, tech-heavy postings on the Internet newsgroup rec.audio.pro. You can find theoretical white papers written by Dan on the Lavry website, discussing such things as word-clock jitter (he's against it) and high sample rates (he's against those, too, at least as far as quadruple rates such as 176.4kHz and 192kHz are concerned).


The DA2002 is Lavry's first consumer product, and is housed in a small chassis with a gold-plated front panel. It incorporates Dan's lateral thinking on audio circuit design (again, see the Lavry website), and handles sample rates from 44.1 to 96kHz. Both balanced and unbalanced analog outputs are provided, and absolute polarity can be inverted with a front-panel button.

...Engineering
The DA2002's power supply is based on two small toroidal transformers. These are mechanically damped by being clamped by the top panel, and are followed by some beefy rectifier diodes, the usual reservoir capacitors, and four three-pin regulator chips that use the chassis as a heatsink.

The three data-input jacks are each buffered by a small transformer. The clock circuit uses two crystals, one for 44.1kHz/88.2kHz data, the other for 48kHz/96kHz data, and these appear to be under the control of an Analog Devices 12-bit DAC, which matches the rate at which the incoming data are clocked out of the jitter-reducing buffer to the long-term average of the incoming datastream. The data are then processed by a Motorola DSP56002 DSP chip, which I assume handles the low-order upsampling and the digital low-pass filtering.

The signal path downstream of the DSP chip is complex. The only DAC chips I could see were a pair of Analog Devices AD7538s. This is a 14-bit part, according to its data sheet; the DA2002's manual states that these are used to calibrate the DACs. So where are the real DAC chips?

The key to the DA2002's D/A conversion is actually buried beneath a power resistor. Rather than use an off-the-shelf DAC part, Dan Lavry uses a custom-made network of laser-trimmed thin-film resistors; the big resistor acts as an oven, heating the network to its working temperature. Each time the DA1002 is powered up, this network is calibrated using the AD7538 DACs and multiplexer chips, using, I imagine, data stored in the adjacent EEPROM chip.

The end result of this activity is that, somewhere within this forest of chips—the resistor network is embedded in an array of Burr-Brown OPA177 bipolar op-amps and Analog Devices AD744 BiFET op-amps—an analog signal that accurately corresponds to the input data emerges, and is fed to a seven-pole analog reconstruction filter and the DA2002's output stages. Burr-Brown OPA627 Difet op-amps (driven by the crystal oscillators mentioned above) are used as sample-and-hold deglitchers ahead of the balanced filter and output circuitry, this using discrete transistors as well as ICs and carried on a small double-sided daughterboard. The two unbalanced outputs each appear to be realized with a Burr-Brown OPA134 Sound Plus op-amp chip.

Sound quality
When I first turned on the Lavry DA2002, it took about five minutes to decide that all was okay with its operating parameters. During this process, first the top two orange sample-rate LEDs flashed, indicating that the DAC resistor network had not yet reached its equilibrium temperature. When it had, the bottom two orange LEDs flashed, indicating that the DA2002 was going through its DAC calibration routine. At the end of this process, the DAC scrolled through its green input LEDs until it found a source carrying valid data, at which point the appropriate input LED remained lit. When presented with two sources, the input button can be used to select the desired source in the usual way. It also has a Mute position.

Out of the box, the DA2002's CrystalLock® jitter-rejection buffer is in-circuit. This can be switched out or in by holding the Polarity button for 1.5 seconds; I did all my auditioning with CrystalLock engaged.

The first disc I played was the Hi-Res Music DVD-Audio of the Ray Brown Trio's Soular Energy (HRM 2011). This disc is unusual in that it allows an in-the-clear 24-bit/96kHz datastream to be transmitted from the DVD player's digital output (if it has one that will handle 96k data). The Lavry had no trouble locking to my Technics DVD-A10's digital output via a 10' length of AudioQuest's VSD-4 S/PDIF cable—and yes, while I admit this disc was mastered from an analog original, there was an analog-like "ease" to the Lavry's presentation. The sound of Ray Brown's double bass had a satisfying purr to its leading edges, yet the upper register of Gene Harris' piano didn't sound too etched.

The Technics has overall proved a disappointment as a disc transport since I purchased it at the end of 2000. Yet it proved more than adequate feeding the Lavry DAC, which seems to suggest that the DA2002's data receiver is effective at cleaning up the timing of the data fed it. The result was an almost "tubelike" midrange smoothness to the Lavry's sound.





As I write these words, I am about to drive to Goshen College in Indiana to record my fourth CD for Cantus, the unaccompanied male choir from Minnesota. One thing I have learned from my previous three recordings of the group is that when singers let it all hang out, lesser CD players and digital processors lend the mid-treble a shouty quality, almost a "rattle." Some of this is real—I hear it when I'm in the hall with the singers. But with the Lavry fed by a Mark Levinson No.31.5 transport, my 2003 Deep River CD (Cantus CTS1203) sounded as sweet in this respect as Cantus had live. Perhaps more important, each of the 11 singers could be clearly heard to occupy his own place in the soundstage. Nice, very nice.

Comparisons
All comparisons were performed with the levels matched to within 0.1dB at 1kHz. First up was my longtime reference, the Levinson No.30.6, which cost $17,500 when last available. Playing Soular Energy again, the Lavry DA2002 didn't have the low-bass reach or ultimate authority of the No.30.6, but Ray Brown's bass was presented both a bit more forward in the soundstage and with a slightly more palpable sense of "being there." The Levinson's image was flatter, with less depth and less lower-midrange bloom. Again, I'm forced to use the simile "tubelike" to describe the DA2002's character, which I preferred.

In the higher frequencies, I was hard-pressed to tell the two processors apart. Carol Wincenc's flute on my Mozart Flute Quartet CD (Serenade, Stereophile STPH009-2) sounded equally sweet through both, with the slight chiff of her breath on the notes' leading edges neither exaggerated nor mellowed. Even on raucous modern rock mixes, such as our June 2002 "Recording of the Month," Wilco's Yankee Hotel Foxtrot (CD, Nonesuch 79669-2), it was hard to tell the two processors apart, other than by the differences in bass and soundstage presentation mentioned earlier.

Next on deck was the sample of the Benchmark DAC 1 that I'd purchased after reviewing it in the May Stereophile (p.117). Like the Lavry, the Benchmark emanates from the pro-audio world, and offers superb sound quality and superb value at its $975 price. Playing the Mozart Flute Quartet again, the Benchmark and Lavry DACs were very close in tonal character. Perhaps the breath-noise chiff was a little more forward in the soundstage with the Benchmark, the treble balance a little harder.

In the low frequencies, there was basically no difference that I could detect between the Lavry and the Benchmark. However, the two processors became a little easier to distinguish with my Deep River CD. As I had found with the Levinson comparisons, the Lavry's lower midrange had more bloom compared with the Benchmark. There was a slightly better sense of "being there," in that the unaccompanied men's voices were more solidly positioned in the hall acoustic.

The bloom wasn't always a benefit. When I mixed Stereophile's Rendezvous CD (STPH013-2), I ended up balancing Jerome Harris' Taylor acoustic bass guitar a little lower in level than he or I were anticipating. Musically this made sense, but the instrument's image is a little fragile as a result. The Lavry's bloom robbed the bass guitar of some of its leading-edge energy compared with the Benchmark. Not unpleasant, and not something I perhaps would notice except in A/B comparisons, but I suspect that if I'd used the DA2002 when mixing the album, I would have raised the bass guitar level slightly.

My final comparison was with the $12,000 Nagra DAC that Kalman Rubinson reviewed in May (p.85). Dr. Kal had enthused over the Swiss DAC's clarity and powerful dynamics; when I plugged the Nagra into my system and hooked up the Levinson CD transport to its AES/EBU input with a DH Labs Silver Sonic datalink, I could immediately hear why it had impressed him.

But that doesn't mean it outclassed the Lavry DA2002. Yet again, I had to sweat to perceive any meaningful differences at all between the two DACs. It was my Mosaic CD (Stereophile STPH015-2) where I thought I had it down: in the slow movement of the Brahms Clarinet Quintet, soloist Antony Michaelson barely breathes the notes. The Nagra was ever so slightly better at defining the space around the image of the clarinet, as each note slid into the next. But in other respects—Ray Brown's bass on the 96kHz Soular Energy, the voices on Deep River—it was a wash.

Problems
I had two minor problems during my time with the DA2002. The first was probably a design flaw: It didn't mute when fed a Dolby Digital bitstream. This was a disadvantage when playing DVD-Video discs, which may have high-quality LPCM tracks for the music but sometimes use a Dolby track for the menus. The result was a disturbing blast of pulsed white noise as I groped for the Mute button on the preamp's remote.

The second problem was that, very occasionally, the DAC would start to overlay zingy-sounding spuriae on the music, as though it were losing lock with the incoming datastream. (The input and sample-rate LEDs remained resolutely lit.) Rescanning the inputs so that the Lavry rediscovered the active one solved the problem, but the fact that it happened at all raised my eyebrows.

Summing up
Impressively engineered and even more impressive-sounding, Lavry Engineering's DA2002 is one of the few digital processors that I could live with in the long term. Yes, its inability to gracefully handle Dolby Digital bitstreams and its lack of de-emphasis should be addressed, and a version with a volume control would cause my audiophile cup to run over. A TosLink input would also maximize its usefulness. But, as it stands, I recommend the Lavry DA2002 very highly. It's definitely a high-end contender, at a relatively affordable price, from an unexpected source.


Sidebar 1: Specifications

Description: D/A processor with balanced and unbalanced analog outputs; two transformer-isolated AES/EBU balanced digital inputs; and one transformer-isolated unbalanced S/PDIF digital input. Sample rates: 96kHz, 88.2kHz, 48kHz, 44.1kHz at ±150ppm lock range. Crystal lock tracking: 1ppm/15s, 40-50kHz wide lock mode (varispeed). Maximum output level: 4.5V RMS into 100k ohms (balanced XLR), 2V RMS into 100k ohms (unbalanced RCA). Frequency response: 10Hz-20kHz, ±0.05dB. Phase linearity: 2 degrees, 10Hz-20kHz. Channel separation: 100dB at 1kHz. Signal/noise: 110dB RMS, 130dB peak spurious response. Distortion (1kHz tone at -1dBFS): 0.0009% peak harmonic amplitude. Output impedance: not specified. Power consumption: 20W.
Dimensions: 17" (432mm) W by 2" (51mm) H by 11" (280mm) D. Weight: 10 lbs (4.5kg).
Finish: Black chassis with 24k gold-plated front panel.
Serial number of unit reviewed: 50686.
Price: $8500. Approximate number of dealers: factory-direct only.
Manufacturer: Lavry Engineering, 945 Hildebrand Lane NE, Suite 110, Bainbridge Island, WA 98110. Tel: (206) 842-3552. Fax: (206) 842-3193. Web: lavryengineering.com.