Chủ Nhật, 13 tháng 2, 2011

Meridian 518 Digital Audio Processor

The High End is a tidily ordered world. There are CD players, transports, and processors used to play stereo recordings and drive stereo preamplifiers. There are stereo or mono amplifiers used to drive a pair of speakers. And then there is the British high-end company Meridian, run by one J. Robert Stuart, one of audio's deeper thinkers and a Fellow of the Audio Engineering Society. Meridian does it their way. They put their amplifiers inside their speakers. Heck, Meridian even puts their D/A processors inside their speakers when they can. And two speakers to play back stereo recordings? Meridian believes in re-creating the original soundfield no matter how many speakers and channels it takes to do it right. And they do it sufficiently successfully that their Digital Theatre system, which does all of the above, was one of Stereophile's joint Home Theater products of 1995. [See also the 2000 review of their Series 800 Digital Theatre.—Ed.]
Then, at the 1995 June CES, I saw a press release for a new Meridian product that seemed even more perverse. The $1650 518 offers digital inputs and outputs only (though it can be expanded to include analog inputs with the 562V switcher/ADC that J. Gordon Holt reviewed last June). It can digitally perform gain and source selection; it can change data with one digital word length to data with another; and it does all these things with 72-bit internal precision. How does the 518 fit within a conventional high-end audio system?
The 518 is a successor to Meridian's 618 Mastering Processor, which has gained no small reputation in professional audio circles for the way it reduces 20-bit audio data to the 16 mandated by the CD standard without sacrificing sound quality. It adds considerable functionality, however, and in its most fundamental role can be used as a digital replacement for a system's conventional preamplifier, accepting the outputs of digital sources, allowing for volume control, and feeding its output either to a D/A processor/amplifier/speaker combination, or to a system based on Meridian's DSP-series powered loudspeakers, which have digital data inputs. (Bob Stuart: "At Meridian, we say there is no preamp like no preamp!")
Or it can be used like the 618: to make 16-bit recordings with resolution and sound quality approaching those of true 18- or 20-bit recordings.
Or it can be used as a sophisticated jitter-reduction unit that also allows the user to increase resolution, change the gain of the digital signal, and even to add or remove pre-emphasis.
I immediately asked for a review sample.
Innards
The 518 is housed in a smart black enclosure identical to that of Meridian's other 500-series components. The circuitry is based on a Motorola DSP-56002 24-bit digital signal processing chip running at 40 or 60MHz (the specs. are unclear here). The selected input signal—there are five inputs and, in Meridian fashion, these have to be assigned to a display alias to become operational—is dejittered using a dual Phase-Locked Loop circuit and a crystal oscillator similar to those in the well-regarded Meridian D/A processors. The reclocked data are then fed to the DSP engine, which performs three main functions, all under the control of front-panel buttons, of other Meridian components via a Comms. socket on the rear, or of a computer via a rear RS-232 port.
First, the 518 can adjust the gain in 1dB steps up to +12dB or -99dB. This volume control operates with 72-bit internal precision, and dither is used to minimize round-off errors. Second, the input and output word lengths can be set to be different. A typical application, for example, would be when a 20-bit A/D converter was used to feed a DAT or CD-R recorder, both of which can only store 16-bit data. In such a case the 518 offers a choice of seven noise-shaping/redithering algorithms, these psychoacoustically optimized to give as audibly transparent a data reduction as possible. Finally, the 518 can add pre-emphasis to the digital data to increase resolution when the analog signal is finally reconstructed. The data output is in both AES/EBU (XLR) and S/PDIF (RCA) formats.
The well-written manual goes into much detail on how to set up and use the very flexible 518 in a number of different situations. I would suggest that its uses are limited only by the imagination of its user. Unusually, the team responsible for the 518's design are mentioned by name in the manual: credited are Phil Boddy, Richard Hollinshead, Duncan Smith, Bob Stuart, and Rhonda Wilson. Now that's a nice touch, and one that more high-end companies should echo.
Something for nothing?
I was mostly interested in the 518's resolution enhancement abilities. So, Sherlock, does the 518 do what is claimed for it? Can it give more than 20-bit precision from a datastream limited to 16-bit word lengths?

Before I discuss the results of my auditioning, I'll present the measured results. These were all taken in the digital domain using an Audio Precision System One fitted with the DSP option. The Meridian 518 was fed a data signal from the AP's AES/EBU output, set to a 20-bit word length. The 518's AES/EBU data output was looped back to the AP's digital input.
In the digital domain, the data word length is inexorably tied to the ultimate level of the noise floor. The longer the word, the lower the noise, with an approximate increase of 6dB in the system S/N ratio every time the word length is increased by one bit. Another way of looking at this is to realize that a linear-PCM system is deaf below a level related to the word length. Thus in an 8-bit system, like early computer soundcards, nothing below -48dBFS can be encoded—"FS" stands for Full Scale"; a 0dBFS signal exercises the full dynamic-range capability of a digital system—and the maximum S/N ratio is also 48dB. With the 16-bit words stored on a CD, the maximum S/N ratio is 96dB; and without the use of record dither (discussed at length in this magazine over the past 10 years), the system goes deaf below -96dBFS.
This is shown graphically in figs.1 and 2, the waveform and FFT-derived audioband spectrum of an undithered 1kHz tone recorded at a level of -90.31dBFS: the resultant waveform can be seen to toggle among just three levels, 0 ("digital black") and ±1 LSB. As a result, as well as showing a component at the fundamental level, the spectrum features harmonic distortion components at 2kHz, 3kHz, 4kHz, 5kHz, 8kHz, and 9kHz. The individual noise components in the fig.2 spectrum can each be seen to lie around the -120dBFS mark: if added in a Root-Mean-Square (RMS) manner, their sum would lie at the -96dBFS level, confirming that this is a 16-bit system.
Now look at fig.3: this is the same kind of FFT-derived spectrum, but now with 20-bit data words representing the -90.3dBFS 1kHz tone. The entire noise floor can be seen to have dropped significantly, with the individual components now lying below -140dBFS and their RMS sum at -120dBFS. The magic worked by the 518 is revealed in fig.4. This is a spectrum derived from 16-bit data, yet over almost all the audioband the noise components are as low as with true 20-bit data. Although the data word length is restricted to 16 bits, as it would be on a CD or DAT, the signal resolution is closer to 20 bits!
Are we getting something for nothing? Of course not. Look at the right-hand side of fig.4. Above 4kHz the noise rises with frequency, slowly at first but with increasing rapidity until, at the 22kHz band edge, each component is on average as high in level as the fundamental tone. What the 518 has done is "shape" the noise floor, pushing as much of it as possible to higher frequencies, where the ear is less sensitive, and dropping it in the midrange and treble regions, where the ear is most sensitive.
This can be seen in figs.5 and 6. For reference, fig.5 shows the waveform of a 1kHz tone at -90.3dBFS encoded with 20-bit precision. By contrast, fig.6, plotted to the same vertical scale, shows the waveform produced by the Meridian 518 when processing digital silence: far from there being silence, there is a considerable amount of high-frequency noise present, its peak-peak amplitude several times that of the reference sinewave in fig.5. The overall RMS level of this "blue-colored" noise is high, around -60dBFS. But as I said before, all this energy lies in a region where the human ear is very insensitive. And if you can't hear the high-frequency noise, it might as well not be there!
Note that this increase in effective resolution does not depend on the nature of the input signal. The source may have acoustic or electrical noise well above the 16-bit noise floor, but the 518 can't distinguish noise from music: everything in the source data is processed as though it were the wanted signal. On the master of Stereophile's Concert recording, for example, the analog tape hiss lay around the -60dBFS level. The difference between the use of noise-shaping (via the 518's predecessor, the Meridian 618) and straight truncation from 20-bit to 16-bit data was clearly audible, however, even in what turned out to be a double-blind test (see "As We See It," February 1995, Vol.18 No.2, p.3).
What it does depend on is the performance of the D/A converter downstream of the Meridian. But if your D/A is capable of true 20-bit resolution—and the measurements in this magazine's reviews are a good source of this information—you can perform the apparently impossible, as is shown in fig.7. To produce this graph, I fed the 518 with 20-bit data representing a 1kHz tone at -90.31dBFS, noise-shaped those data to 16-bit precision using the "B" algorithm, but then used the Meridian's digital gain-adjust function to reduce the signal level by 24dB. The result is a 16-bit datastream, nominally deaf to anything below -96dBFS, that carries a sinewave encoded with a peak amplitude of just -114dBFS. Wacky, wonderful stuff, eh?
But remember, this magic depends on the shaped noise not being audible—which depends on the playback level being correct. Slap in gobs of analog gain after your D/A and you'll start to hear the shaped noise—and it's not pretty. But as long as the recording engineer has chosen the optimum noise-shaping algorithm and you're playing back the 518-processed digital recording at a playback level around what was intended, you're home free.

The Meridian offers seven different noise-shaping algorithms, shown in Table 1. The spectral effects of each on the noise floor, hence the ultimate resolution, are shown in figs.8 through 14. (All of these were made using our old friend, the 1kHz tone at -90.31dB.) Which will be optimum depends on the nature of the original signal, the playback situation, and individual taste. For example, I ended up using Shape A on the Concert CD, which was originally recorded on analog tape. For Festival, for which I used a Nagra D 20-bit digital recorder, I used Shape C. And the late recording engineer Gabe Wiener, of Quintessential Sound, used to use curve D for all his audiophile projects.
Table 1 Meridian 518, Noise-Shaping Algorithms
Noise Advantage (44.1kHz)
ShaperNormalPre-emphasis
Flat0.0dB3.4dB
(White-spectrum TPDF dither)
High-pass3.5dB7.0dB
(Additive, high-pass TPDF dither)
Shape A6.4dB12.2dB
(Flat dither, 2nd-order MAF*)
Shape B19.0dB22.6dB
(Flat dither, 9th-order MAPÝ)
Shape C15.1dB17.9dB
(Flat dither, 9th-order MAF)
Shape D15.3dB21.5dB
(High-pass dither, 9th-order MAF)
Shape ENot specified
(High-pass dither, 2nd-order MAF)
* MAF = Minimum Audible Field, the threshold curve for human listeners with stereo loudspeakers.
Ý MAP = Minimum Audible Pressure, the threshold curve for human listeners using stereo headphones.
Sound
All my auditioning of the 518 was done in my usual reference system: Mark Levinson No.31 CD transport feeding a No.30.5 D/A processor via a 1.5m length of Illuminati's excellent new Orchid AES/EBU datalink ($750 [choke, cough, splutter!]). With the Meridian in the chain, a 1.5m length of Madrigal AES/EBU cable took the output of the transport to the 518. The D/A processor was connected to a Levinson No.38S preamplifier via 0.5m lengths of AudioTruth Diamond x2, with then 5m lengths of AudioTruth Lapis or XLO Signature 3.1 feeding a Levinson No.333 power amplifier (review underway) or a Cary CAD-300SEI single-ended amplifier. Loudspeakers were B&W Silver Signatures (mainly), the four pairs of speakers I am currently reviewing—JMlab Micron Carats, Totem MANI-2s, Joseph RM7sis, and PSB Stratus Mini Mk.IIs—or the Epos ES25s that JE reviews in the next issue. The B&Ws were used with their dedicated silver speaker cables, the other speakers with an 8' bi-wired set of Cardas Cross.
Masterful: To investigate the subjective effect of the 518's resolution enhancement while reducing data word length, I used the 20-bit master data for both Stereophile's new Festivalrecording and for the forthcoming Robert Silverman Liszt Piano Sonata recording. These were played back from a Sonic Solutions hard-disk editing system mounted in a Macintosh Quadra 650, with 6 Gigabytes of hard-drive space available.
I have no doubt that, along with other devices/algorithms to achieve the same end—such as Apogee's UV22 processor and Sony's Super Bit-Mapping process, neither of which I have yet gotten any hands-on experience with—Meridian's 518 (and its predecessor, the 618, which I used for our Concert project) achieves a revolution in CD sound quality.
After spending many hours listening to the 20-bit Sonic Solutions data fed straight to the '30.5 via a Sonic Frontiers UltraJitterbug, simply truncating the word length by setting the Sonic Solutions' output to 16-bit and switching off its dither gave a horrendously audible degradation. Midrange textures sounded clangorous rather than sweet, treble acquired an extra helping of fine glassy grain, and the recorded soundstage shrank. And most tellingly, applause no longer sounded like the sound of many hands clapping, but was reduced to a bland impulse-noise-like, er, noise.
Enter the 518 with its input word length set to "20" and its output length to "16." With the dither shape set to "Flat," some of the sound's dimensionality returned and digital applause started to sound more like applause. "High-Pass" gave more improvement, as did "Shape A." And with Shapes B through E, I could not hear much of a difference between the original and the processed data, nor could I hear any consistent differences between the different dither algorithms.
Mission accomplished; the magic became real. I therefore used the Meridian to convert the edited 20-bit data for Festival to 16-bit.
Meridian recommends using the 518's adjustable digital gain to ensure that the signal occupies the full dynamic range of the output medium. As the Festival data already peaked at 1 LSB below 0dBFS, I didn't add gain. But as I had recorded Aaron Copland's Appalachian Spring6dB hotter than the percussion-heavy Milhaud and Kohjiba works in the same program, I used the 518 to reduce its level by 6dB when I transferred the 20-bit data to 16-bit CD-R. In this way, not only would I achieve the correct level match among the three works, but I would do so with the minimum resolution penalty. As the 518 indicates when the signal reaches 0dBFS, you don't have to worry about clipping.
Meridian also recommends using the 518 to add pre-emphasis, Bob Stuart correctly pointing out that wrapping the digital data in a pre-emphasis/de-emphasis loop will result in a 1-bit (6dB) improvement in S/N, a 2-bit improvement in HF resolution, and a more than 10-fold reduction in D/A converter noise. I therefore tried mastering my data with pre-emphasis (adjusting the overall gain to avoid clipping). To my surprise, not only could I not hear any difference between emphasis and no emphasis played back through the Levinson, occasionally I thought it sounded worse. There did seem to be an improvement with my budget reference, the Assemblage DAC-1, however.
I suspect that the reason for this disparity between theory and practice is due to how the D/A processor implements the de-emphasis. The whole idea is to place the de-emphasis as far back in the chain as possible. In this way, all the high-frequency errors and artifacts produced by a processor or CD player's digital filter and DAC will be rolled off, reducing their audibility. The logically optimal de-emphasis implementation will therefore be in the analog domain, right before the processor's output amplifier. Indeed, the Assemblage's de-emphasis is achieved with a FET-switched network in the feedback loop around the I/V converter op-amp. But the Levinson's de-emphasis, as in almost all processors, is achieved in the digital domain, within the digital filter chip. While this is not subject to the vagaries of analog component tolerance, giving a perfect de-emphasis curve, it sort of throws away the reason for using it in the first place (as Bob Stuart points out in the 518's manual).
Glitches: Any recording engineer using A/D converters with better than 16-bit resolution but storage media restricted to 16 bits—CD, CD-R, DAT, etc.—owes it to their sound-quality-conscious customers to use a Meridian 518. However, I did have one operational problem using the 518 in that I could not get my Panasonic 3700 professional DAT recorder to lock to its output, even with the channel bit set to "P" for "Professional." For the Festival project, I used a Meridian CDR as an interim storage medium (footnote 1), as this would lock onto the 518's data output with the channel bit set to "C" for "Consumer." When I had the same problem with the earlier Meridian 618, this was solved by an EPROMectomy. I assume, therefore, that the next 518 software release will fix the bug; I eagerly await the arrival of a new chip before my next recording projects gets underway.
Although the 518 has a female 9-pin RS-232 socket on its rear and the manual tells you how to control the unit with a computer, other than once getting it to completely freeze up, I couldn't get the 518 to respond. This was despite all possible combinations of comms programs, RS-232 cables, and gender changers I tried. Don't blame the 518 in this regard, however, as I am singularly unlucky in matters RS-232.
Magic: During the editing and mastering for Stereophile's Festival CD project, I grew very familiar with the Meridian 518's positive effects when used as a resolution enhancer to reduce digital word lengths. The limited time I had available during the review period meant that I wasn't able to try it as a full-blown digital control center. But one thing I did try at the end of the review period was the inverse: feeding it 16-bit data and instructing it to output 20-bit data to my Levinson No.30.5 (which does have 20-bit resolution) with noise-shaping algorithms "A," "D," or "E" (which one worked best seemed dependent on the source material).
I was expecting nothing to change. After all, both the S/PDIF and AES/EBU datastreams carry 24-bit words at all times. The nominal word length only means that the specified number of MSBs are significant, the rest being set to zero (ie, the digital representation of "digital zero" encoded with 16-bit precision in the datastream is 111111111111111100000000; with 20-bit precision, 111111111111111111110000). In both cases, the actual word length in the signal is 24 bits wide/long. And remember that the Levinson '30.5 uses a FIFO data buffer with a length of 1/8 second to minimize input-data word-clock jitter (footnote 2).
But strap me if there wasn't something quite magical happening. First, the bass became more solid, more deep. The rather lightweight bass synth lines and kickdrums on Steve Winwood's "Valerie" (from The Finer Things boxed set, Island Chronicles 314 516 860-2), for example, became more physically moving, more dancelicious. Orchestral bass drum and organ bass pedals purred more palpably, and the low orchestral strings acquired more bloom. Second, little details in the soundstage, such as the echoes from the walls of the hall, became both more noticeable and paradoxically better integrated with the sound sources. Image depth deepened. But most important, the flow of the music just became more enveloping. It was goosebump city, even with recordings that usually strike me as just being notes by numbers.
I switched to Bypass mode: the sound reverted to being merely very good. The magic had diminished. I switched back to the 16-20 dithered word-length increase. Back came the magic.
I swear I am not making this up! Because I didn't expect anything to happen, I didn't even tryit during most of the time I had the 518! So what the heck was going on? Yes, the 518's dual-PLL receiver design means that it acts as a conventional jitter-reduction unit, with the added benefit of stripping out the subcode, including the SCMS bit. (You can see in any ofStereophile's digital processor reviews that the subcode data in the S/PDIF signal produces a ubiquitous 7.35kHz jitter content.) But surely that would also happen with the unit in its bypass mode. I reread the 518's manual. There on p.8 was a warning: "Warning! Systems like this can sound incredibly good!" But this appeared to refer to the 518 used as a digital preamp, where its reduction in jitter is coupled with the complete elimination of the analog preamplifier. I can only assume that using the 518 to increase the word length helps the digital filter in the following processor to do its math more optimally, with reduced round-off error.
Or something.
Was there a downside? But of course. HDCD-encoded recordings would not pass intact through the 518 in its resolution-enhancement mode. But that was trivial given the presence of its Bypass button, which disables any DSP guarantees that all 24 bits presented at the input are passed through to the output.
I generally didn't use the 518's gain-adjust function during CD playback; I routinely monitor recorded level using a Dorrough AES/EBU-input meter, and it's rare that a modern CD doesn't hit -1dBFS or higher some of the time. But there are CDs which are way under-recorded—A Meeting by the River, Kavi Alexander's recording of Ry Cooder and V.M. Bhatt (Water Lily Acoustics WLA-CS-29-CD), for example—and which did benefit from a little noise-shaped volume increase.
Conclusion
When I heard about the Meridian 518, it struck me that it was in some ways an orphan product. While it is an essential tool for quality-oriented engineers, I couldn't see what kind of role it would have in the context of an ordinary high-end audio system.
Now that I've spent time with the 518, I'm going to have to get out my checkbook. Meridian isnot getting this review sample back! Yes, it is invaluable used for mastering 16-bit CDs from 20-bit originals. But its effect on CD playback is almost as staggering. And at $1650, it's even a bargain. The Meridian 518 is the Swiss Army Knife of digital audio.


Footnote 1: Putting Pioneer's $25/disc price for the blank 60-minute "consumer" CD-R media into perspective—see SS's review of the new PDR-99 elsewhere in this issue—the latest price we paid for 74-minute "professional" blanks was just $7.50! No doubt this is a consequence of RIAA pressure not just on Pioneer, but on all companies offering consumer-oriented digital recorders.
Footnote 2: It's hard to believe that any changes upstream of that mighty buffer could affect the sound of the Levinson, yet in previous experiments the differences made by either an Audio Alchemy DTI•Pro or a Sonic Frontiers UltraJitterbug were clearly audible. (Which sounded better, or none at all, was very dependent on the music.)