Wadia 27 Digital Decoding Computer (DAC) & Wadia 7 CD Transport (part 2)

Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting





<< part from part 1

LISTENING

My first experience with the 27 was with it driving the VT1SOs without the SFL-2 preamplifier in the signal path. The transport was the Mark Levinson No.31 feeding the 27 via glass-fiber (ST type) optical output. After a day of warmup, I put on Mighty Sam Mc Clam’s Give It Up to Love (AudioQuest AQ-CD1O15). Seconds into the first selection, I knew the 27 was something special. There was a magic to the presentation that made the music come alive. The more I listened, and the wider the variety of music I played, the more my appreciation for the 27 grew.

Specifically, the 27 presented a palpability and immediacy I hadn’t heard in my system before. The 27 wasn’t for ward and up-front in a way that gave the false impression of presence but quickly became fatiguing. The sense of palpable presence was instead created by the 27’s extraordinary transparency, resolution, and natural reproduction of timbre. I also heard an openness that was the antithesis of veiled, thick, or congealed. Instrumental images were right there in the room, vivid and alive.

Although these qualities were apparent on all music, the 27-excelled at re producing acoustic guitar Two reference-quality acoustic guitar recordings that are also great musically are Ralph Towner’s instrument on Oregon’s Beyond Words (Chesky JD130), and “Lonesome Road” from the previously m Mighty Sam McClain CD. If the system has any veiling, thickness, or slowness, the guitar loses its edge and life. If the component under evaluation—is hard, bright, and etched, the guitar becomes metallic and an affront to the ears. The 27 struck a perfect balance between resolving the bright transient attack of the guitar and sounding brittle. This combination of vividness without etch is a rare quality. In short, the 27 produced a greater sense of the guitar existing between the loudspeakers than I’ve heard from my system. (Credit for resolving what the 27 could do in this regard goes to the Genesis II 5s, which are extremely quick and clean through the mid-band and treble.)

In a related aspect of its musical performance, the 27 had a wonderful ability to present lots of detail without sounding analytical. Indeed, it is this quality that distinguishes the very best converters from the merely excellent ones. The 27 presented a huge amount of musical information in a completely natural and unforced way. When a processor or CD player gets this aspect of the musical presentation right, you don’t immediately notice it. Rather than call attention to its resolution with a hyped or etched sound, the 27 gently and subtly revealed musical nuances in a way that drew me into the music. I found myself time and again riveted to the musical performance with a total immersion in the musical expression. Products of lower powers of resolution just don’t produce the same sense of surprise at the end of a piece when you return to the real world and realize where you’ve just been.

Although highly resolving, the 27 had a wonderful sense of ease and grace. The treble had a finely filigreed quality that resolved the inner detail of instrumental timbre without becoming analytical. The 27’s upper midrange and treble were extremely smooth and free from glare or metallic hardness. The synthetic treble sound we often hear from digital playback was completely missing from the 27’s rendering. Digital processors often impart a hardness to the treble not unlike that of cheap dome tweeters. Such a presentation calls attention to the treble as a separate component of the music rather than rendering it as an integrated organic whole. The 27 was superb in this regard, which is one reason I enjoyed it so much. I could listen at high levels without fatigue, and never felt a sense of relief after turning down the volume.

Compared with my memory of the Spectral SDR-2000 Pro, however, the 27 had just a trace of grain in the mid range and treble. The Spectral processor also had a more open and extended top octave, with a greater feeling of un limited air and extension. The 27’s treble was more like that of the No.30.5: smooth and detailed, but less open- sounding than that of the SDR-2000 Pro. The 27’s treble was livelier than the No.30.5’s, but the contrast wasn’t enough to make one sound wrong in relation to the other. I enjoyed both presentations.

The overall perspective was more laid-back than those of the Classe DAC-1 and Spectral SDR-2000 Pro, but not quite as relaxed as the Mark Levinson No.30.5’s. The 27 put the first row of musicians at a point just forward of where the No.30.5 positions them.

Spectacular is the best word to de scribe the 27’s soundstaging. The processor threw a wide and deep presentation with exceptional definition, and recreated a wider soundstage than I’ve ever heard from the Genesis II.5s. The entire presentation spread out later ally, giving instrumental images and the air between them more room. My listening notes describe the 27’s soundstage as “W—I—-D—E.”

Soundstage depth was also well portrayed; the 27 not only produced a tremendous sense of distance, but presented instrumental images along a continuum from front to rear. The sound- stage was infused with a huge sense of size and transparent air around image outlines. Even when the music got loud and complex, the 27 maintained its exceptional soundstage clarity and definition. I heard none of the common tendency for digital processors to congeal images and shrink the soundstage during dense, high-level passages. Moreover, the 27 presented its soundstage against a dead-quiet and black background. The 27 produced a wide contrast between the music and the deep black silence.

In comparison with the No.30.5, however, the 27’s soundstage didn’t have the last degree of depth revealed by the Levinson processor. The ’s stage was wider but slightly less deep; the Levinson produced a greater sense of distance to the very rear of the soundstage. The No.30.5 was perhaps helped in this regard by its more laid-back spatial perspective, which set the entire sound- stage farther back than the 27’s slightly more immediate rendering.

‘When I first listened to the 2000 back in 1989, I was struck by the processor’s tremendous bass solidity impact, and effortlessness. That quality has been extended in the 27; the new processor had stunning power and dynamic impact. Kick-drum had a “center-of-the-earth” solidity that offered a firm rhythmic foundation for music. Bass guitar had a sense of power, weight, and drive that worked with the kick-drum to produce a thrilling, visceral experience. Commendably, the 27 had tremendous bass weight without being thick or overblown. The palpability noted ear- her was also evident in the bass; listen, for example, to Stanley Clarke’s solo on “Nevermind,” from Stereophile’s Test CD 3. Stanley was in the room. The bass had wonderful weight, articulation, and dynamic expression. I also enjoyed Patrick O’Hearn’s virtuoso acoustic bass playing on the amazing instrumental “The Ocean is the Ultimate Solution,” from Frank Zappa’s Sleep Dirt (Rykodisc RCD 10527). “Weight without bloat” sums up the 27’s bass presentation.

The remarkable bass dynamics ex tended into the rest of the spectrum. The 27 not only had a good dynamic agility, but presented a more coherent portrayal of transients. The edges seemed to line up in a way that gave the presentation a more lifelike quality. This character was partly responsible for the realism noted earlier on acoustic guitar. When the sharp transient edge was re produced correctly, the sound gave my brain fewer clues that I wasn’t hearing a live instrument. The result was a greater sense of realism.

One aspect of the 27’s performance that’s hard to describe is the sense of music-making I experienced when listening through it. There was a natural ness and vitality to the music that made me forget about the sound and focus on the music.

An Unfair Advantage?

I found myself enjoying music so much through the system that I didn’t want to change any thing. I had to discover, however, how much of the fabulous sound was due to the 27 and how much was from removing the SFL-2 preamplifier from the signal path.

As good a preamplifier as the SFL-2 is, putting it in the signal path degraded the sound. The soundstage became slightly narrower, less deep and spacious, and the bloom around instrumental out lines diminished. The soundstage narrowing was apparent on “Les Douzilles,” from Oregon’s Beyond Words. The oboe, which hung in transparent space at far right, moved a little closer to the center with the SFL-2 in the signal path. It also came slightly forward, lost some of the air that had surrounded it, and was less vivid in timbre. On the same track, the acoustic guitar sounded “slower” and thicker, with a somewhat blunted transient attack. The top octave was also less airy, open, and extended with the SFL 2. The system still sounded superlative, but the magic was gone.

It could be argued that the 27 didn’t really sound as good as I’d experienced it driving the VT15Os directly; it had the advantage of no preamp in the signal path — a condition not enjoyed by other processors under audition. (The comparisons with the No.30.5 were made with both units driving the SFL-2 at matched levels.) Conversely, one could assert that because the 27 has the ability to drive a power amplifier directly, that advantage is intrinsic to the 27’s performance. Why should it be burdened with a preamplifier when it doesn’t need one?

I’ll accept the second argument, but also reiterate the functional limitations of having no preamplifier. If you want to play analog sources you must either digitize them and accept the degradation that that imposes, or reconnect your preamplifier every time you want to listen to a record or the tuner or make a tape. There’s no free lunch.

But once you experience the removal of that last veil between you and the music, it’s hard to live without it. Running the 27 straight into the power amplifiers took me a significant step closer to the music — the last step between great sound and goosebump-raising magic.

My reservations about digital-domain volume controls have been greatly allayed by my experience with the 27. Whatever degradation the digital volume control imposed, it was far less than that introduced by even such an excellent preamplifier as the SFL-2. More over, I heard no degradation when the volume control was used toward the upper end of its range (numbers above 75 on the 27’s front-panel display).

Enter the 7 CD transport:

These impressions were made with the Mark Levinson No.3 1 driving the 27. Would the sound get even better with the $12,500 7 transport? Considering the 7’s apparent cost-no-object design approach and the fabulous Esoteric P-2S mechanism, I had high hopes.

Although the 7 was superb in many ways, I preferred the No.31. The Levinson transport sounded smoother, more liquid, and produced a greater sense of ease. The 7 was more forward overall, with a livelier mid-band and a brighter treble. Piano sounded a little more mechanical through the 7 than through the No.31, with a trace of hardness not heard through the Levinson transport. The No.31 was, overall, more refined and polite.

To its credit, the 7 had deeper bass extension and better bass definition than the No.31. The transport’s bottom end was fuller and tighter with greater rhythmic drive. The 7 had a powerful and dynamic presentation, particularly in the bass.

I also heard precise image focus and exceptional transparency from the 7. Images were well-defined, razor-sharp, and presented against a jet-black back ground. Although the 7 had good soundstage width and transparency the No.31 was more spacious and presented a richer bloom around image outlines. The 7 sounded drier, with less space between images.

Detail resolution was superb; the 7 had a highly detailed rendering that presented me with lots of musical information. The detail was, however, more analytical than the No.31’s more subtle (but no less resolving) perspective. This, along with the smoother treble, gave the No.31 its greater sense of ease and musicality. That’s why the No.31/ 27 match was so musically satisfying.

Although the No.31 suited my tastes more than the transport had the best dynamics, bass definition, and rhythmic coherence of any transport I’ve auditioned. The 7/27 combination had tremendous dynamic impact and drive that I greatly enjoyed. But I was willing to give up some of the 7’s visceral, physically involving pace for the No.31’s greater musicality.

MEASUREMENTS

My measurements of previous digital processors have revealed unusual behavior. In addition to the nearly 3dB rolloff at 20kHz (a tradeoff in the filter to achieve perfect time-domain performance), I’ve measured poor low-level linearity, high levels of noise above the audio-band, and a noise floor that shifts in level and spectral balance as a function of input level. Interestingly, the linearity errors measured in the original 2000 could be corrected with software; sent me different EPROMs that resulted in better low-level linearity but made the unit sound less good.

I thus approached the 27’s measurements with particular interest: Why should poorer bench performance result in better sound? That question, however, remains rhetorical [limited to single-ended triodes]; the 27 exhibited superb technical performance in every parameter.

In the gain position I selected for matching the 27’s analog output level to the rest of my system, the 27 produced 2.45V from the balanced outputs and 2.48V from the single-ended jacks with a 1kHz, 0dBFS sinewave. The similar out put level between the unbalanced and balanced jacks makes sense because the two outputs can be used simultaneously when driving a system (such as mine) that uses a separate woofer amplifier. The 27’s output stage uses separate buffers for the balanced and unbalanced outputs.

This output voltage was sourced from an impedance of 7 ohms (unbalanced) and 14 ohms (balanced) at any audio frequency. Although this is a very low out put impedance, it isn’t the 1 ohm claimed by . DC levels were minimal, and the 27 doesn’t invert absolute polarity. The unit locked to and properly decoded 32kHz and 48kHz sampling frequencies.

As expected from the Digimaster filter algorithm, the 27’s frequency response (fig.1) shows a rolloff of nearly 3dB at 20kHz. The actual -3dB point is 22.05kHz, half the sampling rate. Although this rolloff is confined to the top octave, you’d expect it to be audible as a slight loss of air. I didn’t notice any “closed-in” character during the auditioning, but the 27 did lack the sense of top-end extension I hear from the Spectral SDR-2000 Pro. Fig.1 also shows the 27’s de-emphasis error, which is surprisingly large considering that the de-emphasis is performed in the digital filter. There’s a rise of 0.15dB over about two octaves tight in the region where the ear is most sensitive to amplitude changes, followed by a treble rolloff. This combination of de-emphasis response errors will make pre-emphasized discs sound a little more forward and bright, and reduce their treble ex tension. Because so few CDs have been recorded with pre-emphasis, this won’t be significant.

The 27 had superb separation between its left and right channels, as seen in the crosstalk plot of fig.2. Note that the scale had to be expanded to —140dB (from our usual -130dB) to show the 27’s low crosstalk. The unit had nearly 110dB of channel separation at 20kHz, and 123dB separation at 1kHz —impressive performance.

A spectral analysis of the 27’s output when decoding a 1kHz, -90dB dithered sinewave (fig.3) shows the 27 has an extremely low noise floor, good linearity, and a complete absence of power-supply noise in the audio signal.

We often see peaks in the trace at 601-Iz or 120Hz that indicate the power supply is polluting the audio signal path. There’s not even a hint of this phenomenon in the 27. The peak at 2kHz is second-harmonic distortion in the DAC, which we see in all processors using the Burr-Brown PCM63 or PCM1702 DAC.

Repeating the spectral analysis with an input signal of all zeros and a wider bandwidth (fig.4) reveals that the 27 is well-behaved above the audioband. The newer products use more aggressive ultrasonic filtering than previous de signs, which should reduce system-to- system sound-quality variability. Power amplifiers respond differently to ultrasonic garbage, making products with ultrasonic spuriae less sonically predictable in different systems.

The 27’s linearity (fig.5) was also excellent, with near-perfect behavior to -110dB. The PCM1702 DACs are inherently linear and won’t go out of adjustment over time due to their “co linear” architecture.

Fig.6 shows the 27’s reproduction of a 1kHz, -90dB undithered sinewave with 16-bit precision. The DACs per form well at this low level, with no differential nonlinearity. The 27’s low noise is also apparent in this waveform. Fig.7 is the same waveform, but with 20-bit precision; the 27 gets close to correctly reproducing the 20-bit waveform.

The 27’s noise-modulation performance was outstanding. In flg.8 we can see a virtually perfect overlapping of the traces, indicating that the 27’s noisefloor remains constant regardless of input level. This is one of the best-looking noise-modulation plots I’ve seen, and a far cry from the original 2000’s poor noise-modulation performance. There is, however, an indirect relationship between good low-level linearity and good performance on the noise-modulation test. The better linearity inherent in the PCM 1702 DACs confers an advantage not enjoyed by the mid- 1988—era DACs in the original 2000.

Fig.9 is the 27’s intermodulation distortion spectrum, made by driving the 27 with a hill-scale mix of 19kHz and 20kHz —a real torture test. The IMD products are virtually nonexistent, even the usually strong 1kHz component (20kHz minus 19kHz). Again, this is exemplary performance.

We stopped publishing the square- wave responses of digital processors in Stereophile because they all tended to look alike. The 27, with its time- domain—optimized filter, is a different story. I therefore captured the 27’s re production of a 1kHz full-scale square-wave (fig.10) and the squarewave as re produced by a processor that uses the common NPC 5803 filter (fig.11). Note the complete lack of ripple in the filter By contrast, the NPC filter shows both continuous “Gibbs Phenomenon” ripple (due to the filter not having infinite bandwidth), and that the ripple is clipped. One of the tricks used in the Pacific Microsonics PMD 100 HDCD filter/decoder is to attenuate all signals by 1dB to prevent the ripple from clipping.

I also looked at the 27’s impulse response, which confirmed ’s claim of no pre- or post-echo. An impulse is a signal of all zeros with one sample at hill-scale. Putting this signal through a device under test produces the device’s impulse response, from which frequency-domain information can be obtained using the Discrete Fourier Trans form. Processors with conventional filters produce squiggles in the signal be fore and after the impulse, the so- called pre- and post-echo. It’s odd to think of an artifact appearing before the signal that created that artifact, but the delay through a digital filter makes this seemingly impossible phenomenon a reality.

The musical significance of the 27’s unique time-domain performance is open to debate. I did hear, however, one quality in the 27 that I haven’t heard be fore in a digital processor: a greater sense of palpable presence — “more there there”. Whether or not that quality is related to the 27’s radically different filter and its time-domain—optimized performance is an interesting question.

Because the 27 has no RCA coaxial in puts, to measure jitter I connected the RCA output of a PS Audio Lambda transport to the 27 through an RCA-to BNC adaptor. For consistency, we use the same transport and digital cable for all of our jitter measurements. But introducing an adaptor can significantly worsen the processor-under-test’s jitter performance, something I discovered in my review of the Resolution Audio Quantum processor and Cesium transport. Is it fairer to measure the 27’s jitter with the same cable that other processors are measured with and have it suffer from the RCA-to-BNC adaptor, or to drive the 27 with the Lambda’s ST-Type optical glass-fiber output?

I measured the 27 both ways. How ever, I believe that the results made with the glass-fiber input are more representative of the 27’s intrinsic jitter performance. Most users will opt as I did, to use the 2 glass-fiber input. Don’t even consider using an RCA-to-BNC adaptor.

The following measurements were made on the 27’s 16x word clock (705.6kHz) rather than on the 8x (352.8kHz) clock found in most digital processors. The 27’s custom filter re samples at 16x, not the 8x optional filters.

Fig.12 is the 27’s jitter spectrum made with the coaxial cable and RCA-to BNC adaptor. The input signal was a 1kHz, -90dB sinewave. Note the correlated jitter at the test-tone frequency of 1kHz and its harmonics. The RMS jitter level, measured over a 400Hz - 20kHz bandwidth, was 65 picoseconds. This is a respectable RMS figure, but the spectrum has strong periodic jitter components correlated with the input signal. In other words, the data representing the 1kHz sinewave caused the 27’s clock to be jittered at that frequency—not a good thing.

Now look at the spectrum under identical conditions, but with the 27 fed from the Lambda’s ST-type glass-fiber output (fig.13). The spectrum is virtually free of correlated jitter, save the tiny jitter component at 1kHz. Although we have much to learn about relating clock-jitter spectra to sound quality, I suspect that the difference between figs.12 and 13 is audible, and responsible for the much better sound quality I heard through the glass-fiber interface. The RMS jitter dropped only slightly, to 50ps. The difference between 65ps and 50ps may not be significant, but the difference in jitter spectra certainly is.

With a 1kHz fill-scale sinewave and the ST connection, the 27’s clock-jitter spectrum was completely free from correlated jitter (flg.14). The RMS jitter level was 4Sps (130ps with the coax adaptor). Driving the 27 with a signal of all zeros (which introduces no signal- correlated jitter, but shows the processor’s intrinsic jitter) produced the plot of fig.15. Again, the spectrum is clean. The RMS jitter level dropped even further, to just 26ps (also 26ps with the coax adaptor).

This is outstanding jitter performance, with clean spectra, low R1v1S levels, and excellent isolation of the clock from data-pattern—induced jitter. In fact, the 27’s measured performance was text book in all respects. The unit had a very low noise floor wide channel separation, freedom from power-supply noise in the audio circuits, excellent linearity, superb time-domain performance as seen in the impulse response and squarewave reproduction, and low jitter. In short, the 27 is an extremely well-executed piece of engineering.

The 7’s tracking performance, checked with the Pierre Verany Test Disc (PV.788031&2), was okay but not outstanding, the sound breaking up on track 34, which has a gap in the data 2mm long. Other high-end transports, the Meridian 500 and the Mark Levinson No.31, for example, have per formed much better on this test. While there is no immediate correlation between a transport’s tracking ability and sound quality it does mean the 7 might not be as tolerant of discs with damaged surfaces as other high-end transports. It was also surprising, given the construction and design of the Esoteric P-2S mechanism used by the transport. (Perhaps the transport was out of alignment.)

CONCLUSION

’s 27 Digital Decoding Computer has joined a select group of digital processors I hold in the highest regard. In my experience, the three best-sounding converters are the Spectral SDR 2000 Pro, Mark Levinson No.30.5 (with the HDCD upgrade and its 6dB of attenuation turned off), and the 27. These three form an upper echelon a step above the other Class-A-rated processors in Stereophile’s “Recommended Components.”

I won’t restate its specific qualities here, but must reiterate how much I enjoyed listening to music through the 27. Driving the power amplifiers directly, the 27 was immensely communicative and involving. Throughout the auditioning I had the sense of sitting in on music-making, with a riveting realism and spontaneity.

The 27’s design, execution, and build quality were all world-class. The impressive engineering techniques that fill the 27 are reflected in its outstanding technical performance.

I was less enthusiastic about the 7 CD transport. Although the 7 had great bass, dynamics, and detail resolution, it fell short of the standard set by the Mark Levinson No.31 transport in smoothness, ease, soundstage depth, and bloom. There’s also the matter of the 7’s $12,500 price tag. For that amount of money, the product should be without peer.

The 27 has earned a strong recommendation, even if you must use it with a preamplifier. But those listeners who can drive a power amplifier directly from the 27 are in for a special treat.


Fig. 1 Wadia 27, frequency response (top) and de emphasis response (bottom) (right channel dashed, 0.5dB/vertical div.)


Fig.2 Wadia 27, crosstalk (R-L channel dashed, 10dB/vertical div.)

Fig.3 27, spectrum of dithered I kHz tone at -90.31 dBFS, with noise and spuriae (20-bit data, V analysis, right channel dashed).

Fig.4 27, spectrum of digital silence (20-bit data, V analysis, right channel dashed).


Fig.5 27, departure from linearity (right channel dashed, 2dB/vertical div.)

Fig.6 27, waveform of undithered 1 kHz sinewave at -90.31 dBFS (16-bit data).

Fig.7 27, waveform of undithered I kHz sinewave at -90.31 dBFS (20-bit data).


Fig.8 27, noise modulation, -60dBFS to -100dBFS (10dB/vertical div)

Fig.9 27, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 0dBFS (linear frequency scale, 20dB/vertical div)


Fig. 10 27, 1 kHz square wave at 0dBFS.

Fig. 11 Processor equipped with NPC 8013 digital filter, 1 kHz squarewave at 0dBFS.


Fig. 12 27, word-clock jitter spectrum, DC-20kHz, when processing 1 kHz sinewave at -90dBFS; PS Audio Lambda transport, RCA/BNC connection (linear frequency scale, 10dB/vertical div., 0dB=1ns).


Fig. 14 27, word-clock jitter spectrum, DC-20kHz, when processing 1 kHz sinewave at 0dBFS; PS Audio Lambda transport, ST optical connection (linear frequency scale, 10dB/vertical div., 0dB=1ns).

Fig. 15 27, word-clock jitter spectrum, DC-20kHz, when processing digital silence; PS Audio Lambda transport, ST-optical connection (linear frequency scale, 10dB/vertical div., 0dB=1ns).

Notes:

2. Adapting centuries-old shipbuilding techniques to run filter algorithms for converting digital bits to analog audio is a perfect example of James Burke’s thesis in his fascinating book and PBS television series, Connections.

3. The Meitner Intelligent Digital Audio Translator (IDAT) solved this quandary by the interesting approach of using two filters: an FIR optimized for frequency-domain performance, and an IIR optimized for optimum time-domain performance. The IDAT’s software looks at the signal characteristics and routes transient signals to the IIR, signals that were more steady-state in nature to the FIR. The $15,000 IDAT both sounded wonderful and measured well, with perfect square-wave reproduction and impulse response, along with flat frequency response.

Reviewed by Robert Harley (courtesy Stereophile, Oct 1996); Related articles at Stereophile: 2002 Follow-up

Home


Updated: Wednesday, 2019-07-10 8:27 PST