When it comes to network streaming music, there are many ways to - quick, cover Fluffy’s ears - skin a cat. Network streaming refers to an architecture of music playback, where the functions are distributed: a music server catalogs and manages the music files, but then transports them to a streamer, or endpoint that renders the music by sending them to a DAC for playback.
This distributed architecture has been widely adopted, and has spawned outstanding products like the Sonore UltraRendu and the SOtM sMS-200ultra, to name just a couple. These streamers have been designed to run on low power, and produce as little electrical noise as possible. Typically, these run a stripped-to-purpose version of Linux, and expose a simple web UI to configure and manage. While everything matters when it comes to sound quality, a couple of the primary factors that affect sound quality profoundly, and go hand in hand with these streamers, are PSU quality and clock quality.
Over the last few months, a couple of new factors have been discovered by hobbyists here on CA that enhance sound quality even further. First is the finding that running Audiolinux (an audio-focused variant of Arch Linux) fully in RAM (the so-called ramroot mode) enhances the sound quality of both the streamer and the music server. Second is the discovery that Intel’s 7th generation NUC (next unit of computing) boards - when combined with a fanless case, running Audiolinux in RAM, and powered by a good PSU - sound shockingly good as network streamers, and even as music servers.
The DS-1 Streamer
It is against this backdrop that we do a quick take of the DS-1 streamer from The Linear Solution (TLS), a small Seattle-based audio company, led by its founder, Adrian Wun. Adrian was one of the first proponents of the RAM-resident OS as a means to enhance SQ, and he demoed an experimental setup to me at AXPONA last April. He’s been hard at work to bring his products to market, which include a standalone music server, The One Music Server, and the DS-1 streamer. He recently sent over one of the first batch of the DS-1 for me to evaluate.
What is the DS-1? Basically a NUC on steroids, configured to function out of the box as a Roon and/or NAA endpoint. At its heart is the familiar (to CA hobbyists) NUC7PJYH board with a J5005 Pentium CPU. The unit comes with 8GB of RAM and boots from a USB stick. But this is where the similarity with the stock NUC ends. The case is a custom design built in Switzerland to TLS's specifications, and I have to say - it is gorgeous. TLS claim it is a "Custom design fanless enclosure with noise treatment to reduce vibration and resonance," and I can quite believe it. Additionally "EMI/RFI Treatment" is claimed, but I need to get more details from Adrian.
Another difference is the memory modules. TLS have worked with a memory vendor to source modules customized to their latency specification.
The final, and most important difference from DIY NUCs is the OCXO clocking. TLS is using oscillators from Connor-Winfield, which Adrian tells me are not consumer grade, but closer to mil-spec. However, I have no further details on their phase noise characteristics. More notably, TLS have worked with Intel to enable them to use their OCXO clock to apply not just to USB, Ethernet, or the system clock, but to all 3. As you'll read below, this is a significant outcome.
Here are my initial impressions, after about 100 hours of burn-in. Given that this unit has an OCXO clock, I expected that it would need some time to stabilize thermally and burn-in. Indeed, the unit's sound improved fairly dramatically over the first 24 hours, after which I stopped really paying attention and have just been using it.
But how does it sound? With the DS-1 in the chain, sound quality is really excellent! I very recently had a NUC7CJYH in an Akasa Newton JC case in-house for listening. This was my first chance to hear these Intel NUC boards. In addition to a tonally rich and dense presentation, one of the attributes that stands out is dynamics. The sound quality is big, open and unconstricted.
Great as that NUC sounded, the DS-1 sounds significantly better. The improvements are exactly what I have heard with clock improvements before. Clearly the OCXO is the reason for the bulk of this improvement:
- more airy, open,
- bigger, deeper soundstage,
- more micro-detail,
- more refined treble, and
- better bass definition and clarity.
Now here is the interesting thing. As I have previously reported here on the forum, the NUC7i7DNBE in the Akasa Plato X7D case sounds significantly better than the CJYH. So, how do the DS-1 and the NUC7i7DNBE compare? The DS-1 is still the better sounding, though by a smaller margin. But...
- the improvements, as listed above, are on another axis than the i7's strengths
the i7's strengths over the CJYH are still apparent when compared to the DS-1:
- more dense, fleshed out, and dimensional,
- more bass extension.
Obviously, as audiophiles who are never satisfied, we would love to get the benefit of the OCXO AND the benefit of the i7DNBE. Adrian is aware of this, and is looking at a follow-on product based on this idea.
A note about power supplies. Like all streamers I’ve tried, the DS-1 scales to higher levels of sound quality with increasing quality of PSU. It sounds great with the included SMPS supply, but SQ rises to another league once powered with a quality PSU like the Uptone LPS-1.2, set at 12V. The LPS-1.2 has a bigger sound stage, with sweeter mids and a more refined treble. This PSU is only able to supply 1.1A max, which may or may not be sufficient for the DS-1, depending on what the connected devices, especially the DAC, draw from the USB VBUS. In my setup, the connected USB device is a SOtM tX-USBultra, which has its own PSU to power the USB VBUS to the DAC, and consequently draws very little from the USB port it is connected to upstream. In this system, the LPS-1.2 worked successfully.
Stepping up to the Paul Hynes SR-4 yielded another uptick in treble refinement, where music sounded even more relaxed and effortless, emerging from a blacker background.
Sadly, I no longer have streamers like the sMS-200ultra Neo on hand to directly compare. I expect the DS-1 to fare very well in a head to head comparison, although if reports of SQ improvements in the latest Enuhasu firmware are substantiated, we may have a race on our hands. Either way, it's great for us customers to see SQ going up!
Here are some functional notes on the DS-1:
Since it is at heart a NUC7PJYH, you can still tune the BIOS to your heart's content.
- As per findings here on the forum, I went ahead and turned off SpeedStep, Bluetooth, Wifi, SATA, Audio. This improves SQ.
- While TLS supply their own USB stick to boot from, you can substitute your own, if you like dabbling. I booted off my own customized Audiolinux stick, with no issues.
- The DS-1 Ethernet port currently negotiates to 100Mbps. I'm not sure if this is a tuning choice, or related to the next point.
- DXD streaming (24/352.8, 24/384) does not work - I encountered severe skipping. This is a known issue that TLS are working to resolve with Intel.
The DS-1 retails at $1599, which seems like a very fair price. Since I'm a Roon user, the USB stick that came with the box just boots up as an immediately-usable RoonBridge or NAA, requiring no intervention or configuration. If you want to use a different ecosystem - like UPnP, LMS/squeezelite, MPD, etc, and are not adept at doing Linux command line operations, I'd strongly recommend contacting Adrian before buying to get a clear understanding of how he will support you to get up and running.
While it is easy to characterize the DS-1 as a reaction by manufacturers to the recent findings here of NUCs' sound quality, Adrian deserves more credit than that. He has had his AudioLinux-based DreamOS in the pipeline for some months now. It takes time to bring a product to market, and the incorporation of OCXO clocking further distinguishes the DS-1 from the DIY versions some of us have been playing with.
In summary, the DS-1 is worth serious consideration as a NUC-based network streamer and endpoint, to DIYers and non-DYers alike. The appeal to non-DIYers is obvious. You get an attractive, well-engineered turnkey product that allows you to experience and enjoy excellent sound quality. For DIYers, the attraction is the OCXO clocking, which is not easy to accomplish via DIY.
Music Server: Dell XPS 8700 running Audiolinux in RAM
DAC: Ayre QX-5 Twenty
Headphone Amplifier: Cavalli Liquid Gold
Headphones: Sennheiser HD800 (Super DuPont Mod), Abyss AB-1266 Phi CC, Focal Utopia
USB Regenerator: SOtM tX-USBultra
Ethernet Switch: The Linear Solution OCXO switch
Reference Clock: Mutec Ref 10 10MHz clock driving the tX-USBultra
Power supplies: Utpone LPS-1.2 for switch and tX-USBultra
Paul Hynes SR-4 for the DS-1 and other NUCs
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, QX-5, and all PSUs)
USB cables: Phasure Lush^2 USB
AES/EBU cables: Cardas Clear
Clock cables: Habst 5N Cryo Pure Silver
Ethernet cables: SOtM dCBL-Cat7 (switch to NUC), TLS cable (switch to router)
DC cables: Audio Sensibility Signature Silver (LPS-1.2), Paul Hynes DC3FSXLR fine silver (SR-4)
Interconnects: Cardas Clear XLR balanced (DAC to Amp)
Headphone cables: Cardas Clear balanced and SE cables for all headphones
Many thanks to Cardas Audio for providing a full loom of Cardas Clear cables for this review!