Wadia Digimaster X-32 digital processor Page 3
In essence, the staggered DAC array linearly interpolates three additional points on the waveform in the analog domain between the seven oversampled points calculated in the digital domain by the DSP chips and the "DigiMaster" software. The resultant output signal has 32x the number of samples (8x4), but with some of the samples created by computer calculation (with software) and some by the DACs (with hardware).
The X-32's eight DACs (four per channel) are custom-made to Wadia's specifications. Rather than perform I/V (current to voltage) conversion on a separate chip (usually an op-amp) as is typical, I/V conversion takes place in the DAC. Since the four DACs must essentially operate as a unit, matching is critical. Wadia claims the DACs within each array are matched to within a tenth of an LSB (Least Significant Bit). In addition, DAC consistency is critical since a single MSB trimmer adjusts four DACs. Any performance differences would show up as poor low-level linearity since the MSB trimmer can be set optimally for only one DAC.
In the final analysis, how a product achieves its technical goals is unimportant: what matters is how it reproduces music. However, I have gone into this discussion to dispel any confusion about marketing claims of high oversampling rates.
After all this technical discussion, I'm ready to ditch my plastic shirt-pocket protector and get down to some music.
When comparing equipment, precise level matching is vital to reaching accurate judgments. Before auditioning digital processors, I play the 1kHz, 20dB (peak) tone on the Stereophile Test CD and measure its voltage output at the power amplifier's input terminals. This procedure is repeated for the processor to which it will be compared. I then adjust the stepped level controls on the EVS Stepped Attenuator until the output levels of both processors are identical. The number of level steps on the EVS are counted, allowing precise matching during auditioning. If the EVS steps do not permit accurate level matching, I use the same procedure with the EVS Penny & Giles Attenuator, an infinitely variable attenuator. Usually, however, the Stepped Attenuator's resolution is sufficient. Coincidentally, the X-32's output level with this signal was identical to the Wadia 2000's level (89mV) and very close to the Stax X1t's output level (87mV), which simplified the auditioning. In addition, the Theta DSPro Basic's output level was exactly 89mV with eight attenuation steps. Of course, identical interconnect types and lengths connected the processors to the EVS passive control unit.
Comparing the X-32 to the 2000, to which I have listened extensively in the past five months, was interesting. It was apparent that the two processors share a similar sonic signature as well as technology. I was surprised by just how close the two processors' presentations were. The characteristics one expects from high-end digital processorstextural liquidity, soundstage depth and transparency, lack of high-frequency hashwere all immediately apparent in the X-32. The differences between the 2000 and X-32 were quantitative rather than qualitative. The X-32 had the ability to successfully convey the music much the way the 2000 does. Not surprisingly though, I found the 2000 more musical than the X-32.
First, the X-32 had a more forward presentation with less of the sense of depth and air that characterizes the 2000. The X-32 threw just as wide a soundstage, but didn't separate instruments in the front-to-rear perspective quite as well. The sense of images floating in air and surrounded by space that makes the 2000 so enjoyable was present in the X-32, but to a lesser extent. Instruments tended to be grouped together toward the front of the soundstage, imparting a slightly drier rendering. Textures tended to be slightly harder and with less liquidity. Low-frequency drive and dynamic impact were excellent through both processors, but the nod has to go to the 2000. There was, however, a distinct similarity in overall musicality between them that was surprisingly close considering the price differential. Considering the cost differences, the fact that the X-32 was able to hold its own was remarkable.
One should keep this description in perspective: the 2000, at $7800, costs nearly four times more than the X-32. I felt, however, that a comparison was warranted since the X-32 is, in essence, a scaled-down version of the 2000; potential buyers should be aware of their sonic differences. No, one doesn't get the performance of the 2000 for the price of the X-32.
The X-32 fared better in a comparison with processors in its price range and below, offering substantially better performance than every less-expensive processor I've auditioned. The X-32 goes a long way to reducing many of digital playback's problems so apparent in lesser machines.
Although I wouldn't characterize the X-32's treble presentation as sweet, it was sufficiently devoid of grain and hash to make music listenable over extended periods. The overall tonal character was a little on the bright side, with a forward presentation. For example, cymbals assumed a more prominent position at the front of the soundstage, but never sounded spitty or harsh. Compared to the Proceed PDP, favorably reviewed in Vol.13 No.6, the X-32 was livelier and more immediate.
I attribute the X-32's forward presentation both to the marginally bright treble and to a slight coloration through the midrange that sounded euphonic on some recordings and instruments and unpleasant on others. The midrange had a slightly veiled, dark character in relation to every other processor to which it was compared. Female vocals just didn't sound pure and unencumbered. Compared to the astonishing midrange purity of the Stax DAC-X1t, this was glaring. Next to the 2000 and the Proceed PDP, however, the X-32's slight midrange veiling was much less apparent, though inescapable. These characteristics reduced the sense of ease and liquidity heard through the Stax and 2000.