AA合并机推雨后超五 [复制链接]

When I wrote that review, I had no way of measuring transport jitter. Now that we can measure the DTI's effect on jitter, let's see how close our listening impressions were to the DTI's measured performance; it may shed some light on how much jitter is audible and what the sonic effects are.
Fig.11 is the SV-3700's jitter spectrum with the DTI in the S/PDIF signal path. The DTI reduced the jitter by a factor of 10, from 4250ps to 450ps (with a 1kHz, full-scale signal), easily seen in the comparison with fig.4 (the SV-3700's jitter without the DTI). The jitter reduction was less apparent, but still significant, with music as the test signal. Compare fig.12 (with the DTI) with fig.5 (without the DTI). The reduction in broad-band RMS jitter was from 3830ps without the DTI to 289ps with the DTI. Overall, we can confidently say the DTI produces a significant reduction in jitter—when fed with a high-jitter source.

[upload=jpg]Upload/20053914572455270.jpg[/upload]
Fig.11 Panasonic SV-3700 DAT recorder with Audio Alchemy DTI, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting digital silence (bottom solid trace), a 1kHz sinewave at -90dB (middle, dashed trace), and a 1kHz sinewave at 0dBFS (top, light dotted trace) (vertical scale, 1ps-2ns, 100µV = 1ps). Compare fig.4. [upload=jpg]Upload/2005391458053468.jpg[/upload]
Fig.12 Panasonic SV-3700 DAT recorder with Audio Alchemy DTI, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting music #1 (Firebird, solid) and music #2 (Steve Morse, dashed) (vertical scale, 1ps-2ns, 100µV = 1ps). Compare fig.5.

But look what happens when we put the DTI in the digital signal path from the PS Audio Lambda, a low-jitter source. Fig.13 is the Lambda's jitter spectrum with the DTI. Compare it to fig.6 (Lambda without the DTI) shown earlier. Now the jitter level increased, from 32ps to 133ps with the DTI (with a full-scale 1kHz test signal). Moreover, the DTI imposed a very similar jitter signature—seen as the curve's characteristic shape—on both the SV-3700 and the Lambda. The effect will be to make the high-jitter source sound better, the low-jitter source sound worse. Further, the DTI's common jitter signature will reduce audible differences between transports and interfaces, making them all sound similar—and relatively mediocre. With music as the test signal, the Lambda/DTI combination produced the plot of fig.14. The RMS jitter level with music was 37ps without the DTI and 129ps with, both measured using musical signal #2. (Note that the 129ps figure is lower than the CS8412's 200ps of intrinsic jitter owing to the 30kHz measurement bandwidth.)

[upload=jpg]Upload/20053914582866421.jpg[/upload]
Fig.13 PS Audio Lambda with Audio Alchemy DTI, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting digital silence (bottom solid trace), a 1kHz sinewave at -90dB (middle, dashed trace), and a 1kHz sinewave at 0dBFS (top, light dotted trace) (vertical scale, 1ps-2ns, 100µV = 1ps). Compare fig.6.
upload=jpg]Upload/2005391501252474.jpg[/upload]
Fig.14 PS Audio Lambda with Audio Alchemy DTI, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting music #1 (Firebird, solid) and music #2 (Steve Morse, dashed) (vertical scale, 1ps-2ns, 100µV = 1ps). Compare fig.7.

In addition to revealing that the DTI quantitatively improves some transports and degrades others, as was suggested by the listening, these measurements suggest something else: That the DTI imposes its intrinsic jitter on its digital output signal, and still passes some of the jitter in the incoming data stream. Note the very much higher jitter at the DTI's output when fed by the high-jitter Panasonic SV-3700 compared to the low-jitter Lambda.

Further, these test results demonstrate that the differences seen in the jitter spectra and levels with and without the DTI are easily audible. As described in my review, the DTI's sonic effects—good or bad—aren't subtle. I'm at a loss to explain why some reviewers have gone on the record to state that the DTI improves the sound of every transport and processor combination—even that of the Mark Levinson No.30 and No.31!

I repeated these measurements using a prototype jitter-reduction box that uses the new Analog Devices AD1890 Asynchronous Sampling Rate Converter chip described in last May's "Industry Update" (Vol.16 No.5, p.41). The prototype box, designed by Bob Katz (Digital Domain, Chesky), greatly reduced the jitter of high-jitter sources (from 4250ps to 45ps—an order of magnitude greater reduction than the DTI), and very slightly degraded the low-jitter Lambda (from 32ps to 42ps). All jitter-reduction devices are, however, limited by the fact that their reduced-jitter outputs must still then pass through another jitter-inducing interface between their output and the digital processor.

On to the transports
For the RMS jitter figures for each transport with each test signal, see Table 1.

Table 1: Transport Jitter (figures in picoseconds) Transport 0dB -90dB M1 M2 Silence Average
Panasonic SV-3700 4250 1110 3830 3800 180 2634
SV3700 w/DTI 450 289 334 278 111 292
JVC XL-Z1010 96 201 116 93 71 115
Meridian 200 44 126 63 44 38 63
Pioneer CD-65 43 126 70 44 32 63
Proceed PDT 3 39 127 70 43 33 62
Denon DP-X 59 80 75 54 30 60
Meridian CDR 36 109 69 38 18 54
Mark Levinson No.31 34 98 66 38 29 53
CAL Delta 35 110 40 34 32 50
Proceed PDT 1 37 81 62 40 32 50
C.E.C. TL 1 30 90 61 33 24 48
PS Audio Lambda 32 51 66 36 29 43
PS Lambda W/DTI 133 158 129 120 139 136

SMERT 在 2005-3-9 5:16:19 发表的内容 海阳 在 2005-3-8 22:12:41 发表的内容 我觉得如果喜欢米各解码的声音,还不如上顶级的米格解码2手也不过5K出. 好像这款T-DAC已经是当年的旗舰了。你说的那台是不是近期在音响发烧站上出的Micromega DUO PRO？T-DAC好像是在比DUO高一级的TRIO停产之后才出来的，应该比DUO时间晚。

呵呵,早不一定就不好,认真听过就清楚那种好,我觉得兄的器材那么高,应该上好点的解码.T-DAC可能刚好匹配兄的器材.个人觉得T-DAC比兄的CD好不了多少.

SMERT何时安排下试听啊？

今晚回去对比

The "bits is bits" camp rejects this thesis, claiming that transport and interface jitter is completely removed by the digital processor's input receiver. They consider the PLL an absolute barrier to jitter. Consequently, they argue, transports, digital interfaces, and CD tweaks can't affect sound quality.

I conducted a little experiment to test this hypothesis. I measured a digital processor's word-clock jitter (with the Meitner LIM Detector described in Vol.16 No.1) when driven by two different digital sources. One source has low jitter (the PS Audio Lambda transport), and one source has high jitter (the Panasonic SV-3700 professional DAT machine). Fig.2 shows the jitter spectrum of the processor's word clock when driven by the Lambda. For contrast, fig.3 is the same processor's jitter spectrum—measured at the DAC with the identical test signal and conditions—but with the high-jitter Panasonic SV-3700 driving the processor. Note the vastly cleaner spectrum and fewer discrete-frequency jitter components when the processor was driven by the Lambda. Moreover, the overall RMS jitter (measured from 400Hz to 22kHz) increased from 145ps with the Lambda transport to a whopping 561ps when driven by the high-jitter SV-3700. Clearly, jitter in the S/PDIF signal driving a digital processor does greatly affect word-clock jitter inside the processor.

[upload=jpg]Upload/20053914314358854.jpg[/upload]
Fig.2 PS Audio Reference Link, DAC word-clock jitter spectrum, DC-20kHz, when driven by PS Audio Lambda CD transport (linear freqeuncy scale, 10dB/vertical/div., 0dB = 1ns.) RMS jitter (400Hz-22kHz) = 145ps. [upload=jpg]Upload/2005391432873543.jpg[/upload]
Fig.3 PS Audio Reference Link, DAC word-clock jitter spectrum, DC-20kHz, when driven by Panasonic SV-3700 DAT recorder (linear freqeuncy scale, 10dB/vertical/div., 0dB = 1ns.) RMS jitter (400Hz-22kHz) = 561ps.

Incidentally, the digital processor used in this experiment was the PS Audio Reference Link, which uses the Crystal CS8412 input receiver in perhaps the best possible implementation. The difference would have been even more dramatic if I'd chosen a processor with the Yamaha YM3623 chip, or one that had a lower-quality implementation of the CS8412. Note that these measurements don't reflect poorly on the Reference Link: any processor with the Yamaha or Crystal input receiver (ie, just about all of them) will pass these differences in transport jitter to the recovered DAC word clock (footnote 3).

Because a processor's clock jitter changes so dramatically when driven by different digital sources, some important questions are raised about how jitter is specified in digital processors. First, when digital-processor manufacturers quote jitter numbers in their literature, what transport or digital source do they use? With what test signals? And over what measurement bandwidth? Finally, what test instrument do they use to measure jitter? It's too easy for manufacturers to offhandedly claim a low jitter number without even knowing what the jitter levels or characteristics really are. Be suspicious of any jitter claims made in manufacturer's specification sheets and promotional literature.

As clearly demonstrated in the experiment described above, transport and interface jitter end up at the DAC's word clock—the point where jitter affects sound quality.

Test methodology
Using the UltraAnalog jitter analyzer is simple. A CD transport or other digital source is connected to the analyzer's input, and the analyzer's output is fed to an Audio Precision System One. The System One is configured to perform a 1/3-octave spectral analysis of the transport's jitter. This technique plots the jitter's energy as a function of frequency in 1/3-octave bands from 20Hz to 50kHz. This is exactly the same technique we use to look at a processor's output when decoding a 1kHz, -90dB dithered sinewave in all our CD-player and digital-processor reviews.

After the spectral analysis is performed, the overall RMS jitter amplitude is measured using the System One. The System One's bandpass filters are invoked to band-limit the measurement to 10Hz-30kHz. The measured RMS voltage of the jitter energy in the 10Hz-30kHz band indicates the transport's RMS jitter level, expressed as a single number in picoseconds (the analyzer is calibrated at 100mV per nanosecond). We thus end up both with a spectrum of the jitter and a number that reveals the "area under the curve" of that spectrum.

Because the jitter analyzer doesn't have optical inputs, the measurements were restricted to coaxial and AES/EBU outputs. The transports' AES/EBU outputs generally had lower jitter than their S/PDIF outputs. In some cases there was little or no difference; in other products—the Meridian CDR and SV-3700, for example—the jitter was significantly lower from the AES/EBU jacks (footnote 4). Except for the SV-3700, which was chosen to illustrate a poorly implemented S/PDIF interface, I chose the lower-jitter AES/EBU measurements for presentation.

The test terrain
Selecting appropriate test signals is difficult for any measurement, never mind one as new as transport jitter. We settled on showing two graphs for each transport: one made using three test signals, and one made with two musical selections. The test signals were all taken from the CBS Test CD: digital silence (all data words are zero); a 1kHz sinewave at full-scale; and a 1kHz sinewave at -90dBFS. I tried other test signals—intermodulation twin tones and squarewaves, for example—but decided to publish only the three signal conditions (no signal, very low-level signal, and very high-level signal) to keep the graphs from getting too cluttered.

The musical selections were chosen for their very different signals and levels. Music #1 (the solid trace in all the music plots) was the first minute and a half of Stravinsky's The Firebird Suite on Sheffield Lab CD-24. The levels are extremely low (about -35dB) at the beginning of this disc: the signal is mostly ambience and noise, with instruments playing very softly. The second musical selection (the dotted trace in the graphs) was chosen for its opposite signal characteristics. "Cut to the Chase," from Steve Morse's Southern Steel (MCA MCAD-1-112), begins with the levels at almost full-scale and stays there for the whole piece. Most music falls somewhere between these extremes.

Jitter plots
A typical jitter-measurement plot is shown in fig.4, measured on the highest-jitter source I found: the Panasonic SV-3700 professional DAT machine (a Stereophile-owned sample). The horizontal scale is frequency in Hz, the vertical scale is amplitude in volts. The vertical scale is calibrated so that the "100u" at the bottom is equivalent to 1ps, the "1m" a third of the way up is 10ps, the "10m" point is 100ps, and "0.1" is 1000ps, or one nanosecond (1ns). Note that the vertical amplitude (jitter) scale is logarithmic. Below the "1m" division, each horizontal division equals 1ps. Between "1m" and "10m," each horizontal division is 10ps. Between "10m" and "0.1," each horizontal division is 100ps. The topmost division—0.1 to 0.2—equals 1ns (1000ps). Differences in the trace levels toward the graph top represent much more of a difference in jitter levels than those at the graph bottom. Incidentally, you can't infer the overall RMS jitter level from looking at where the trace lies; it takes some complicated math to make that conversion. Instead, I've presented the RMS jitter levels for each product with each test signal in Table 1. [upload=jpg]Upload/20053914372976263.jpg[/upload]Fig.4 Panasonic SV-3700 DAT recorder, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting digital silence (bottom solid trace), a 1kHz sinewave at -90dB (middle, dashed trace), and a 1kHz sinewave at 0dBFS (top, light dotted trace) (vertical scale, 1ps-2ns, 100µV = 1ps).

The solid trace in fig.4 is the SV-3700's jitter when transmitting digital silence (track 4 on the CBS Test CD), the heavy dotted trace is the jitter spectrum when the transport is transmitting a -90dB, 1kHz sinewave, and the lightest (top) trace is made with a 1kHz, 0dB sinewave. Note that the jitter isn't randomly distributed with frequency: the spikes in the trace at 100Hz and multiples of 100Hz indicate that there are jitter components with energy at those frequencies. Moreover, we see a huge change in the jitter level and spectrum with different test signals. The jitter's "signature" is quite different with low- and high-level signals.

Fig.5 is the SV-3700's jitter spectrum when playing the two musical selections. The RMS levels are: 4250ps (0dB, 1kHz signal), 1110ps (-90dB, 1kHz signal), and 180ps (digital silence). The musical selections have an RMS jitter value of 3830ps for music #1 (Firebird) and 3800ps for music #2 (Steve Morse) (footnote 5). [upload=jpg]Upload/20053914392811808.jpg[/upload]

Fig.5 Panasonic SV-3700 DAT recorder, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting music #1 (Firebird, solid) and music #2 (Steve Morse, dashed) (vertical scale, 1ps-2ns, 100µV = 1ps).

For contrast, Fig.6 is the PS Audio Lambda's jitter spectrum with the same test signals. The RMS jitter level was 51ps (worst case) and 29ps (best case). The musical signals produced the jitter plots in fig.7, which had RMS jitter values of 66ps (music #1) and 37ps (music #2). Note how the low-level musical source produced more jitter than the high-level one. The Lambda and SV-3700 are representative of very good and very poor S/PDIF jitter performance (footnote 6).

[upload=jpg]Upload/2005391439190355.jpg[/upload]
Fig.6 PS Audio Lambda, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting digital silence (bottom solid trace), a 1kHz sinewave at -90dB (middle, dashed trace), and a 1kHz sinewave at 0dBFS (top, light dotted trace) (vertical scale, 1ps-2ns, 100µV = 1ps). [upload=jpg]Upload/20053914401595441.jpg[/upload]
Fig.7 PS Audio Lambda, jitter in S/PDIF data signal, 20Hz-50kHz, when transmitting music #1 (Firebird, solid) and music #2 (Steve Morse, dashed) (vertical scale, 1ps-2ns, 100µV = 1ps).

Jitter Bugs
As described by Dr. Fourré in his article last month, the jitter from a transport and interface is highly correlated with the encoded audio signal. If the transport is putting out the digital code representing a 1kHz sinewave, we see additional jitter energy at 1kHz. The large peak seen in some plots between 7kHz and 8kHz is the subcode carried in the S/PDIF data stream. Subcode is non-audio data such as track time, track number, and whether the data have been pre-emphasized. The data rate for each subcode channel is 7.35kHz, producing jitter at 7.35kHz.

The signal-correlated jitter is greatest when the test signal is lowest in amplitude. Low-level signals produce a greater number of bit transitions than high-level signals, which induce more jitter in the interface. (Dr. Fourré explained the mechanism behind this last month.)

Before proceeding to the test results, I must caution readers not to jump to conclusions about a transport's sound quality from these measurements. Although the UltraAnalog jitter analyzer is a very sensitive and accurate instrument, there are several factors beyond the transport's intrinsic jitter than can affect a digital front-end's sonic characteristics. First, a particular digital processor may present a different impedance from the jitter analyzer's tightly specified and correct impedance. Second, the clock-recovery performance of different digital processors varies greatly, affecting the jitter spectrum and level in the recovered clock. Another variable is the different comparators used in the input circuits of different products. All these factors may affect the jitter level and spectrum passed to the recovered clock in an unknown way.

Finally, different DAC architectures (1-bit and multi-bit) respond differently to different jitter levels and the spectral distribution of that jitter. The identical word-clock jitter could produce different sonic effects, depending on the DAC and the manner in which its word clock has been recovered. Consequently, the measurements presented here should be viewed on a comparative basis only, not as an absolute quantification of a transport's intrinsic sound quality. Further, these measurements are so new that we don't fully understand how differences in measured performance affect musical perception.

With that caveat, here is how some popular CD transports performed on this new test.

Measurement surprises
I had planned to try measuring jitter differences in digital interconnects only after I'd finished measuring transports. If there were measurable differences in cables, I thought they would be revealed only by averaging many measurements with each cable (to reduce the influence of random noise), and then processing the data to uncover the tiniest of differences. The System One has a "Compute Delta" function that extracts only the difference between two measurements. My preconception was that any measurable differences between different coaxial digital interconnects would be marginal at best.
After measuring the first two products (the PS Lambda and the Panasonic SV-3700), I went back and repeated my measurements to make sure the analyzer was giving consistent results, and that my test setup was correct. When I remeasured the SV-3700, I got about half the jitter than when I first measured it!
What caused this reduction in measured jitter?

Changing the direction of the digital interconnect between the transport and the jitter analyzer.

This phenomenon was easily repeatable: put the cable in one direction and read the RMS jitter voltage, then reverse the cable direction and watch the RMS jitter voltage drop. Although I'd heard differences in digital-cable directionality, I was surprised the difference in jitter was so easily measurable—and that the jitter difference was nearly double.

To confirm this phenomenon, I repeated the test five times each on three different digital interconnects. One was a generic audio cable, the other two were Mod Squad Wonder Link and Aural Symphonics Digital Standard, both highly regarded cables specifically designed for digital transmission. The generic cable wasn't directional: it produced the same high jitter in either direction. But both the Wonder Link and the Aural Symphonics had lower jitter levels overall, but different jitter levels depending on their direction. Moreover, the generic cable had higher jitter than either of the two premium cables—even in the latters' "high-jitter" direction.

Fig.8 shows the jitter difference between cable direction in Wonder Link using the Panasonic '3700 as the source (the difference was about the same in the Aural Symphonics). Note that, at these high levels, small differences in the trace are significant. Between "10m" and "0.1" on the vertical scale, each horizontal division is 100ps. The overall RMS jitter was 4050ps with the Wonder Link connected in one direction, and 2700ps with the cable reversed.

[upload=jpg]Upload/20053914464535490.jpg[/upload]
Fig.8 Panasonic SV-3700 DAT recorder, jitter in S/PDIF data signal, 20Hz-50kHz, with Mod Squad data cable one way around (solid) vs the other (dotted). RMS jitter measured 4050ps vs 2700ps (vertical scale, 1ps-2ns, 100µV = 1ps).

I also plotted the SV-3700's jitter spectrum through the generic audio cable, Wonder Link, and Aural Symphonics Digital Standard (fig.9). The Wonder Link and Aural Symphonics were both in their "low-jitter" directions for this plot. The top trace (highest jitter) is the generic cable, the next-lower trace is Wonder Link, and the lowest is Digital Standard. You can easily see that the premium digital interconnects had significantly lower jitter than the generic cable. As we saw earlier, in figs.2 and 3, this jitter in the S/PDIF signal directly affects a digital processor's word-clock jitter, which in turn degrades sound quality.
[upload=jpg]Upload/2005391447394163.jpg[/upload]
Fig.9 Panasonic SV-3700 DAT recorder, jitter in S/PDIF data signal, 20Hz-50kHz, using a generic audio cable (solid), Wonder Link (dashed), and Aural Symphonics Digital Standard (dotted) (vertical scale, 1ps-2ns, 100µV = 1ps).

These measurements confirm the reports of critical listeners—see elsewhere in this issue—that digital interconnects sound different when connected in different directions.

I performed the same tests using the low-jitter PS Audio Lambda transport as source. The results were very different. With a good source, cable direction didn't make a difference in the measurable jitter (fig.10). This suggests that the SV-3700—or any poor-quality transmitter—reacts with the cable's impedance to create jitter-inducing reflections in the interface. The directionality was probably caused by differences in the way the two RCA plugs were soldered to the cable; any bumps or discontinuities in the solder or RCA plug will cause a change in the characteristic impedance, which will cause higher-amplitude reflections in one direction than in the other. These reflections set up dynamically changing standing waves in the interface, introducing jitter in the embedded clock. These problems were reduced by the Lambda's higher-quality output circuit.

[upload=jpg]Upload/20053914475874378.jpg[/upload]
Fig.10 Panasonic SV-3700 DAT recorder, jitter in S/PDIF data signal, 20Hz-50kHz, with Mod Squad data cable one way around (solid) vs the other (dotted) (vertical scale, 1ps-2ns, 100µV = 1ps).

In short, the worse the transport, the more cables—and their direction—affect sound quality. Incidentally, a $2.99 Radio Shack 75 ohm coaxial video cable had lower jitter than the generic audio cable, but higher jitter than either the Wonder Link or the Aural Symphonics (footnote 7).

While we're on the subject of the digital interface, I should point out that the engineering for transmitting wide-bandwidth signals was worked out nearly 50 years ago in the video world. In video transmission, the source has a carefully controlled output impedance, the cable and connectors have a precisely specified characteristic impedance and are well-shielded, and the load impedance is specified within narrow tolerances. If these practices aren't followed, reflections are created in the transmission line that play havoc with video signals. This issue is so crucial that a whole field called Time Delay Reflectometry (TDR) exists to analyze reflections in transmission lines.

The audio community should adopt the standard engineering practices of video engineering for digital interfaces. This means designing transports with a carefully controlled 75 ohm output impedance, precisely specified characteristic impedance of the cable (75 ohms with a narrow tolerance), and junking RCA connectors in favor of true 75 ohm BNC connectors. By applying standard video engineering techniques—in use for decades—the high-end product designer can greatly improve the performance of the transport/processor interface. We've seen what happens with a poorly implemented interface with the SV-3700 and different cables: higher jitter in the recovered clock and degraded sound quality. The engineering needed to optimize the digital interface is readily available. Let's use it.

Audio Alchemy DTI
The next job was to measure the effect of the $349 Audio Alchemy DTI "jitter-reduction" box on the measured jitter—using the same digital interconnects in the same direction for each test. In my review of the DTI last May, I concluded that it improved some transport/processor combinations and made others sound worse. I wrote: "The Audio Alchemy DTI's ability to improve the sound of a digital front end varied greatly with the transport and digital processor used....Although the DTI can improve the sounds of some digital systems, it is no substitute for a topnotch transport....Careful evaluation in one's own system is mandatory before purchasing the DTI." (emphasis in original). JA's independent auditioning also revealed that the DTI degraded the sound of some transport/processor combinations.

天啊，这个文章就没完了吗？？？？？

略过图片
Incidentally, the jitter analyzer wouldn't lock to the Runco/MSB laserdisc player, the Museatex CD-Deck, or the Meitner IDAT when I attempted to use the latter's digital input/output (which has the jitter-reducing C-Lock receiver and transmitter circuits) as a jitter-reduction device. (These transports' output carrier frequencies were probably shifted from the tolerance allowed in the S/PDIF interface.)

Analysis
Although we can't quantify a CD transport's musical performance just by looking at these curves, some broad correlations with sound quality emerged. The four transports held in highest sonic regard (the No.31, C.E.C. TL 1, Lambda, and Proceed PDT 3) all had low jitter. The Meridian 200, one step lower in "Recommended Components," had higher jitter than all the Class A and B transports that I measured. The JVC XLZ-1010 (replaced in Class D by the similar XLZ-1050) had higher jitter still. Further, I thought the PS Audio Lambda was slightly better-sounding than the PDT 3—and the Lambda had slightly lower jitter. Finally, the Audio Alchemy DTI's effect on jitter—whether an increase or decrease—was clearly audible.

After examining the data and relating the measured results to my listening impressions of transports I was familiar with, I concluded that the data's graphical presentation doesn't have fine enough resolution to show small differences in jitter. Moreover, I'm led to believe by some respected engineers that sound quality is affected even by the very small differences in jitter performance revealed by these measurements. These differences aren't resolved by the graph's vertical scale, which was chosen to accommodate the very high jitter from the Panasonic SV-3700's S/PDIF output.

I've therefore examined the differences between well-known transports by using finer resolution in the graphical presentation. Fig.35 is a comparison of the $1695 Meridian 200 with the $8495 Mark Levinson No.31, shown with an expanded amplitude scale. The Meridian is the solid line, the No.31 the dotted line. The No.31 has a much smoother spectrum—the jitter is more random in nature—and is lower in level overall except in the 100Hz-600Hz band. The Meridian 200's spectrum is spikier, indicating that the jitter is more periodic than random. Remember, random jitter is much more sonically benign than periodic jitter. Random jitter raises the noise floor; periodic jitter creates discrete tones around the signal frequency. These tones, harmonically unrelated to the musical signal, add unpleasant artifacts to the music. Is the 200's spikier spectrum sonically significant? Or are these differences meaningless when we consider the variables involved in a transport driving different processors?
Although the No.31 and the $4650 C.E.C. TL 1 appeared to have similar spectra and jitter levels in the graphs presented earlier, we can see in fig.36 that the C.E.C. in fact has lower jitter in the treble, particularly between 3kHz and 40kHz. The No.31, however, has lower jitter and a smoother jitter spectrum in the bass, seen in fig.37. In my listening comparisons of the two products, I noted that the C.E.C. did indeed have a softer, more laid-back treble than the No.31, but that the No.31 had tighter and better-controlled bass. Could these differences in measured jitter performance explain the differences in their sound?
Conclusions
There is now no question that jitter in CD transports and digital interfaces affects digital audio sound quality. Not only do different transports and interfaces sound different, they produce varying amounts of jitter and have their own "jitter signatures," seen in the jitter's spectral distribution.

Moreover, we can see that transport jitter goes right through the digital processor's input receiver (even the Crystal CS8412) and affects the amount of jitter at the DAC's word clock—the point where jitter makes an audible difference. If the word-clock timing is different, the sound will be different.

The revelation that digital interconnects and their direction can introduce large differences in measured jitter was quite a shock. The differences heard between digital interconnects—and in their directionality—have now been substantiated by measurement.

Although the CD-transport measurements presented here are fascinating, it is impossible to draw conclusions about how a transport will sound solely by looking at its jitter measurements. Based on the measurements and listening impressions of the Audio Alchemy DTI, we can confidently conclude that the jitter differences the DTI imposes on both high- and low-jitter sources are easily audible, and that lower jitter always correlates to better sound. But when examining the jitter performance of other transports, a direct correlation is less clear.

As described earlier, there are many variables that influence how much jitter, and jitter of what spectral distribution, appears in the recovered clock. This could be a significant factor, suggested by the example of the PDT 1 transport. The PDT 1 had slightly lower RMS jitter and an almost identical spectrum compared with the Mark Levinson No.31, despite the No.31's vastly better sonic performance (footnote 8).

This paradox illustrates the problem of interpreting an entirely new set of test data. We don't know what's significant in the measurements and what isn't. Remember, these are the first transport-jitter measurements ever presented by any publication: It will take some time and a lot more experience to determine which jitter characteristics are of actual sonic importance. Do the slight differences between the C.E.C. TL 1's and the No.31's jitters (figs.36 and 37) make an audible difference? Are we using an appropriate amplitude scale in examining jitter differences between products? Does jitter in a certain frequency band produce a much greater audible change than a similar amount of jitter in a different frequency band? If so, what are the subjective effects?

These are all unanswered questions. With traditional measurements we have a well-established framework for drawing conclusions about the audibility of measured performance. For example, we know that a frequency-response rise of 0.1dB over an octave of bandwidth is just audible. But what is the threshold of audibility for transport jitter? Our examination of transport jitter may be analogous to looking for a 0.1dB amplitude difference using a vertical scale of 20dB per division—it's there in the data, but we can't see it. Similarly, tiny differences between traces in our graphs may produce large subjective differences. We just don't know.

The measurements presented here are far from the last word in quantifying a transport's technical or musical performance. Instead, this article should be considered a primitive first step toward understanding jitter and its effects on sound quality. We need a concerted effort by critical listeners and audio engineers to understand transport jitter—and to correlate measured data with its subjective effects on the musical presentation. Only then will jitter be eliminated as a source of variability in the quality of digitally reproduced music.



--------------------------------------------------------------------------------
Sidebar: High Jitter vs. Low Jitter
What is the sonic effect of jitter in the S/PDIF datastream? To answer that, I listened to the same music through the same system, but switched between the PS Audio Lambda and the SV-3700 as the S/PDIF data source.

The difference was not subtle. On the Harmonia Mundi recording of Handel's Water Music (HMU 907010), the violins sounded almost unrecognizable through the high-jitter SV-3700. Instead, they sounded like a single synthesizer, devoid of all nuance, detail, and the sonic cues that tell you a bow is moving across strings. In place of these cues was a steely hardness and screechiness that literally made me cringe during loud passages. It sounded as though the violins were being played with hacksaw blades.

In addition, low-level detail was lost; the harpsichord seemed to disappear, and all the instruments blended into a giant roar. The presentation was drier, and reverberation decay became truncated. Overall, the music was sterile, mechanical, and uninvolving.

Moving to a very different selection, Michael Ruff's Speaking in Melodies (Sheffield CD-35), I heard a similar steeliness to the treble. Sibilance was emphasized, sounding like spikes of barbed wire sticking out on vocal "s" and "ch" sounds. The treble was edgy and aggressive. The vocal was more forward, drier, and didn't sound integrated with the band.

One change between the Lambda and the SV-3700 not heard on the Handel but clearly revealed on the Ruff disc was how jitter affects bass. The SV-3700 had more bass, but it was fatter, slower, and devoid of pitch. Through the Lambda, I could clearly hear all the notes the bass player was playing at the beginning of track 11. Through the SV-3700, the bass in this same passage was a featureless morass. Moreover, the SV-3700 destroyed this music's upbeat rhythmic intensity. The band seemed less enthusiastic through the SV-3700.

When we hear the effects of jitter, we're hearing voltage errors at the DAC's analog output caused by word-clock timing variations. This magnitude of the voltage error is a function of the signal amplitude and the jitter. Although low-level signals introduce more jitter in the interface, we're less likely to hear jitter on low-level signals. The voltage error may be on the order of a few microvolts with a low-level signal, but several millivolts with a high-level signal. Consequently, the audibility of jitter will vary greatly with the music's spectral content, level, and dynamics. According to HDCD® developer Pacific Microsonics—who has done extensive research into jitter audibility—the jitter's frequency and spectral distribution largely determine a transport's sonic signature. Jitter at one frequency may produce a certain sonic characteristic, while jitter at another frequency will create a very different subjective effect.

Clearly, jitter in the data stream driving a digital processor is audible—and a significant contributor to "digital sound." Because jitter adds artifacts we associate with digital audio in general, reducing jitter is an important step toward truly musical digital reproduction.—Robert Harley

从图片上看，我的T-DAC背面比这台机器新，特别是镀金部分，上面这张图片的接口镀金部分磨损比较大。

SMERT 在 2005-3-9 15:18:38 发表的内容 别贴了，大部分人看起来太累

就想你来帮忙总结下啊，专家：）

Raxel 在 2005-3-9 7:41:33 发表的内容 音源当然是最重要的,garbage in, garbage out. 我DAC已经买了3个了.MF DAC 21留在上海用,Benchmark DAC1接笔记本听mp3.Benchmark号称1000美元价位首选. 第3台是刚收的,接我的XA9000ES放CD. [upload=jpg]Upload/2005397403763890.jpg[/upload]

Raxel兄，Benchmark好象吹得也很厉害，可说说表现吗？跟MBL比如何？

别贴了，大部分人看起来太累