Having recently completed the “My Quest for a New DAC” series of articles, my first response to @The Computer Audiophile when he asked me if I wanted to review another DAC was: “¡no más!” “Even Snowmass?” he asked. Well, now hold on a minute. While I have long been aware of the PS Audio DirectStream DAC (abbreviated from here onward in this article as DS DAC), my listening had been confined to short sessions at audio shows. At the time of this conversation, Snowmass, the 8th major firmware update, had just come out and was garnering positive attention.
My interest was piqued. The ~$6000 price point is a crowded one for DACs, and buys outstanding quality. Has the series of upgrades culminating in Snowmass allowed the DS DAC to remain competitive in a tough field? Plus, promising free upgrades is one thing. Delivering them every 6-9 months over 5 years is quite a feat. It was high time I tried it out in my own system.
If you’ve read my other DAC reviews, you’ll know that I’m drawn to DACs that pay attention to the basics like the analog stage, the power supply, and the clocks. Furthermore, I’m intrigued by DACs that eschew chips and implement their own digital logic in FPGAs. The DS DAC ticks all these boxes.
Since a lot has been written and reported about the DS DAC, I won’t go into great detail regarding its design other than to highlight designer Ted Smith’s approach, which in the Snowmass release “upsamples everything (PCM and DSD) to 20 x the nominal SACD rate (56.338Mhz) and then back down to quad rate DSD (11.2896MHz).” The final stage is a passive low-pass filter.
Of course, to pull this off requires paying attention to both the basics - a low phase noise clock, well-regulated and low noise power supplies, and a well-engineered analog output stage - as well as the secret sauce - the code in the FPGA to implement Ted’s digital logic. The fact that this code can be - and has been repeatedly - upgraded to deliver refinements and improvements to sound quality is one of the most compelling aspects of this DAC’s design.
Technically, the DS DAC is a DAC/preamp, as it incorporates a “100% bit-perfect” volume and balance controls, and an output stage that can directly feed a power amp if the user chooses. The DS DAC has a comprehensive set of features to support modern connectivity and sample rates:
- I2S and USB input: PCM supported up to 24/384, DSD up to 128 (DoP)
- AES/EBU and S/PDIF inputs: PCM supported up to 24/192, DSD up to 64 (DoP)
- Single-ended RCA and balanced XLR analog outputs
- Control either via front touch panel, or a full-function remote control.
Aesthetically, the DS DAC just oozes elegance and quality. In addition to its rounded, brushed metallic sides, the highlight for me was the high-gloss, piano black top panel. This is a component you want to display prominently in your rack for all to admire, not hide away!
Network Bridge II
My review unit came pre-configured with the PS Audio Network Bridge II, which is a modular, user-installable $899 MSRP add-on to the DS DAC. Externally, it simply presents an Ethernet input for the unit. Internally, the Bridge II is a purpose-designed card, that contains an embedded computer that implements streamer/endpoint/renderer functionality for both UPnP/DLNA as well as Roon. It accepts sample rates up to 24/192 PCM and DSD64. For Roon users, the Bridge II, once attached to the network, just shows up as an available Roon endpoint. The card interfaces with the DAC through an internal I2S connection.
The Bridge II also implements full decoding of MQA streams, albeit up to the above maximum of 24/192. Why does this unit only support input rates and decode MQA up to 24/192 rather than 24/384? This is a constraint of the ConversDigital SoC (system on chip) used on the Bridge II card. While I’m atypical, and have a growing collection of DXD (24/352.8) music in my library, this is a small subset, so I don’t consider this limitation to be significant for most users. Similarly, the MQA constraint is a non-issue for me, as I prefer to stream high-resolution music through Qobuz.
Review System Topology
Primary System: Here is a diagram of how the DirectStream DAC integrated into my system.
Additional listening was done on a speaker-based setup. More detail is contained in the Associated Equipment section below.
Key Questions for this review
- How did the DS DAC sound in my system?
- How much was the DS DAC improved by upstream optimizations?
- How did the Bridge II input sound compared to other inputs?
- How does the DS DAC compare to other excellent DACs in its price range?
- How does the DS DAC compare to my (admittedly higher-priced) reference DACs?
DirectStream DAC Review Playlist on Qobuz (US)
To enable you to listen to the same tracks that I did, I have created a public playlist on Qobuz USA. This playlist includes the tracks mentioned in this review, as well as others I listened to in the course of this evaluation. Please note that in some cases, the Qobuz track will only stream at 16/44.1, whereas I may have used a local hi-res version. Still, this gives you a sense for the music I listened to for evaluation.
You’ll notice that many, if not most, of the tracks are not audiophile-grade demo-quality recordings. I don’t buy DACs only to listen to demo tracks. Most users buy gear to listen to the music they love, so this is how gear should be evaluated, in my opinion.
I spent several weeks just enjoying the DS DAC in my system, and it was indeed a joy to listen to. At this price point, sonic expectations are brutally high, but the DS DAC delivers on them resoundingly. Its sonic character is muscular and organic. I’m not one to throw around phrases like “analog-sounding,” so I will instead say that the music has effortless pacing, and instruments sound natural and well fleshed-out.
Jordi Savall’s new album Ibn Battuta: Le Voyageur de l’Islam (Alia Vox, 24/88.2) is a tour de-force, with musical traditions represented from West Africa to China. One of my favorite tracks from the album is Kouroukanfouga, a haunting, mesmeric piece played on the valiha and oud, with tabla accompaniment. Catch the video at this link. The DS DAC rendered this beautifully, from the rich, woody tone of the oud to the intricate plucking of the valiha. There was plenty of solidity in the bass for the tablas. While this track has a low dynamic range, the DS DAC conveyed a sense of excitement and power even at modest listening levels. I have found this to be a characteristic of DACs with excellent dynamics.
I’m a sucker for submarine war movies, and The Hunt for Red October is up there as my all-time favorite. Hymn to Red October (16/44.1, Geffen) is the Russian male choral piece that is immediately recognizable to anyone who’s seen the movie. While this isn’t the greatest recording, the DS DAC reproduced it with satisfying growl of the male voices, and excellent dynamics on the crescendos.
And so it went. I threw everything from 16/44.1 PCM to DSD128 content at the DS DAC, and it handled it all with aplomb and authority. The more I listened, the more I appreciated just how musical, how natural this DAC is. Tonality was spot on across the spectrum, and the presentation had calmness and authority.
Perhaps the biggest indicator of how good the DS DAC is was how little time I spent thinking about the DAC, compared to thinking about what to listen to next.
Effect of the Upstream Chain - USB
This is a test I’ve done in my past reviews, but I’ll repeat the rationale here. If you glance at my system topology above, you’ll see I have optimized my digital transport chain upstream of the DAC. This is apparently a sticking point with some, who argue that this part of the digital chain shouldn’t matter to sound quality, especially with DACs that have been engineered with galvanic isolation and buffered inputs. All I can say is that in my listening experience, this part of the chain makes a profound impact on sound quality, and I have invested accordingly.
The experiment is to compare 2 digital chains, both using Roon as the music server:
- The optimized chain shown in the diagram below, and
- The upstream chain replaced by a MacBook Pro laptop running on battery, driving the the DS DAC directly with an Amazon USB cable.
The DS DAC is no exception to other DACs I’ve tested: it sounds much better with the optimized chain. The improvements are fairly comprehensive:
- A reduction in harshness and glare,
- A more holographic (bigger, wider, deeper) image,
- An increase in clarity and detail, and
- A deeper and better articulated bass.
The key takeaway from this section, and my recommendation, is that investments in upstream USB optimization will enhance the already excellent sound quality of the DS DAC. While my upstream optimizations are extreme, even modest enhancements will pay rich dividends.
Effect of the Upstream Chain - Ethernet
During the time I was reviewing the DS DAC, I also had on hand a review sample of the SOtM sNH-10G High End Audio Switch. This switch was recently reviewed in Audiophile Style, and the reviewer found it delivered significant sound quality improvements.
This too is a contentious topic, but my experience has been the same as the reviewer’s. With my reference DACs, using the SOtM switch with a good PSU upstream of the NUC endpoint as shown in the system diagram, the sound quality improvement was just startling!
SOtM designed this switch from the ground up for audio chains, using upgraded components like Evox capacitors, linear regulators, and using their low-phase noise sCLK-EX clock board for the Ethernet clock. Additionally, this clock board can be disciplined by a 10MHz reference clock. In my system, I used my existing Mutec Ref-10 OCXO reference clock
Could using the SOtM switch upstream of the DS DAC’s Ethernet input produce similar sound quality improvements? And would this propel the sound quality of this optimized Ethernet chain to the same level as the optimized USB chain?
To find out, I did a listening experiment through the DS DAC’s Bridge II input, comparing:
- The optimized Ethernet path, connecting the DS DAC Bridge II to a port on the SOtM switch, vs
- The DS DAC Bridge II ethernet input connected directly to my router.
And indeed, the results were consistent with my USB optimization findings. The improvements with the optimize Ethernet path via the SOtM switch were: less harshness, a bigger image, more clarity, and better bass articulation.
My final test was to compare:
- The optimized USB chain, fed to the DS DAC USB input, with
- The optimized Ethernet chain, fed to the DS DAC Bridge II input.
This was very interesting. In my reviews of other DACs where I’ve tried this - the Ayre QX-8 and QX-5 Twenty - I found a significant gap between the USB and Ethernet inputs, with the USB sounding significantly better.
On the DS DAC, the results were much closer. I found the optimized USB chain still had the best sound quality, but it was close enough where I could see a user making some tradeoffs. My USB chain is extreme and involves a fair amount of expense and so-called spaghetti, i.e. accessories, power supplies, cables etc. For a lot less cost, and a far simpler topology, a user could use the optimized Ethernet path. Yes, you give up a small dollop of sound quality, but save a lot of expense.
What about I2S?
As my experiences have shown me, the question of “which input is best?” for a given DAC often devolves to a question of “what upstream optimizations do you have?” Just like with USB and Ethernet, optimal sound quality with the I2S input in my system would require a high-quality USB to I2S DDC (digital to digital) converter, with low phase noise clocks and good power supplies. I did not have such a device on hand to try it, so I can’t comment on the I2S input of the DS DAC.
Comparisons with Other DACs
This is where the rubber meets the road. As excellent and enjoyable as I found the DS DAC to be, it is competing with some outstanding products in its price range. How does it compare to these DACs in head-to-head listening tests in my system? And I do mean head-to-head listening, with both DACs on hand, and with careful matching of levels. I did many of these comparisons both on my primary system, as well as the speaker-based system listed in the Associated Equipment section below.
Let’s look at the competitive lineup. All prices listed here are US MSRP. Competing head to head with the DS DAC ($5999 or $6899 with Bridge II) are the Chord Hugo TT 2 ($5795) and the Ayre QX-8 ($4950 with USB, $5450 adding Ethernet).
As for reference DACs: my reference until recently has been the Ayre QX-5 Twenty ($7650 with USB only, $8950 adding Ethernet). My current reference “DAC” is actually a combo of the Chord Hugo M Scaler ($4995) and TT 2 ($5795).
Let me be clear about the purpose of comparison with these higher-priced reference DACs. The intent isn’t to shame the DAC under review. I don’t expect the DS DAC to exceed or even match the SQ of my reference DACs. Rather, it’s to see how close it comes in different aspects of performance, and to give more points of comparison to readers who may be familiar with these DACs.
It’s important to keep some things in mind in the following sections. First, every DAC featured sounds excellent in isolation. There are no mediocre DACs here. It needs to be emphasized that any differences in sound quality I describe are relative, and do not change the inherent quality of these DACs. Finally, at this price and quality point, we are well into the diminishing returns of the price-performance curve. Whether sonic differences are worth the price differential is not something I’ll get into, as that is a personal calculation for every buyer to make.
While I’ll indicate my preferred DAC in each comparison, you as a prospective buyer would be ill-advised to just take my word for it. My intent here is to describe the character of each piece, but every DAC in this section is worthy of an audition in your own system before making any purchase decisions.
Comparison with Ayre QX-8
Ayre and PS Audio are located 10 mins from each other in Boulder, CO, but when it comes to DAC design, they’re miles apart. In contrast to the DSD-upsampling, FPGA-based DS DAC, the QX-8 is based on the late Charley Hansen’s approach of marrying an ESS chip with Ayre’s proprietary minimum-phase filter and his zero-feedback, Diamond output stage. The characteristic both companies share is deep expertise in the meat and potatoes of building high-end equipment: power supplies, analog stages, isolation, and chassis design.
Sonically, both DACs have very different strengths.
On Respighi’s Ancient Airs & Dances (DSD64, Mercury Living Presence), the QX-8 projected a bigger and, ahem, airier image. I’ve found one of its enduring strengths to be excellent imaging. The DS DAC excelled in portraying the music with more density and solidity. On the Laura Soave movement of Suite 2, P. 128, the lute had more heft, and the rhythm was better articulated and went deeper.
As Freddie Mercury’s final album, Queen’s Innuendo (2011 remaster, 16/44.1 Qobuz) has bittersweet memories for me, and is one of my favorite Queen albums. On I’m going slightly mad, the QX-8 projects the ambience of the recording really well. Mercury’s voice has a holographic, if slightly ethereal, quality. On the DS DAC, the bass lines and drums had a satisfyingly visceral feel to them. While the image was smaller, and the keyboards a bit less soaring, there was also a calmer, less fatiguing tonality that is very pleasing.
I was honestly hard pressed to prefer one of these DACs over the other. Since I promised not to waffle, I’ll give the nod to the DS DAC for its physicality and calmer, natural presentation.
Comparison with Chord Hugo TT 2
We have a saying here in Texas that goes: “I wasn’t born in Texas, but I got here as fast as I could.” It came to mind as I was contemplating this match up. Applied to incoming music data, substitute DSD for Texas and you describe the DS DAC. Substitute PCM, and you have the Chord TT2. These diametrically opposite philosophies continue in the filtering space. The DS DAC is based on the idea of a simple LPF (low pass filter), while the Chord’s WTA (Watts Transient Aligned) is a very long (98,304) tap-length filter. Could these two extremes really produce similar levels of sonic excellence?
The track Mombasa from the soundtrack of the movie Inception (16/44.1, Reprise) can be very revealing of DAC differences. The deep pounding beat is mixed in with delicate cymbals and fine brush strokes. On the DS DAC, the drum beat had just a bit more body and weight. Where the TT2 excelled was with the texture and detail of the cymbals and brushes. There was more clarity and coherence to the music, as if a knob had been turned to improve focus.
On Suscepit Israel from Arnesen’s Magnificat (DSD64, 2L), the DS DAC’s rendition, when heard in isolation, was gorgeous. Yet, the TT2 allowed the voices of the chorus to be heard more clearly as a collection of individuals rather than an amorphous whole. The soaring soprano’s voice was more natural and real. On the DS DAC this track was beautiful; on the TT 2 it was downright moving.
These were not major differences. With DACs of this quality, this is the level of subtlety with which one must make distinctions. On balance, the transparency and coherence of the TT 2 tipped the scales for me, but someone whose ears favor tonal richness over transparency could well have picked the DS DAC.
Comparison with QX-5 Twenty
Next up was the Ayre QX-5 Twenty, which for the last two years had been one of my reference DACs, and is still one of the best I’ve heard. What its higher price buys you is better power supplies (4 separate transformers, better regulation), better clocks (SC-cut crystal oscillators), and upgraded components.
On the Tempo di bolero movement of Concierto Andaluz, from the Rodrigo album by Los Romeros (DSD64, Mercury Living Presence), excellent as the DS DAC was, there was clear daylight between it and the QX-5. The QX-5 had more micro-detail, and a better delineation of instruments. Its image was bigger in all 3 dimensions, although the DS DAC matched it on density and weight.
A.R. Rahman’s genius transcends languages and cultures. His album Vande Mataram (16/44.1, Columbia) is a perfect example. In Revival, Rahman fuses the Indian national hymn Vande Mataram with a hypnotic beat, decorated with interesting percussive flourishes. Both DACs rendered this piece stunningly. They were neck and neck in the extension of the bass, the taut and controlled pacing and rhythm, and the mellow tonality of the song. Where the QX-5 pulled ahead was in the ambience and textures. The sax was just that little more soaring, the bell notes had more micro-detail, and seemed to linger a bit longer.
I said this was an unfair comparison, and it was. The DS DAC never put a foot wrong, but economics does have a role in high-end audio. You get what you pay for (usually). In this case, the much higher-priced QX-5 Twenty buys you that extra dollop of refinement and sound quality.
Comparison with Chord Hugo M Scaler/TT 2 combo
After recently reviewing the combo of the Chord Hugo M Scaler with the TT 2 DAC, it become my new reference DAC solution. In a previous section, I described how the DS DAC fared relative to the comparably-priced TT 2 DAC by itself. I have found the M Scaler, which implements Chord’s million-tap WTA filter, and upsamples all streams to 705.6/768 kHz, to be a real game changer. The combo of the HMS/TT2 surpassed even the mighty QX-5 to become my new reference. At a price point of $10,790, this combo is 80% more costly than the DS DAC. How did the DS DAC compare?
My favorite recording of Mahler: Symphony No. 8 is by Sir Georg Solti conducting the Chicago Symphony Orchestra (24/96, Decca). This so-called “Symphony of a Thousand” puts ridiculous demands on the entire audio chain, and it is hard for any system to capture the thunderous dynamic range of a Mahlerian orchestra in full voice. Lesser DACs struggle to project a big enough or dense enough image, to portray instruments that have the flesh and bones to sound realistic, and most importantly, to maintain coherence during dense passages.
The DS DAC did all that with flying colors. It matched very closely with the HMS/TT2 combo on the size and density of the image, and the dimensionality of the instruments. It even had a small edge on the deep bass of the tympani, double basses, and tuba. Where the HMS/TT2 pulled ahead was in three main areas - first, the ability not just to resolve multiple instruments, but to follow instruments even as they recede into the background. Second, the HMS/TT2’s image had more front-to-back depth, and the ability to place instruments in space was spooky. And third, the HMS/TT2’s particular strength is in the portrayal of voices. On the track Neige, neige, du Ohnegleiche, the clarity and the emotion of the soprano was better conveyed on the HMS/TT2.
To reiterate, I had no expectation that the DS DAC could surpass a combo almost twice its price. The fact that it held its own in some aspects is impressive.
The PS Audio DirectStream DAC is a 5 year old design that, with the latest “Snowmass” release, still competes with the best DACs in its price class. How many DACs can make that claim? Many DACs are user-upgradeable, but very few have received the steady stream of free firmware updates that delivered sound quality improvements - not just bug fixes - that enable the DS DAC to retain its competitive edge. Plus, designer Ted Smith tells me he still has some (albeit shrinking) capacity left in the DS DAC’s FPGA to continue delivering improvements in the future. These include trickle-down findings from the cost-no-object Obsidian “Ted Smith Signature” DAC he is developing. That is a stunning value proposition, and one very few companies can match.
The DS DAC is an intensely enjoyable piece of gear that allows you to connect deeply with your music. Its biggest strength is its ability to make music sound natural and real. Refined tonality, dimensionality, dynamics, and grunt - it has it all. And to my pleasant surprise, this was the first DAC I’ve reviewed where I could optimize the Ethernet input to a level of sound quality close to that of USB.
Did the DS DAC surpass its competition in every attribute? No it did not, but that’s not how I measure success. The DS DAC’s success comes from having earned a spot on the shortlist of any audiophile shopping for a DAC in this price range. And just like thousands of delighted current owners, I fully expect many more will find it to be the one that speaks best to their soul. Highly recommended!
Product: PS Audio DirectStream DAC (US MSRP $5999, or $6899 with Network Bridge II)
Documentation: DirectStream DAC Manual
Roon Core: Custom server: H370M-ITX/ac, i7-8700T, 8GB RAM, HDPlex H3 case,
32GB M10 Optane SSD for OS, JCAT Net Card Femto,
running Audiolinux (in RAM) and Euphony OS
Music Storage: Synology NAS DS916+ 4-bay, attached to router via Ethernet
Endpoint : Intel NUC7i7DNBE in Akasa Plato X7D case, 8GB RAM, running
Audiolinux (in RAM) and Euphony OS
Headphone Amplifier: Cavalli Liquid Gold
Headphones: Sennheiser HD800 (Super DuPont Mod), Meze Empyrean
USB Regenerator: SOtM tX-USBultra SE
Ethernet Switch: SOtM sNH-10G SE switch
Reference Clock: Mutec Ref 10 10MHz clock driving the tX-USBultra and sNH-10G
Power supplies: Utpone LPS-1.2 for switch & tX-USBultra, Paul Hynes SR-4 for JCAT Net
Card Femto, HDPlex 400W ATX LPS for Roon server
Power Details: Dedicated 30A 6 AWG AC circuit, PS Audio P5 PerfectWave Regenerator
Power Cables: PS Audio AC-12 (wall to P5), Cardas Clear Beyond (Cavalli Amp),
Cardas Clear (Mutec Ref-10), Cardas Clear to all DACs under test
Cardas Clear (SR-4, SR-7), Pangea AC-14SE MkII to all PSUs,
Pangea AC-14XL to Dectet
USB cables: Phasure Lush & Lush^2 USB
AES/EBU cables: Cardas Clear
Clock cables: Habst 5N Cryo Pure Silver
BNC cables: Habst 5N Cryo Pure Silver
Ethernet cables: SOtM dCBL-Cat7, Supra Cat 8
DC cables: Audio Sensibility Signature Silver (LPS-1.2)
Paul Hynes fine silver (SR-4, SR-7)
Interconnects: Cardas Clear XLR balanced (DAC to Amp)
Headphone cables: Cardas Clear balanced and SE cables for all headphones
Accessories: Synergistic Research Tranquility Base XL UEF with Galileo MPC
Alternate system for speaker-based evaluation
Roon Player: Innuos ZENith MkII SE
Preamp: Audio Research Reference 6
Power Amp: Hegel H30
Speakers: Magnepan 3.7i
Subwoofers: 2x Rhythmik 12g
USB Regenerator: SOtM tX-USBultra
Reference Clock: Mutec Ref 10 10MHz clock driving the tX-USBultra
Power supplies: Paul Hynes SR-7 for tX-USBultra
USB cables: Phasure Lush & Lush^2 USB
BNC cables: Amphenol and Digi-Key
Ethernet cables: SOtM dCBL-Cat7
DC cables: Paul Hynes fine silver (SR-4, SR-7)
Interconnects: Cardas Clear RCA and XLR (DAC to preamp)
Accessories: Synergistic Research Tranquility Base UEF
Many thanks to Cardas Audio for providing a loom of Cardas Clear cables to aid evaluation.