Reprinted, with permission, from Audio Electronics, (P.O. Box 576, Peterborough, NH 03458), (c) (1997), Audio Amateur Publications, Inc. All rights reserved.

(note: The following text is the article obtained directly from Mr. Carter and does not include changes made by the editors of Audio Electronics. Also the photos of Kevin's X-DAC 3.0 linked are not always the same views featured in the paper version of the article.)

Breaking The Ice

It took me over 8 years after the Compact Disk was introduced to give into its charms, which were, at the time, relatively limited in number. It became a necessity to equip myself for CD to get the benefit of new performances that I read about and longed to partake of. So I went out and bought a generic Magnavox player, swapped the op amps for Burr Brown OPA2604s, POOGEd the power supply, and replaced the output caps with polypropylene units. After modification, the player didn't sound particularly offensive any longer, but clearly left a lot to be desired in the definition area compared to my modest analog set-up.

Sometime later, I invested in one of the few new pieces of audio gear that I have ever purchased, a Monarchy 22A D/A Converter. It was a sonic revelation compared to the player, but after a few months it was clear that it, too, left something to be desired in the area of midrange and high frequency smoothness. On many CDs the presentation was harsh and bright, especially as the music became louder. Musical climaxes were often more than just annoying, they were downright unlistenable.

Analytical System

My system borders on the analytical side, although I use a tube amplifier and line stage, probably due to the Focal drivers in my Aria 5 speakers. The undamped Kevlar midbass drivers tend to have some lower level resonances in the higher end of their range that aggravate any tendency toward brightness in a component introduced into the system. So I decided that I either needed a new set of speakers or should investigate the latest technology in the D/A conversion area for a sign of improvements that could be incorporated by the DIY type.

I got interested in the Analog Devices AD1890/1891 sample rate converter (SRC) and followed a thread on the rec.audio.tech Usenet discussing its potential for the alleviation of what is felt to be one of the chief technical weaknesses of the Sony/Philips system, that of jitter imposed on the data stream in the S/PDIF digital transmission interface. (For a discussion of the jitter reducing principles embodied in the AD1890/1891, see Gary Galo's column in TAA 3/94.)

Norman Tracy of the Audio Crafters Guild (ACG) was one of the main contributors to that thread. After corresponding with and talking to Norman live I got a loan of an early prototype of his DAC design to demonstrate at a Piedmont Audio Society meeting and I was hooked. The X-DAC 3.0 looked like it had the potential to take me to the next level of music listening using the CD format.

I had already obtained some of the parts for a DAC based on the Crystal Semiconductor chips, so I didn't need a kit, but I obtained an AD1891JN and a Crystal CS8412CP input receiver from Norman, as well as his printed wiring board, digital input transformers, and a kit manual. Most of the rest of the stuff necessary I had obtained from Digi-Key previously. Norman has several levels of the kit available (with and without the AD1891 and with 1 to 4 digital inputs) that allow the builder to customize his/her project to their needs and budget constraints.

Which One?

The most basic kit without the SRC chip and with one input is $571, which includes everything except the case. A fully drilled and laser-etched lettered case is available for $230. The full-on kit with the SRC and four digital inputs sells for $762. Ultimately, the version I built corresponded to the "deluxe" kit with four digital inputs and the AD1891 SRC. This report will review the process and the results of the effort. (Photo 1)

Crystal Chips

The X-DAC is based on the Crystal Semiconductor CS8412CP input receiver and the CS4328CP D/A converter. The Crystal chips result in a simpler implementation than most other DACs because the digital filter and the anti-alias filter are built into the CS4328. So, at the simplest level, only a small number of devices are necessary to make a working DAC. The X-DAC is somewhat more complex than this for reasons that pertain to sonics.

First of all, the quality of the clock signal is a key factor in the maintenance of low jitter. Norman has placed a low jitter crystal-driven ECL clock immediately adjacent to the DAC chip. The use of this clock circuit helps keep the jitter levels attained by the SRC low.

Second, the use of regulators at all key points in the circuit reduce power supply mediated interaction between the various stages of the DAC circuitry. And, of course, the presence of the AD1891 SRC to reduce jitter adds complexity to the DAC.

The CS4328 has onboard audio output buffers, such that no further audio signal manipulation is required prior to the output jacks, however, Norman has used AD811 video buffers at the audio outputs to provide better drive capability for the succeeding component on the audio chain. These also have dedicated three terminal regulators. I hope you have gotten the picture that while this is not a device with the sophistication and complexity of a Mark Levinson No.30.5, it is far from a barebones DAC.

The entire circuit, less the raw power supply, is contained on one 6.5 inch by 10 inch double-sided printed wiring board. The board is professional quality with plated through holes and solder mask. Norman has implemented the design suggested by Crystal in their application notes that carefully maintains separate analog and digital supplies and ground planes. This is obviously critical to avoiding the contamination of the output with digital artifacts.

Take Your Pick

You can either construct a power supply on a separate board or use one of the dreaded "wall-wart" supplies that provide + and - 12 volts, as well as 5 volts (provided with the kits). There is a place on the board for a 5 pin DIN connector that is commonly used on such canned supplies. If you choose to use a "wall-wart" supply, select one with the largest current capability that you can afford. It makes a sonic difference.

There is provision for the mounting of input and output connectors directly to the board, if you use the PC mount connectors supplied with the kits. Up to four digital inputs are possible (one is configured for TOSLINK optical, if desired, while the others are BNC) and a single pair of RCA audio outputs are provided on-board. The connectors themselves are all high quality gold-plated, but the BNCs are the common 50 ohm type, rather than the more desirable 75 ohm version (more on this later).

The front panel controls consist of an input selector switch with indicator LEDs, a MUTE switch, a POLARITY switch with indicator LED, and input receiver status LEDs for ERROR and DEEMPHASIS. The LEDs and MUTE and POLARITY switches can be board mounted, if desired, also. With an enclosure of the correct size, the board mounting of components can save much of the wiring, making very little wiring required to complete the kits.

The passive components provided with the kit are all of high quality. Metal film resistors, Panasonic low Z electrolytics, and WIMA polypropylene capacitors are provided for all important audio circuitry points. Other components are consistent with high quality digital circuitry.

Assembly

Having described the kit and its components, what did I use to assemble my X-DAC? Well, I followed Norman's parts selection philosophy pretty closely. Because I chose to mount all of the panel stuff off of the board I had more flexibility in selection of hardware. I used Kimber RCA jacks for the audio outputs and left holes for an additional pair of single-ended audio outputs and balanced audio outputs. I configured my digital inputs with one AES/EBU professional balanced 110 ohm digital input (which requires a special order transformer), two 75 ohm BNC S/PDIF inputs, and one RCA S/PDIF input. This allows me maximum flexibility in interfacing with digital sources.

I have fitted my CD player/transport (Magnavox CDB-650) with a 75 ohm transformer coupled S/PDIF output and have found that an all-75 ohm digital transmission system is sonically superior to all RCA connector-based cables I have tried. Low level detail is recovered better with less fatiguing highs using the 75 ohm BNC connectors and RG-179 (or RG-59) cable. I used an IEC AC socket for power and put a high quality Amphenol connector on the back panel to use later for battery power (if I ever get around to trying it). I also put a switch on the back for the AC ground wire to allow for ground lift, resistive (10 ohm)ground, or true ground, in order to optimize the grounding status. (Photo 2)

I upgraded the resistors in the audio output stage to Caddock metal-films, otherwise I used the same components as Norman uses in his kits. I constructed a power supply on a separate 0.1" square grid printed wiring board. I used a discrete power line filter, separate 10/10 volt AC and 6/6 volt AC semi-toroidal power transformers, common mode inductors, 3A fast recovery diodes, and polypropylene-bypassed 10,000 uF filter caps. This gave me + and - 15 volts for the analog stages and two 9 volt supplies for digital stages (or, later, for tube filaments).

This board is located in the same enclosure as the DAC board. No enclosure is provided with the kit, but can be ordered separately from ACG. Cabinets like those offered by Sescom or other vendors should be suitable, as well.

I built my own enclosure from 1/16" and 1/8" sheet aluminum, 3/8" square aluminum stock, and 1/2" by 1/2" by 1/8" aluminum angle. The combination of these three materials, a table saw with a nonferrous metals blade, and a drill press has given me a system that makes high-quality, custom cabinets easy, although somewhat time-consuming, to construct. (Photo 3)

No Neophytes

The manual that is included with the kits has assembly instructions, schematics, layouts, parts lists, and test results. I found it to be complete and I believe it is adequate for the intermediate level audio amateur. I would not recommend this kit for beginners, simply because the circuitry is complex and the instructions, while good, are not at the level that was used for the old Dyna, Heath, or EICO kits. It presupposes that you know how to solder, understand component polarity, and can make elementary voltage measurements.

Three Assembly Problems

During the course of stuffing the circuit board, I found only three problems. The most significant problem relates to Norman's determination to pack all of the components that need to be in close proximity for electrical reasons as close together as possible. He states this in the instructions, so you are warned, but I bought wonderful quality low-profile Augat IC sockets at a hamfest for not-much-money. They made it very difficult to mount the 10K resistors adjacent to the AD1891, because they were a bit wider than standard low-profile IC sockets. Beware substitution of parts from a mechanical perspective; there are many places on this board where space is very tight Also be very careful mounting components, especially capacitors, to ensure that the polarity is right the first time. There are a couple of engineering change notices appended to the instructions that should be heeded. Unsoldering and removing components is difficult due to the very tight hole clearances used. I got a little careless with one cap, caught my error on double-checking, but had a devilish time removing the cap intact without frying the board.

Two Kit Problems

I only experienced two kit (not assembler-related) problems. The holes in the board for GPD/LS connector were too small for the one specified. Fortunately, this connector is not required if the AD1891 will be used and can be omitted. One of the holes for the DAC clock crystal was partially blocked by what appeared to be a plating through flaw. I simply soldered the crystal in from the top for that hole and all was well.

Modifications

I made a significant modification in the analog buffer stage power supply by using LM317/337 regulators instead of the three terminal fixed voltage units specified. I simply added the voltage setting resistors and a reference bypass cap to little pieces of perf board. Space is a little tight, but it worked. While these regulators are hardly state-of-the-art, they are significantly better than three terminal fixed voltage regulators in the areas of noise and output impedance. At a later date I plan on creating a more compact layout for the SuperRegulators designed by Walt Jung (TAA 4/94) and substitute them, but I couldn't wait for this added improvement.

I also raised the voltage at the AD811 buffers from + and - 5 volts to + and - 12 volts, increasing the sonic performance of the output buffers significantly. To be conservative I glued small heat sinks to the AD811s to compensate for the higher dissipation. Norman tells me that he is currently working on similar modifications for the kits. One could also use higher quality regulators for the analog section of the CS4328, but I had to stop somewhere and get on with the project.

I have to say, that in spite of the complexity of this project, it worked correctly from the first (well, second, if you count leaving the DAC crystal out as the first!) application of power.

Sonic Performance

How does the unit perform sonically? The promise of the prototype was met in every way by my completed DAC. It betters my Monarchy in every significant area of performance. Most striking are the improvements in soundstage depth and width and dynamic clarity. The portrayal of instruments in classical recordings is similar in scope to my analog set-up (VPI HW-19 Mk II, Fidelity Research FR-64, and Audio Technica OC-9) without the surface noise to obscure the lowest level details.

The soundstage is as deep as it is wide and instrument localization, where it is strongly present, is very, very good. No more squashed "digital" imaging. The musical dynamics are more realistic. When orchestral climaxes occur, one can get a sense of what it might have sounded like in the performance hall, whereas the Monarchy rendered louder sounds more harshly, with a collapsing soundstage.

At the low amplitude end, the X-DAC has a kind of dark quiet between notes that I have rarely experienced with audio. Overall, it was as if I had installed an expander circuit in my system, without any of the deleterious effects associated with these. If you have recently gone to a live classical performance, you were probably surprised by the dynamic range possible in real life. Most of us listen to our audio systems much more frequently than we hear live music, so I believe that unconsciously we come to expect live music to sound like our systems. Even though I attend our local symphony concerts regularly, I am always pleasantly reminded anew of my bias at every concert. The X-DAC gave me a significant piece of that experience. It also has made a large number of my CDs listenable at more realistic volume levels; previously I cringed at the thought of some of my favorites played at these levels and would flee to the next room to "filter" out the harshness.

Meeting of the Converters

The Piedmont Audio Society had a meeting at which the theme was D/A converters. Several converters were brought by members and two rounds of listening were conducted to accommodate all of the participants. Four different music selections were used (orchestral, piano, solo vocal, and R&B) and changeovers were made as rapidly as possible. The DACs were all warmed up for at least 60 minutes prior to listening and level matching was done by group acclaim.

Obviously, this is not an ideal comparative listening scenario (and certainly not one that would please the hardcore double blind advocates), but differences were apparent and reproducible between the groups. The converters included:

Adcom GDA-600

Audio Alchemy DAC-in-the-Box

Audio Crafters Guild X-DAC 3.0

Monarchy 22A

Member-built unit based on Burr Brown PCM63 DAC

Proceed PDP-2

Sony DAS 703ES

Theta DS Pre Gen III

Theta Pro Basic II with Theta DIL

The results were not expressed as a forced ranking, rather the DACs were placed in groups of like sonic value, notwithstanding their individual strengths and weaknesses. This turned out as follows:

Upper group:

Adcom GDA-600

Audio Crafters Guild X-DAC 3.0

Proceed PDP-2

Theta DS Pre Generation III

Middle group:

Member-built unit based on Burr Brown PCM63 DAC

Theta DS Pro Basic II

Lower group:

Audio Alchemy DAC-in-the-Box

Monarchy 22A

Sony DAS 703ES

Significant Differences

It's worth mentioning that these conclusions were arrived at in one system (mine) with all of the inherent weaknesses and strengths that it possesses as they interact with the DACs' characteristics. DACs in a group had significant differences in performance, especially in the lower group.

The Sony reproduced music at a more primitive level than most CD players, whereas the DAC-in-the-Box is a marked improvement and the Monarchy yet again improved the perception of the performance.

In the highest group, the Theta had the most liquid sound, while the X-DAC had the greatest clarity, with the other two somewhere in between. However, I could live with any of them in this group and enjoy the music. One of the biggest advantages of DIY is your ability to tailor the sound of a component through modification to harmonize (bad pun) with the rest of the pieces in the audio reproduction chain.

Summary

The X-DAC 3.0 has advanced my level of listening enjoyment with the digital medium a giant step. It was also my first substantive digital audio construction project; I learned a lot and enjoyed it. Assuming that you don't want to breadboard your own, ACG's kit is undoubtedly the most cost-effective way to get the sonic benefits of the Analog Devices SRC and get a great performance/cost converter in the bargain.

The AD1891 seems to be a very effective device from the standpoint of "curing" many of the ills associated with digital sound reproduction. I am going to build an outboard SRC using it to see if that will positively affect the quality of other converters, such as the Monarchy. Perhaps I will have a chance to report on the results of that at a later date.

In spite of my enthusiasm, I will not be selling my LP collection, so don't call. Analog still has a place in my system and still allows music through in a way not yet equaled by the digital medium. Maybe 96 kHz will do the trick, but we'll have to wait and see.

Manufacturer's Response

Let me begin with the thanks to a few without whom X-DAC and Audio Crafters Guild would not exist. First to thank is my wife Marian who puts up with this hobby turned all consuming passion. Next is Ed Dell the godfather of American DIY audio for the inspiration his magazines provide and designer authors like Walt Jung and Ben Duncan who share their knowledge and upon whose broad shoulders so many of us stand. Most important are the X-DAC builders like Kevin Carter, Erland Unruh, and all the others who over the last two years have built our kits.

Their combination of enthusiasm for X-DAC's sound and careful technical scrutiny as typified in Kevin's review is exactly the kind of interactive relationship I hoped for when including the word guild in our name. The feedback from guild members is a vital part of our R&D process resulting in continuous improvement of our products.

A good example of this interaction in action is the power supply and regulator options now available. Our upgrade power supply, X-Pwr, is even more radical than Kevin's with double LC filtering and we offer a shunt regulator to power the critical CS4328 DAC chip. In the 18 months since Kevin built his unit X-DAC has become my vision of a component which can be configured to the owners exact needs and desires, from elegant simplicity to extreme statement.

I am, of course, thrilled with Kevin's review of X-DAC. In many ways his description of the product is more accurate than my own brochures. The concept is simple and well known to Audio Electronics readers, great parts, great power and a fully optimized design yield great sound.

That the rest of the audio industry has overlooked the amazing power of onboard jitter reduction through asynchronous operation is their loss and our gain. The only clarification I can offer is that all four digital inputs can be setup for coax S/PDIF or balanced AES/EBU interface. Toslink ™ optical is an option for the fourth input which accepts optical or electrical input components.

I absolutely agree with Kevin that 75 ohm BNC connectors are the interface of choice so our PCB mount BNC connectors are 75 ohm characteristic impedance. For customers whose needs dictate RCA for the digital inputs these connectors fit the same PCB layout.

Of course the bottom line will always be how well we reproduce the power, majesty, subtlety, and emotion of great music. That Kevin and other members of the Piedmont Audio Society found X-DAC 3.0 superior to so many other DACs and the equal of the best tested shows we have succeeded in our goal of offering audio craftsmen a first quality base upon which to build their vision. And they have not even heard the X-DAC 3.0 Signature version yet!

Norman Tracy

Audio Crafters Guild

Owner/Chief Designer