Digital Processor Reviews

To echo Audio Cheapskate Sam Tellig, who was in turn paraphrasing Thomas R. Marshall, what the world needs is a great $299 CD player. Certainly there's no shortage of expensive units vying for your attention—most of them consisting of separate transports and D/A converters.

Sam, of course, has often advised against spending big for CD playback. That's an easy recommendation to make if you have a bazillion LPs and consider CD to be a secondary (and inferior) program source. There was a time when I would have agreed with Sam's low opinion of digital, but no longer. Neither does Madrigal Audio Laboratories, if we can judge by the fact that they make some of the best—and most expensive—CD playback devices around.

When Madrigal introduced their Mark Levinson No.30—now the No.30.5—its price put it out of reach of any but the most affluent or debt-tolerant audiophiles. Add their No.31 transport, and you have a terrific combination which, unfortunately, turns up mostly in homes with four-car garages and security gates. And when the less-expensive No.35 was introduced, we were still talking champagne and Brie.

Now, however, with the introduction of the No.36 (and the companion No.37 transport), Madrigal has a CD playback combination available to the merely terminally obsessed among us. You can, of course, reduce your outlay more—a lot more—and still get quality digital replay. But if the you want your less-pricey thoroughbreds to come from the Madrigal stable, they will bear the name "Proceed," not "Mark Levinson."

Features & design
With inputs for every popular type of digital connection, the No.36 is ready for use with any transport or other digital source. The selected input signal fades in slowly to prevent extraneous switching noise. A front-panel LED display indicates the status of the selected input. The No.36 also may be programmed to display an appropriate name for that source—CD, DAT, DCC, or what-have-you. Other front-panel functions are Phase, Teach, and Mode. Phase, as you might expect, inverts the phase (polarity) of both channels. Teach allows the No.36 to program an outboard learning remote to control its functions (no remote is provided with the No.36, but the remotes furnished with other Madrigal products are suitable). And Mode selects or turns off the digital output and serves as a control switch during processor setup (naming the inputs).

In addition to its inputs, and balanced and unbalanced outputs, a rear port labeled "Master Communications Port" allows the No.36 to be linked with other 30-series components such as the No.37 and '31 transports, No.38 and '38s preamplifiers, and No.33-series power amplifiers, further expanding functional flexibility.

One of the HDCD specifications calls for some form of level control to compensate for the 6dB difference in peak output level between conventional CDs and decoded HDCDs. The No.36 may be set up to do this automatically (Auto mode), in which case it drops the level of standard CDs by 6dB in the digital domain to match the average perceived level of HDCDs. In its Manual mode, it feeds the unaltered signal to the analog outputs, meaning that you must boost the preamplifier level by 6dB. Madrigal strongly prefers the latter, analog-domain level-change as the most sonically transparent.

If you use the No.36 with the most recent versions of the No.38 or 38S preamplifier, the communications link referred to in the preceding paragraph will automatically trigger this level change in the preamp when the No.36 senses an HDCD. In my case, since I used a preamp from a different manufacturer, I used the manual mode on the No.36 and made the level change myself at the preamp. (Incidentally, the No.36 is delivered in the Auto mode—with the level change made digitally and internally—because of HDCD requirements. This can be easily changed to Manual mode by the user.)

Madrigal makes much of the fact that the No.36 uses the new Pacific Microsonics PMD100 HDCD, 24-bit throughput, 8x-oversampling HDCD decoder/digital filter chip, which is said to be inherently superior, sonically, to previous digital filters even when used with conventional CDs. The overall resolution of the No.36 is limited to 20 bits by its D/A converters, two per channel operating in a balanced configuration. (We should never forget, with manufacturers throwing references to 20- and 24-bit resolution around with abandon, that higher-bit converters can never recover more than 16 bits from CDs—which are inherently limited to that resolution unless some kind of noise-shaping/redithering has been used in the mastering process. They simply make it more likely that we'll get the full benefit of those 16 bits).

The No.36 also makes use of something called a "smart" FIFO (First In, First Out) buffer to minimize jitter. FIFOs are basically a way of re-clocking the incoming data with high precision, minimizing jitter which may be inherent in the incoming signal, and are not unusual in high-end D/A converters. But if the FIFO is too small, the buffer can either overflow with a subsequent loss of data, or be underfilled with a resultant need for additional error correction. If the buffer is large, there may be an unacceptable time delay (okay for audio but unacceptable with the audio signals on a laserdisc). With the No.36's "smart" FIFO, the data release rate from the buffer is controlled by software that tracks the signal's long-term data rate. The amount of data in the buffer memory will therefore never approach the full or empty states. As the output data clock can be made to be of very high stability and accuracy, any time variations in the received data clock—jitter—are simply ignored. Madrigal claims less than 20 picoseconds of jitter for the No.36.

This "smart" FIFO operates at the 44.1kHz sampling rate of CDs and laserdiscs and the 48kHz sampling rate of DATs. Though the No.36 will lock onto and decode the 32kHz sampling frequency proposed for digital satellite transmission—the latter little used and of no real concern in any event for high-end reproduction—the FIFO buffer is not functional at this rate.

Elsewhere, the No.36 shunts unselected digital inputs to ground to minimize interference. The internal circuitry is also fully balanced in both analog and digital (unbalanced digital inputs are converted to balanced before any processing is performed).

Sound
When I first put the Mark Levinson No.36 in my system, the Levinson No.35 D/A converter had been a long-term resident there. My initial impressions were that the No.35 was a little more open and sparkling on top—though those who tend not to like this quality may refer to it as "crispness." But I had the same impression of the No.35 when I compared it with the original No.30 (Vol.16 No.11), so the No.36 was certainly in good company.

Most of my listening to the No.36 was done in its unbalanced mode. Direct comparisons between unbalanced and balanced modes in my system resulted in no significant differences (footnote 1). If I thought I heard anything, it was perhaps—just perhaps—a slightly more full-bodied sound in the unbalanced configuration. But it was a very elusive thing. Similarly, I used the coaxial digital input (RCA) for most of the auditioning. The differences between coaxial and ST optical were subtle and about what I've observed in the past. The coaxial—with the Kimber digital cables—was slightly more immediate and sparkling, the ST a little softer and sweeter.

In any mode, the longer I listened to the No.36, the more I liked it. When the No.35 went back to the manufacturer, I had some initial separation pangs, but they lasted about two days. The No.36 picked up the slack without missing a beat. I've used it not only in the system listed above, but also as the primary signal source in every review I've done in the past several months—both for Stereophile and for the listening-room portions of my reviews for the Stereophile Guide to Home Theater. It has not been a limiting factor in any of these reviews.

How do I describe the sound of the No.36 without describing the sounds of the components with which I used it? Only with great difficulty. The bass was deep and powerful through the Energy Veritas loudspeakers, full-bodied and extended through the KEF Reference 107/2s, and tight and punchy with the Thiel CS7s. From the solid bottom-end whack of the drums on Däfos (Reference RR-12 CD) and the bite of Dean Peer's bass on Travelogue (Fahrenheit FR2451), to the low-end extension of the organ on Jean Guillou's organ transcription of Mussorgsky's Pictures at an Exhibition (Dorian DOR-90117), I found nothing to criticize. And if the top end of the No.36 had seemed a bit softer and sweeter in my initial exposure to it, the crisper-sounding No.35 was now just a memory: The No.36 revealed itself to have all the subtlety and detail I could hope for, while never sounding aggressive or harsh (except on the worst recordings).

But as with the frequency extremes, the sound of the No.36 through the midrange—including soundstaging and depth—appeared to be pretty much that of the rest of the system. It never imposed anything on the sound that I could point to and say "Ah-ha—that No.36 sound, again." It simply did not call attention to itself.

High Definition Compatible Digital
As to HDCD, the performance of the No.36 remained top-class. Pop an HDCD into the transport, and the letters "HDCD" light up the No.36's LED display screen. But color me undecided when it comes to HDCD vs standard CD. To date, there's simply not enough material available in both formats (make that almost none) to make a really intelligent comparison. Sure, the HDCD recordings I've heard have been uniformly terrific in sound. But they've been excellent with and without decoding. Comparisons between their undecoded or decoded playback are not only invalid—because the coding results in subtle changes to the sound of an undecoded disc over that which might be expected from a conventionally mastered one—but also nearly impossible to make. You can't shut off the HDCD processing in any HDCD decoder I know of, and comparing the modes using two processors—one HDCD, the other not—requires that the processors be otherwise identical: an impossible condition to meet.

---

Footnote 1: This is consistent with my past experience. In an environment with long interconnect runs and a lot of RF (downtown New York, maybe) the advantages of balanced lines might be more evident. For most listeners and with most equipment, however, I'm not convinced that it justifies the complication and expense.