As mentioned in my Benchmark DAC1 comments, I had been interested in trying the DAC2, with its significant improvements in components and specifications. Benchmark ran a promotion for a few weeks in August 2015 when they were offering a pretty good trade in credit on a DAC1 if one purchased a DAC2 from their own online store. I took the plunge and am very happy.
Benchmark had a solid resume of prior work in the DAC1 that was technically outstanding if slightly soft and dry sounding, and reviews of the DAC2 were strong, so along with the trade-in credit, the risk in upgrading seemed low.
The DAC 2 is definitely not euphonic, but it's highly accurate and revealing. Initial impressions are that there is vastly more usable resolution and clarity. Noticeably greater accuracy and detail in the reproduction of the percussive attack of piano note strikes, for example, could be a consequence of greater time and dynamic resolution. Voice is also much more realistic sounding, with a more faithful rendition of the original signal, particularly breath and vocal dynamics, but also in fine details of timbre. So far everything sounds vastly more real.
The DAC2 measured specifications are a significant improvement over the already very impressive DAC1, due to the very high resolution ESS SABRE DAC, better components and improved circuit topology. Noteworthy, there is more power supply regulation, something I felt was lacking after a casual examination of the DAC1. More sophisticated on-board regulation is added to the DAC2 with more switch-mode regulator sections compared with relatively few linear regulators of the DAC1. Properly implemented, this can improve signal integrity by reducing power supply modulation effects. An expected result should be cleaner sound, particularly during complex passages such as those with concurrent loud and soft dynamics. The DAC2 does seem less "confused" by loud, complex passages than the DAC1, and that could be aided by significantly improved voltage regulation.
However, the impressively improved measurements don't really prepare for the emotional impact of music played through the DAC2. Improvements in objective aspects of performances are there: I can hear keyboard, voice, drum, string and jazz saxophone articulation, timbre and instrument sound much more clearly. The intention of the artists through the mechanics of their playing is much more clearly audible.
But the communication of emotion is even more greatly improved. Jazz swings better. Classical storms, soars and sings better. Rock would rock harder if most of it didn't have such dismal sonics to begin with. Great playback can't really improve bad recording. That said, well-recorded pop like the 96 kHz x 24-bit Fleetwood Mac Rumours DVD-Audio remaster reveals great writing and outstanding musicianship. 2014's fantastic 96 x 24 remaster of Elton John's Goodbye Yellow Brick Road absolutely rocks in its many special ways. At times, bluesey, orchestral, rocking, funky, overtly beautiful, and occasionally silly, the added resolution of this recording and this playback is magical. The title track really engaged me in a way I had never experienced before; a spine-tingling, outrageous, swirling serenade. The whole album charmed me in unexpected ways, and I hadn't really considered myself much of an Elton John fan until hearing this album this well. Eric Johnson's song Trademark from the 192 kHz x 24-bit HDTracks.com FLAC digital download of his album Ah Via Musicom definitely rocks better than the CD version, even if it's not a perfect digital-age recording. The rest of the album is presented superbly in this format over DAC2, and the added resolution of both the recording and playback is readily apparent.
With the DAC2, emotional content is presented in a way that's laid bare; a major barrier to the sentiment of the music due to sonic degradation has been removed. All this is improvement over the already very good DAC1.
The improvements in music enjoyment are not linear. The DAC2 may have 9 dB better signal to noise ratio than the DAC1, which is huge at almost 100 times better, and 3 dB lower noise and distortion, but the improvement in re-creating an artistic moment is a seemingly effortless quantum step.
(That step wasn't actually effortless; it was the result of a lot of careful research, design and component improvements within a strong system concept; hard, solid, engineering work.)
The combination of a much lower noise floor and greater resolution lets me hear much deeper into the music. 10 decibels deeper actually sounds about right, but these are now so close to to the practical limits of hearing and technology that it gets harder to make authoritative judgments. DAC2 sounds about that much better, however. The DAC1 approached the limits of reasonably priced professional equipment during its era, but ran up against some audible and technical limits. It was a reasonable price/performance compromise for its time, price point and available technology. But Benchmark has helped push the state of the art much further along since that time with the DAC2. The increase in price over DAC1 is more than borne out by the increase in performance. It's significantly more expensive but even more significantly better. That makes it a good value even at the higher price.
Test conditions were informal A/B tests using CD and 96/24 DVD over Toslink from Pioneer Elite DV-58AV, and lossless FLAC CD rips over 96/24 USB from Foobar2000 Windows audio player, on Sennheiser HD 600 headphones with Cardas cable. Also, having listened daily through the DAC1 for several years I was highly familiar with its sound, so the DAC2 differences were immediately apparent. It just sounds much better, and that leapt out to me immediately. It was not at all subtle, and that surprised me.
The DAC2 can't be perfect (what can?), but it's extremely good. For its price, it's a relatively great value for how close it gets to perfection. It's also highly accurate, and that's exactly what high fidelity was always supposed to be about. It doesn't color; it doesn't flatter; it tells what's there, honestly and with relatively little artifice. It resolves like crazy, right down to the emotional essence of the music. This is something the DAC1 approached, but didn't reach nearly as convincingly. If the DAC1 was a breakthrough, the DAC2 is a much greater one because of that quantum leap in emotional connection to the music.
One of the questions I had from DAC1 was whether improved jitter performance in the DAC2 might make computerized music over USB more viable. While DAC2 does an outstanding job handling jitter (as did DAC1 within its respective resolving power), I still hear better clarity and resolution when playing optical discs directly from the disc player versus the same discs ripped and played from PC over USB. Again the sources of jitter could be many from the USB medium, to a computer bus, to any number of places where timing inaccuracies can creep in, all the way from the ripping process to the USB output and everything in-between.
The sonic character of jitter remains familiar: less clarity, more hash, nosier sound, less detail, less realism. Everything about DAC2 seems improved over DAC1, including jitter handling, but jitter does still seem to degrade sound quality somewhat. The good news is that the overall performance improvement of the DAC2 is large, so the effect of jitter may be lessened in an absolute sense, but it does still have some effect in my current system. (I have not yet tried the Lynx AES PC interface board, and the DAC2 HGC lacks the dedicated XLR AES input of a different DAC2 model.)
As an interesting side note, Benchmark's add-on USB 2.0 driver, which enables 176 and 192 kHz x 24-bit audio rates over USB, does seem to improve sound quality over the native Windows 7 USB 1.1 driver. The effect is subtle but seems to be a slightly gentler, softer sound which is often an indicator of better performance.
Presumably the timing requirements of USB 2.0 are more stringent than 1.1, due to the much higher data rates. If so, that should imply greater timing accuracy and smaller timing errors in general, leading to lower jitter.
DAC2 follows many design concepts from DAC1, but differs significantly in the sophistication of the power supplies. For example the excellent National Semiconductor (now owned by Texas Instruments) LME49860 op amp is carried over, but is used in even more positions in the DAC2. (It seems to replace the 5532 used in the headphone circuit of the DAC1, for example.) The headphone amps themselves remain the National Semiconductor (TI) LME49600.
The USB 2.0 input receiver is the SMSC USB3318. An XMOS GT1423L1 (possibly read incorrectly) USB audio processor follows the receiver. The Sample Rate Converter is now the Burr Brown SRC43921. A Xilinx Spartan-6 XC6SLX9 FPGA seems to be the brains of the operation. The DAC is the aforementioned ESS SABRE 32-bit. There are several high quality crystal oscillator modules to provide clocking to various digital sections.
The power supply is changed from a traditional analog supply with a Talema toroidal transformer, bridge rectifier and relatively large capacitors feeding linear regulators in the DAC1, to a small but high-quality-looking, single voltage output, switching power supply feeding a more elaborate multi-component, multi-voltage regulator section on a front corner of the main board. (The single voltage output from the switching power supply apparently floats since its carried over two wires, and it's not obvious at a casual glance how the power and signal grounds are resolved.) That section of the board is as far away as possible from the analog outputs which are at the opposite back corner. The new regulation features large computer style inductors and power supply bypass capacitors liberally scattered throughout the board.
The multiple, low-voltage regulation includes a pair of Linear Technology LTC3633A step down regulators. Presumably these are switch mode also, since there are some SCR packs along with the big inductors and capacitor chips. (It looks like a switching power supply component entourage.) About 20% of the board seems dedicated to this much more elaborate power supply section, so it's pretty clear that the local condition of the power got a lot more attention in this design, and rightly so.
A separate control panel circuit board attached to the front panel holds the buttons and displays. It connects to the main board with a ribbon cable via headers. The front panel motorized ALPS potentiometer and headphone jacks are board mounted. At the rear, the USB, Toslink and XLR connectors are board mounted, and the good but not obsessive quality RCA connectors are panel mounted with soldered solid wire connecting them to the main board. The RCA solid wire leads have long ferrite beads on them to block high frequency noise and interference from leaving or entering via the analog connectors. Relays switch the analog inputs, as on the DAC1.
Board layout is very tidy, with discrete components as near as practical to the active parts they serve. Component layouts have a nice symmetry that respects path lengths, shielding and electromagnetic fields. There are also generous ground planes and guard areas between different circuit functions. I did not remove the board to see what's on the bottom side. Most of the components may be on the top side. The board is certainly multi-layer and appears to be very good physical and manufacturing quality that one would expect to find in a professional instrument. It's not a cost-cutting, consumer grade of board. Nor is it extravagant and pretentious "audiophile quality".
The black-painted steel sheet metal case is stiff and tight, if a bit thin. The AC power entrance is nicely sequestered into its own fully-shielded, small cloister which includes fuses and filters. The power supply is universal from 88 to 264 Volts AC at 50 to 60 Hertz.
The CNCed Aluminum front panel is cleaner and classier than the also nice DAC1. Buttons and LED indicators are well-organized, meaningful, functional and usable. Some different colors for the LED indicators of bit depth and data rates might be nice, but the meanings are mostly discernible by position, even without referencing color. For example the 14 input selection, bit depth and sample rate LEDs are all blue. Mute is red, and polarity inversion is yellow. The ALPS volume control has a nice feel, and its knob is functional if a bit utilitarian.
Update: Benchmark published a circuit description of the analog parts of DAC2 on 1 June 2016. A description of the digital circuit followed. A description of the power supply made the third part of the series.
Benchmark DAC2 can play unencrypted DSD with the use of an appropriate ASIO driver. For example HDTracks has a 2.8 MHz DSD of the first Boston album and also much more recent classical recordings in DSD format. They play just fine either with the ASIO driver for direct DSD output or converted to PCM in the PC before being sent to the external DAC. The sounds of DSD and DSD converted to PCM in the PC are different, and I think DSD is softer and more subtle-sounding. Also perhaps a bit more realistic.
The Boston album isn't a great test since it sounds like there are some tape losses from the remastering from aging analog tapes. The album was also remixed by creator Tom Scholz, which makes direct comparison with the original difficult, but the original CD definitely sounds much worse overall, despite the degradation of the master tape which was used to do the recent remix and remaster. (As an aside I enjoyed the Guitar World interview with Scholz who refuted the label of "corporate rock sound" for the first Boston album, when in fact it was recorded and mixed entirely in his basement as a personal labor of artistic love. There was zero input from any corporation in the making of the first album.)
Selecting the ASIO driver in Foobar2000 is done under File -> Preferences -> Playback -> Output -> Device. It's slightly inconvenient to change audio drivers for playing DSD, but perhaps worth it for slightly better fidelity with DSD source material. That said, I don't plan to buy more DSD; it was mainly an experiment borne of curiosity. Well recoded PCM can be very good, and I'm not convinced that the massive ultrasonic peaks inherent in DSD are desirable, even if they can be filtered out effectively in the digital domain. Radical transformations needed for DSD encoding and filtering may have effects in audible ways.
I also used these DSD and PCM comparison tracks published by OPPO. DSD sounds softer and more effortless, but less direct played back in its native format.