Someone interested in buying a digital/analog converter today must make tough choices. Not only are there several competing technologies to choose from—multi-bit, 1-bit, hybrid—but every converter also has its own musical signature. When someone buys a converter, they're locked in to both the technology and the sound.
Counterpoint Electronics has addressed this dilemma with their DA-10 digital processor. The DA-10 is actually a "mainframe" into which different DAC boards can be installed. If a new, better DAC comes along, you can pop out the old one and put in the new. Or if you significantly change your playback system, you can choose a different DAC for the DA-10 that better complements the new system's musical characteristics.
But is this a real advantage? Is a digital processor designed around just one DAC inherently better than one that accepts many DACs? And how much of a digital processor's sound is determined by the DAC chip?
Description
The DA-10 is housed in a sturdy, slim-line chassis with a very attractive front panel. At first glance, the DA-10 looks more like a preamplifier than a digital processor; the unit offers more features and controls than most digital converters. Four of these controls are rotary knobs, furthering the impression of a preamplifier: One knob selects between one of four digital inputs, another provides full digital-tape dubbing (1-2 and 2-1), the third is a digital-tape monitor switch, and the fourth is an invert/operate/mute control. LEDs indicate which input is selected and the position of the invert/operate/mute switch. An additional LED illuminates if the data carries the SCMS flag, which will prevent digital recording of that signal. The front panel also has a feature that I believe is unique to the DA-10: a user-adjustable Most Significant Bit (MSB) trimmer. More on this later. The rear panel has three electrical inputs on gold-plated BNC jacks. A fourth input is an optional ($200) AT&T ST-type glass fiber jack. One of the digital tape-monitor inputs and outputs is on BNC jacks; the second digital tape loop has TosLink input and output. Analog output is via a pair of single-ended RCA jacks. All the rear-panel hardware appeared to be of high quality. Moreover, the DA-10's features, look, and "feel" were all more than I would expect for $1695. Inside, the power supply features separate transformers for the analog and digital circuits and 17 regulation stages. Seven of the latter are three-pin types, which are then followed by ten discrete stages. All the supplies to the analog output section and DAC are doubly regulated, first with three-pin regulators, then with discrete regulation stages. Power-supply filtering is unusual in that it's provided by 44 470µF electrolytic caps instead of a few larger caps. According to designer Michael Elliott, many small capacitors sound better than a few big caps. Overall, this is an impressive supply for a product of this price.
The S/PDIF input receiver is the now-ubiquitous Crystal CS8412 ("C" version). This feeds a Burr-Brown DF1700 8x-oversampling digital filter, which is identical to the popular NPC SM5813 chip.
The DA-10's unique feature is its switchable DAC cards, which allow the user to select from among many DACs simply by plugging in different DAC cards made by Counterpoint. The currently available DACs are the multi-bit UltraAnalog D20400 and Analog Devices AD1862, along with the Crystal CS4328 Delta-sigma part and the Burr-Brown PCM69 hybrid DAC (a combination of multi-bit and 1-bit). Additional DAC cards will be available as new DACs are introduced. The DAC cards are about 4.5" by 2.5", except the UltraAnalog D20400, which is somewhat bigger. The UltraAnalog card sells for $995, the Analog Devices board costs $259, the CS4328 is $355, and the PCM69 costs $595 (footnote 1). I evaluated the UltraAnalog, Analog Devices, and Crystal boards.
To accommodate this wide range of DAC technology, the DAC card sends a signal back to the digital filter telling it to put out either 18- or 20-bit data, depending on the card installed. Further, because the Crystal CS4328 has its own on-chip digital filter, the 8x-oversampling DF1700 digital filter is bypassed when the Crystal board is installed.
Changing DAC cards is easy. The board unplugs from connectors and the new board pops into place. With the UltraAnalog DAC card, the user must plug in a two-conductor cable from the motherboard to the DAC card (engaging a de-glitch circuit). A pair of toggle switches on the motherboard must also be thrown when changing to certain DAC cards.
I mentioned earlier that the DA-10 has a user-adjustable front-panel MSB trimmer. What does an MSB trimmer do? In most multi-bit DACs, the transition between the digital levels 1111111111111111 and the next-higher binary number 0000000000000000 (footnote 2) occurs at the zero crossing point, also called "bipolar zero." If the value of the most significant bit (the leading "1" in our first binary number) isn't exactly the value of the 15 lower bits plus one, non-linearity will result. In practice, it's very difficult to achieve close enough accuracy between steps to avoid a discontinuity at the zero crossing transition. Consequently, DAC manufacturers provide a means of adjusting the value of the MSB so that it's exactly one quantization step greater than the value of the 15 lower bits (in a 16-bit DAC). Although digital-processor manufacturers usually set the MSB at the factory by hand, the optimum MSB adjustment changes with temperature and age. A perfectly set MSB trimmer may be misadjusted when the unit is fully warmed up, or a year later after the circuit has drifted.
Counterpoint has addressed this problem with the front-panel MSB trimmer. Although it requires some test equipment to correctly set the trimmer, Counterpoint suggests it be adjusted by ear (they liken this to setting phono-cartridge VTA). Note that, of the currently available DACs for the DA-10, only the Analog Devices AD1862 DACs need trimming. One-bit DACs work on an entirely different principle and need no adjustment. The UltraAnalog multi-bit DAC has been factory-calibrated by hand-soldering metal-film resistors into the resistor ladder network so that trimming is unnecessary. The UltraAnalog DAC will maintain its virtually perfect linearity over time (footnote 3). You can tell which DACs need trimming and which don't simply by looking at the front panel: an LED above the trimmer illuminates when the DA-10 is fitted with a DAC card that needs trimming.
Another feature unique to the DT-10 is its discrete current-to-voltage (I/V) converter. Most processors use an op-amp for the I/V stage, but designer Michael Elliott wanted to remove op-amps from the signal path as far as possible. The I/V stage also has no feedback. Only the AD1862 DAC card uses this discrete I/V converter: 1-bit DACs inherently have a voltage output and don't require I/V conversion. Further, the UltraAnalog converter has an integral op-amp I/V converter which also doesn't require the DA-10's stage. A manual switch on the pcb engages the I/V converter when the AD1862 DAC is installed.
The analog output stage is a discrete, direct-coupled unity-gain buffer that incorporates a high-speed video amplifier. The DC-servo'd circuit uses no feedback. All the transistors have small heatsinks mounted to them. De-emphasis is passive, switched in by a relay. The output filter is a passive third-order Bessel type. A muting relay prevents noise from appearing at the analog output jacks when the processor isn't locked to incoming data.
Pick a card, any card
I'll start by describing the sound of the three DAC cards relative to each other, then give an overview of those sonic qualities of the DA-10 that were common to all the DAC cards. I'll also compare the DA-10 with the Crystal CS4328 card installed to the $895 Meridian 263 processor, which also uses the CS4328 DAC. Finally, I'll put the DA-10 up against the PS Audio UltraLink, with both the AD1862 and UltraAnalog DAC cards in the DA-10.
The Crystal 1-bit DAC card held no surprises: the sound was similar to what I've heard from other processors using the chip. The CS4328 has a distinctive signature: soft bass, limited dynamics (particularly in the bass), lack of extension, but a beautiful, almost ethereal, midrange and treble.
The DA-10 with the CS4328 had some appealing qualities. The treble was nicely balanced and very clean. In fact, the DA-10 has the best-sounding treble of any converter using the CS4328 I've heard to date. The top end was open, airy, and integrated with the music. Moreover, the midrange and treble were free from grain and hash. Strings weren't overlaid with the gritty texture often heard from digital. Similarly, cymbals had a nice shimmer and detailed character rather than sounding like bursts of undifferentiated white noise.
Soundstaging was also excellent. The DA-10 with the CS4328 DAC board spread out before me an open, spacious, and very transparent soundstage—the presentation had a real sense of depth and space. Instrumental images were surrounded by an impression of air and bloom rather than sounding like cardboard cutouts. Moreover, the soundstage was transparent and crystal-clear. The ability to hear deep into the soundstage was first-rate, and the lack of opacity gave the music a sense of realism I greatly enjoyed.
The overall presentation with the CS4328 was smooth, laid-back, and musically accessible. The sound didn't offend, yet it didn't have the dynamic impact and gripping immediacy of many competing converters.
Switching to the AD1862 DAC card was like changing processors entirely. The AD1862 was incisive, tight, and highly focused compared to the CS4328. The AD1862's bass differed vastly from the Crystal DAC's. Through the AD1862, kick drum took on a depth and power only hinted at by the Crystal—the AD1862 gave some real weight and impact. I could hear the bass drum through the CS4328, but there was no power behind it. But with the AD1862, this important instrument was powerful, better-defined, and sounded vastly more dynamic.
I was also struck by how different acoustic bass sounded through the two DACs. With the AD1862, John Patitucci's bass on Kei Akagi's Playroom (Bluemoon/Moo R2 79342) had a much wider dynamic expression. The AD1862 portrayed the transient attack of fingers on strings—virtually missing from the CS4328—infusing the music with greater life and enthusiasm. The CS4328 tended to mute the attack, thus blurring the distinctions between notes. Moreover, the AD1862's "bouncier" bass quality made music much more rhythmically involving than the CS4328. Bass guitar had more weight, power, and "purr," which, when combined with the deeper extension and wider dynamics, produced a greater awareness of the music's rhythmic characteristics.
The AD1862's soundstage was also more transparent and focused than the CS4328's. Images were more tight and compact compared to the CS4328's slightly unfocused rendering. The soundstage also had tremendous clarity, with a distinct impression of there being less in the way between me and the music. Although the presentation was more forward, instrumental timbre had greater palpability, presence, and realism with the AD1862. Soundstage depth and the ability to portray spatial relationships was superb by any measure. The totally transparent rendering and excellent resolution of spatial detail combined to make the DA-10's presentation like a clear picture-window on the music. I greatly enjoyed this aspect of the DA-10 with the AD1862.
Footnote 1: The Philips SAA7323 and SAA7350 Bitstream DACs are no longer available as Counterpoint cards. I heard them at a CES and thought they sounded horrible. The popular Burr-Brown PCM63 isn't available for the DA-10; Michael Elliott thought the AD1862 was a better-sounding DAC. Footnote 2: Rather than use a simple offset binary 16-bit encoding where the lowest analog voltage level is equivalent to 0000000000000000 (ie, 0), and the highest to 1111111111111111 (ie, 65,535), the CD system uses "twos-complement" coding. This is so that the coding operates symmetrically about the midpoint of the maximum !X voltage swing. With offset binary, where the signal is described by only positive digital numbers, any kind of processing in the digital domain can easily result in an increased offset and thus a loss of accuracy. An additional advantage of twos-complement encoding is that it minimizes the circuit complexity for mathematical signal manipulation. In 16-bit "twos-complement" coding, the most negative voltage before clipping is represented by 1000000000000000, the highest positive voltage by 0111111111111111. The MSB therefore acts as a "sign bit": when set to 1, it indicates that the encoded voltage is negative.$GJA
Footnote 3: See "Industry Update" in Vol.16 No.6, p.57, for a report on how the DACs are hand-calibrated.
The DA-10 is housed in a sturdy, slim-line chassis with a very attractive front panel. At first glance, the DA-10 looks more like a preamplifier than a digital processor; the unit offers more features and controls than most digital converters. Four of these controls are rotary knobs, furthering the impression of a preamplifier: One knob selects between one of four digital inputs, another provides full digital-tape dubbing (1-2 and 2-1), the third is a digital-tape monitor switch, and the fourth is an invert/operate/mute control. LEDs indicate which input is selected and the position of the invert/operate/mute switch. An additional LED illuminates if the data carries the SCMS flag, which will prevent digital recording of that signal. The front panel also has a feature that I believe is unique to the DA-10: a user-adjustable Most Significant Bit (MSB) trimmer. More on this later. The rear panel has three electrical inputs on gold-plated BNC jacks. A fourth input is an optional ($200) AT&T ST-type glass fiber jack. One of the digital tape-monitor inputs and outputs is on BNC jacks; the second digital tape loop has TosLink input and output. Analog output is via a pair of single-ended RCA jacks. All the rear-panel hardware appeared to be of high quality. Moreover, the DA-10's features, look, and "feel" were all more than I would expect for $1695. Inside, the power supply features separate transformers for the analog and digital circuits and 17 regulation stages. Seven of the latter are three-pin types, which are then followed by ten discrete stages. All the supplies to the analog output section and DAC are doubly regulated, first with three-pin regulators, then with discrete regulation stages. Power-supply filtering is unusual in that it's provided by 44 470µF electrolytic caps instead of a few larger caps. According to designer Michael Elliott, many small capacitors sound better than a few big caps. Overall, this is an impressive supply for a product of this price.
I'll start by describing the sound of the three DAC cards relative to each other, then give an overview of those sonic qualities of the DA-10 that were common to all the DAC cards. I'll also compare the DA-10 with the Crystal CS4328 card installed to the $895 Meridian 263 processor, which also uses the CS4328 DAC. Finally, I'll put the DA-10 up against the PS Audio UltraLink, with both the AD1862 and UltraAnalog DAC cards in the DA-10.
Footnote 1: The Philips SAA7323 and SAA7350 Bitstream DACs are no longer available as Counterpoint cards. I heard them at a CES and thought they sounded horrible. The popular Burr-Brown PCM63 isn't available for the DA-10; Michael Elliott thought the AD1862 was a better-sounding DAC. Footnote 2: Rather than use a simple offset binary 16-bit encoding where the lowest analog voltage level is equivalent to 0000000000000000 (ie, 0), and the highest to 1111111111111111 (ie, 65,535), the CD system uses "twos-complement" coding. This is so that the coding operates symmetrically about the midpoint of the maximum !X voltage swing. With offset binary, where the signal is described by only positive digital numbers, any kind of processing in the digital domain can easily result in an increased offset and thus a loss of accuracy. An additional advantage of twos-complement encoding is that it minimizes the circuit complexity for mathematical signal manipulation. In 16-bit "twos-complement" coding, the most negative voltage before clipping is represented by 1000000000000000, the highest positive voltage by 0111111111111111. The MSB therefore acts as a "sign bit": when set to 1, it indicates that the encoded voltage is negative.$GJA
