Assemblage DAC-3.

Manufacturer: The Parts Connection, 2790 Brighton Road, Oakville, Ontario, Canada L6H 5T4; 800/769-0747; sonicfrontiers.com

Price: Kit, $1,499; assembled, $1,599; optional Signature Parts Upgrade Kit, $299 ($399 installed); optional 24/96 Daughter board, $50.

Source: Manufacturer loan

Reviewer: David Milford

Every once in a while, a piece of gear comes along that arouses your curiosity even before it's available. That's what happened when Glenn Dolick of The Parts Connection briefed me on the major design details of the DAC-3 (though some of them actually changed by the time it was in production). I wanted to hear what the massive power supplies, dual-differential operation, and elaborate ground isolation could do for the overall sound quality. Alas, it was wait, wait, and wait some more before I was able to get one to listen to. Part of the delay was necessitated by redesigning to incorporate full 24-bit, 96 kHz capability into the unit.

I don't tend to change equipment as often as some reviewers. I had been happy with the sound of my $650 DAC-2 for more than three years, though I was aware that there were sonic improvements available from some of the more expensive processors. Having been pleased by the level of quality I could afford with The Parts Connection's factory-direct pricing, the DAC-3 was intriguing. Just what kind of sonic results would come out of the differences and enhancements over the DAC-2?

When I first approached KWN about reviewing the DAC-3, he said something resembling: "most of our readers don't see the need for D/A processors, especially not a $1,500 one." At the heart of the matter is whether or not one can get noticeably better sound or not, since there is really no function for a separate D/A processor in a home audio system except to improve the sound. There is certainly more opportunity for reducing jitter when using separates and inserting a jitter-reduction box into the digital path, and this is one reason I prefer this setup to a one-box CD player. I realize that some are under the impression that there are no audible differences between different CD players, nor between different D/A processors. For those I can say try to keep an open mind and listen to what your ears are telling you. If you hear a difference when some "theory" says there should be no difference, believe your ears. If you don't hear any difference at all, lucky you! That means you can save quite a bit of money indeed.

Though some blind listening tests have indicated no discernable difference between CD players, I am not fully convinced of the accuracy and objectivity of all of these types of tests. Many sensitive details need to be properly attended to in order to recreate a good home listening experience. I personally would not be afraid to take such a test, and am certain I would perform well. But, I would need to participate in the planning and design of the test in order to create a truly objective environment in which to listen.

[A few quick points. First, the best way to avoid jitter is with a good one-box unit. Separating things into two boxes makes jitter more of a problem, necessitating a third box to reduce jitter. Simplify, simplify, simplify! Second, I suspect most folks would benefit more by spending $1500 on better speakers, a subwoofer, and/or some room treatment than by spending $1,500 on a DAC. Finally, I am more convinced of the objectivity of most ABX-style double-blind tests than I am of the objectivity of other methods. But yes, it would be nice to be involved setting up and participating in some of these tests, and it would be a wonderful outcome if we would find that we do not need to spend as much money as we thought on electronics, and then were able to free up significant chunks of our budget for better speakers and recordings. But I certainly agree with the two Davids--Milford and Rich--that there is something to be said for having the best-designed and best-performing components in the signal path, and we are all working toward that end. --KWN]

One thing that has become quite obvious over the years is that higher quality parts do usually make an improvement in sound quality. This is one of the areas where The Parts Connection has excelled, and the DAC-3 is no exception. Because they are part of Sonic Frontiers, and they also sell a good quantity of individual parts, they are able to buy and stock larger quantities of parts. These large quantities result in lower per-piece costs, which allows them to lower pricing on finished products. People have come to expect to see some of the world's finest quality parts in Assemblage and Sonic Frontiers products, at comparatively reasonable prices. The Assemblage kits are also made risk-free by the 30-day money back guarantee, assembly guarantee, toll-free technical help phone lines, and two-year warranty.

Some of the major parts found in the DAC-3 include Scientific Conversions pulse transformers and the Crystal 8414 input receiver. The digital filter is the HDCD-decoding PMD-100, which may be popped out of its socket and replaced with the optional $50 DF1704 if full 96 kHz throughput is required. Four 24-bit Burr-Brown PCM1704 DACs (two per channel), for true dual-differential signal handling throughout the analog domain. Linear Technology LT1357 opamps function as I-V converters, followed by Burr-Brown OPA134 and BUF634 high-current output buffers. The low-pass filter is passive and uses Wima polypropylene film caps. In the power supply, three large 25W transformers feed no fewer than 16 big 3300 uF Panasonic HFQ capacitors. Linear Technology regulators and ultra-fast soft recovery diodes with snubber caps round out the elaborate power supply.

Even the finest parts cannot do much unless they inhabit a good overall design. A far larger chassis than the DAC-2.6's houses four separate circuit boards, connected with ribbon-cable jumpers. One board each for digital inputs and filtering, right analog output, left analog output, and power supply. The ground reference of each board is connected via an inductor for high-frequency isolation, and the signal from the digital board is passed through special ground-isolating chips. The analog stage is fully DC-coupled, with servos to eliminate any offset. The power supplies have cascaded regulators for good isolation and low noise. These are augmented by extensive local bypassing, using ceramic caps on the digital board, and electrolytics and film caps on the analog boards.

There are six digital inputs to choose from, Toslink and ST optical, RCA and BNC S/PDIF, AES/EBU on XLR, and I2S-enhanced on 13W3. Each of these has its own LED on the front panel and can be manually toggled with a pushbutton. Or, a jumper inside the unit can be set to auto, making the unit scroll automatically to an active input, then a push of the select button jumps to the next active one. Other front panel controls are phase invert (polarity), mute (a shunt circuit), and power. Other LEDs indicate deemphasis, HDCD, and digital lock.

The unit has true balanced XLR analog outputs, as well as RCA single-ended outputs (paralleled with the XLR) . Much of the benefit of the dual-differential DACs and analog stages can be enjoyed by single-ended users (like myself), and not just those with balanced preamps and interconnects. This is thanks to the cross-coupling in the circuit just after the passive filter. The standard maximum output level is 2.0V unbalanced with or without HDCD, and output impedance is 10 Ohms. If one desires, one may enable 6 dB non-HDCD digital domain attenuation via an internal jumper, making the max output level 1V.

I installed the $299 Signature Parts Upgrade Kit (SPUK) into the review unit as I was building it. This is a comprehensive upgrade, replacing resistors, rectifiers, opamps, wire, RCA jacks, and incorporating Soundcoat chassis dampening tape. The SPUK is something of a challenge to install, mainly due to difficulty in removing solder from holes in the 4-layer audio output boards. A good solder-sucker (such as Radio Shack catalog #64-2060) is an absolute requirement. Even with such a tool you may have difficulty removing solder from certain holes that are connected to the ground plane. For these I recommend re-soldering the empty hole until it is full, then applying the soldering iron to the bottom of the hole while trying to push the part's lead through from the top. If even this does not work, just bend the part's lead so you can solder it to either the new or the old hole on the top side of the board, which is just as good as long as you can get a solid solder joint on there. I actually had to use this last technique when installing the Caddock resistors R223, R224, R323, and R324.

Those of you not too certain about your soldering Kung Fu may wish to ignore completely the OS-CON capacitor upgrade in the SPU. The capacitors that are already in these locations are Panasonic HFS, a very high quality type. This part of the upgrade hardly changes anything, and is difficult and frustrating. In spite of the tricky construction, my DAC-3 worked fine the first time I plugged it in, and I enjoyed the SPUK and consider it a worthwhile upgrade for anyone who has the money.

As for the construction of the standard kit itself, that is much more straightforward. Still, this kit is far more involved than the Assemblage DAC-2.6 (read: more fun!), and should take 3 to 5 hours to complete. In addition to the usual wire and jack soldering, and screwing and bolting together, you need to stuff the power supply board with all of its parts. This circuit board is only 2-layer, and arrives with no solder in any of the holes, so inserting the parts and soldering them in place is easy once you figure out what goes where. The digital board and the two output boards come fully stuffed and nearly ready to install, thank goodness. I would not want to have had to put any of the tiny surface-mount components onto these boards. The manual that accompanies the DAC-3 is very clear and easy to follow. It includes plenty of photos, stuffing guides, and even full circuit diagrams (for those modifiers out there).

The only hiccup in my construction experience was that when I first hooked it up, the phase-invert button was not working. After removing the cover and doing some detective work I discovered a small broken circuit trace on the power supply board, leading to the inactive phase-invert parts. Apparently I had accidentally nicked the board with something sharp. To repair this, I used a small flat screwdriver to scrape off surface mask from the two adjacent parts of the broken trace. Then I pre-tinned both, and then soldered on a tiny sliver of wire to jump the break. Voila!--phase inversion was restored.

I had some ideas for construction tweaks, such scraping the white paint off of the PCB screw holes where a chassis connection was desired, and using no washer here, for best metal-metal contact. I used stainless-steel #6 machine screws and washers instead of the included ones for the digital and analog boards, because stainless-steel is largely nonmagnetic. I taped one end of the SPUK's solderwick (I did not need this copper braid for de-soldering) around the AC wire pair near the connection to the power supply board. Then I spiralled the solderwick around the AC wires and soldered the end to the ground lug attached to the chassis for a good AC noise shield. Even if you don't have a left-over solderwick from the SPUK, you can buy one somewhere.

The DAC-3 played in a system consisting of a Rotel RDD-980 CD transport, Sonic Frontiers UltraJitterbug, home-built STAMINA integrated amps, B&W 804 and Coincident Super Eclipse speakers. My Scanning-Tunneling Audio Microscope Integrated monoblock Amplifiers are basically the Nelson Pass Zen amp and Bride of Zen preamp designs built on the same chassis, with a few circuit modifications. They are ultra-simple, low feedback, pure single-ended class-A MOSFET designs and clip at about 20W each (I've never needed more power). The DAC-3 was connected with either a Sound & Video Digiflex III AES/EBU digital cable, or a Digiflex Gold +Plus BNC. From transport to jitter box was Illuminati DV-75. Analog interconnects were JPS Labs Superconductor, or Sound & Video Audiflex Gold V. Speaker cables were either Goertz MI-2 bi-wire or Coincident CST I.

When first listening to the DAC-3, I was not as blown away by the sound as I had expected to be. Many qualities were wonderful, but it was not quite as detailed or transparent as the DAC-2.6. Actually, it sounded surprisingly similar to my upgraded DAC-2.

I had left the LT1357 and OPA134 opamps that come in the standard kit in their positions, in order to hear the difference that replacing them with the eight OPA627s from the SPUK would make. After plugging the OPA627s into the sockets there was better focus, and a more coherent sound to individual instruments. It was a little more dynamic, and rhythms were punchier and more precise. There was greater bass detail, and the bass was deeper and more solid. Vocals were more velvety and suave.

I was amazed at the subjectively large improvement that changing the main opamps wrought. The DAC-3 was now clearly in another class from the older DAC-2, and sounded like a whole different processor than it had with the standard opamps. Whether you just buy eight OPA627s separately from The Parts Connection for about $100, or buy the more expensive SPUK, which includes them, don't even think about listening to a DAC-3 without OPA627 opamps. If you are going to spend $1500 you deserve the kind of performance that these beauties deliver. I suspect that they provide more of an improvement than all the other parts of the SPUK combined.

Now some very sweet sounds were coming out of this processor. However, my devious mind just can't leave well enough alone, and I was already thinking of ways to modify and improve it. From the included schematic diagram of the output board circuit one can see that the BUF634 opamps, the last active devices in the circuit, are unity-gain buffers. Remember what it says on the label of Dr. Bronner's liquid camping soap: "simplify, simplify, simplify!" (I refuse to use Dr. Bronner's for mouthwash, though.) With a high-impedance preamp input like my 70Kohm STAMINA, an output buffer becomes unnecessary. If a straight wire with gain would be the perfect amplifier, how about a bent wire with no gain?

Thus was born the Horseshoe Mod. It requires U-shaped pieces of wire, and like on a horse you will need four of them. The wire for connecting the input jacks left over from the SPU works perfectly for this, or you can use cut resistor leads or any other thin solid-core wire. Cut the wire to about 3/4" lengths, then bend them into U's. Now remove the BUF634 opamps from their sockets (U211, U212, U311, U312), and plug in the U-wires connecting pin 3 to pin 6 on each opamp socket. These are the pin-holes that are third from the rear of the DAC-3 in each row. If you are not sure which pin is which on the sockets, please do not proceed with this mod.

The Burr-Brown BUF634 is a high quality buffer, but a piece of wire has even less distortion and noise. After hearing what it could do to the sound, I did all of my listening to the DAC-3 with the Horseshoe Mod. I thought this was fair reviewing practice because it's a kit to begin with, so everyone is more familiar than usual with the insides. Also, this mod does not involve any soldering, nor does it use any parts that won't already be lying around. [Please see David Rich's comments about removing thse buffers in his comments below. He strongly advises against this particular modification. --KWN]

After the Horseshoe Mod the microdynamics were even better, and the low bass had more punch. Images were even clearer and better separated from each other. The leading edges of transients were more solid. Voices were more present and human. It just goes to show, sometimes less is more.

Comparing the fully SPUK-outfitted DAC-3 to the much more affordable yet fantastic sounding DAC-2.6 with SPUK was interesting. On Pisces Iscariot by The Smashing Pumpkins, the DAC-3 had slightly more precise imaging, and instruments were more distinct from each other. The top octave was less sharp sounding, textures were more liquid. The low bass was more solid. Instruments had a little more body and character, and inhabited a wider soundstage. The DAC-3 gave more in-the-room immediacy, although the DAC-2.6 had the superiorly detailed treble.

The DAC-3's midrange was nicely more prominent on Tears For Fears' wonderful Raoul and the Kings of Spain. Rhythms were more deliberate. The reverb was more noticeable and complex, partly because the music was arising from a calmer, blacker background. In this system it had the more natural frequency balance, though in other systems the DAC-2.6 might have the upper hand in this department.

Bear with me through just one more change in the unit, which is swapping out the stock PMD-100 digital filter for TPC's $50 option, the Burr-Brown DF1704 daughterboard. Although I had no 24-bit/96 kHz sources to listen to, the change in filter altered the sound of conventional 16-bit CDs, such as Be Yourself Tonight by The Eurythmics. The DF1704 is a very good sounding filter, with a lively, detailed sound. Unfortunately in my system it made the top octave just a little too hot, so the midrange and bass sounded slightly recessed. It did some things very nicely, but the PMD-100 was slightly preferable for its smoother frequency balance.

Overall, the DAC-3 has a suave, refined sound worthy of a very high-class system. Undoubtedly, the DAC-2.6 is a much more sensible choice for many people, and in a budget system that's all you're likely to need. Yet, when equipped with OPA627s, and my Horseshoe Mod, the big DAC-3 gives a more transparent window into the original performance that is hard to resist. After becoming accustomed to its effortless dynamics, vivid soundstage, solid bass, and rich harmonic presentation, I simply can't do without it. I am purchasing the review sample. It is also in many ways the perfect foundation for future modifications, but that is another article. --DM