I looked at the Antipodes DX - it was quite expensive but what ruled it out for me was that it did not have an Ethernet output. This means that you are stuck with the USB output and, if some modern tech end point comes along at a later point which gives better performance, you are stuck. The Zenith permits connection to an end point without a round trip through your network. I found the network round trip to have a negative impact when trying to match my file playback with the Blu CD sound quality so, for future flexibility, a server with two Ethernet ports was essential for me. Surprisingly, there aren't many around. The Melco has it - I have one - but it cannot operate as Roon core and I had concluded that I had to have a Roon core wired directly into BluDave with nothing in between. The Zenith was the only one that I found that operates as a Roon core and has twin Ethernet ports and direct USB out. Fortunately, it sounds ok! Actually, I should add that the build of the server is also significant and that the Zenith Mk II sounded a little better than the Melco whilst the SE is quite a step up again, so it's not just the network and I didn't mean to imply that.
Direct USB out from the Zenith easily betters the Ethernet out to my mRendu into BluDave by USB, but that may not always be the case with the pace of development and I think some degree of future proofing is essential when contemplating not insignigficant sums of money. I think we all hate it when our shiny new device is bettered by some much cheaper new kid on the block.
If the Zenith was comparable to the Antipodes DX in sound quality, that's saying a lot and it gives me some point of reference I can relate to as I hold the DX in very high regard. I have heard the Melco also and while SQ was very good (also better than the microRendu), I agree, lack of Roon was a deal breaker, especially since Melco's interface options are so archaic in comparison.
You are correct, technology in the area of music servers is advancing quickly and rapidly becoming less expensive which is good for all consumers. There is no longer any justification for buying something like a $17k Aurender W20 as I long ago surpassed the SQ I was getting from a W20 at a fraction of the cost. Having opened up and taken apart several of these well-regarded servers and as I have assessed the individual impact of various components, I have drawn the conclusion that it boils down to 3 things: low noise, low latency and low impedance -- but this is easier said than done.
Perhaps the foundation of it all is the power supply and I have yet to find a power supply that can do what Paul Hynes' SR7 can do, better than even ultra-capacitor based PSUs that I have on hand. For example, the Zenith appears to use a well-implemented linear PSU based on a single transformer that then feeds 3 rails. Noise output from its regulators are claimed to be 40uV which is very good as this is nearly 25x less noisy than the LT1083 regulators (about 1,000-2000uV) used by the popular HDPlex LPSUs. However, Paul Hynes' regulator circuits, which are of his own design, have noise levels in the 4uV range, about 10x quieter than the Zenith. Is this audible? Yes, very much so.
Perhaps even more important is output impedance which is, in a simplistic sense, the agility of a power supply to be able to respond rapidly to current requirements that are important not just for macrodynamics but also for microdynamics. It allows for the ability to glean dynamic shadings and subtle nuances within a passage. Notes start and stop more cleanly. Very few PSU makers report this value either because they are unaware that this is important, because their supplies have very high impedance and so they prefer not to advertise it, or because they don't have the equipment to measure it. It would be ideal to have an impedance of zero ohms but the reality is that all electronics have some impedance. A well regarded switching PSU I have on hand has a measured output impedance of about 50 milliohms and so I will assume that some of the finer LPSUs out there will be better than this even though most have no measurements to report (including Zenith). The large Vinnie Rossi ultracapacitor-based PSU has a reported output impedance of about 16 milliohms and indeed, that is a wonderful sounding PSU. The small LPS-1 by Uptone Audio is probably a little better than this but they are unable to report values because they don't own measuring equipment. Paul Hynes' SR7 has an output impedance of <3 milliohms from DC all the way to 100kHz. I have yet to hear anything that can do what the SR7 can do and so not surprisingly, I consider the SR7 as the foundation to any server I build. Should you decide to commission Paul to build a PSU to replace the one in your Zenith, I'm fairly certain you will notice an improvement that will not be subtle.
As far as the server hardware itself, as I see it, it really has 2 functions with regards to audio playback. The first is the more mundane task of music storage, library management and in some cases, DSP. If you are a non-Chord DAC user and have bought into the popular trend of upsampling to DSD via HQP, then this is one more task that falls into this category. These functions require heavy lifting, especially with large music collections, and, therefore, the utilization of noisy components such as a powerful CPU, large amounts of RAM (that consume up to 4A in burst), and noisy hard drives or SSDs (SSDs are not noisy acoustically but noisy in terms of adding HF noise to the signal). As an aside, I find the modern SSDs to be considerably noisier and to the detriment of SQ than a spinning hard drive and should be avoided, imo. Careful A/B testing easily bears this out, especially to my ears which are very sensitive to HF noise and is perhaps partially responsible for the "digititus" that many analog lovers describe. The second function of a server is the rendering of the file so that it can be transmitted to your DAC and this is what is so crucial. Rendering a PCM file requires very little CPU power and very little RAM. In other words, rendering requires no heavy lifting at all.
When you have a single box that serves as both a server and a renderer (i.e. a basic PC or Mac), you are basically unnecessarily using more CPU and RAM than is necessary to render. Along with the noise generated by the CPU and RAM is the noise generated by the hard drives or SSDs and noise generated by all the switching regulators and noisy clocks on the motherboard. Does this reach DAVE, even with its well-implemented galvanic isolation? Yes, my ears clearly tell me it does. I believe the benefits heard by applying ferrite filters to cables proves very well DAVE's galvanic isolation is not 100 percent, and it's not just noise in the ground plane that's the issue but more importantly, noise that gets permanently imbedded into the signal that can't be removed because at some point, I believe, this noise becomes part of the signal. It's like making a photocopy of a photocopy. You keep doing this and eventually, you have no idea what the original should look like. I say this because even using optical via DAVE which is completely impervious to RF in the ground plane, I can still easily hear the impact of noise reducing changes that I implement in the server.
Now, there are several reasons small renderers like the microRendu, ultraRendu or SOtM's sMS-200 have the potential to sound better than large box servers. First, these small renderers are now isolated from the noisy server via a connection (Ethernet) that is inherently galvanically isolated (although this isolation is not perfect either). Second, as previously stated, because rendering requires no heavy lifting at all, these small renderers utilize hardware that require very little current for operation. As such, it is considerably easier and less expensive to build a low noise, low impedance PSU for these devices that draw less than 1A of current. Third, because these devices are small, they have very short signal paths resulting in very low latency but also low impedance. Finally, unlike standard PC motherboards you can buy, these small devices use custom motherboards that are largely devoid of noisy switching regulators and also utilize better clocks with low phase noise characteristics.
There are single box devices (such as the Antipodes and your Zenith) that can sound considerably better than a PC or Mac because what they have done is they have avoided noisy hardware. I can't speak for what is in the Zenith but with the Antipodes DX, I have opened up this machine and found that they are using a small mini-ITX motherboard with an embedded Celeron CPU that draws only 8-10 watts peak. Mark Jenkins specifically found that simple CPUs like a Celeron sound better than more complex CPUs like the i3/i5/i7. My own testing seems to support this. Furthermore, these devices use only as little RAM as necessary. I have found, for example, that 2GB of RAM sounds better than 4GB and considerably better than 8 or 16GB. 16GB of RAM can draw as much as 4A in bursts! The Antipodes uses an SSD for its OS. Interestingly, I found that the OS drive has a greater impact on SQ than the drive that stores your music collection (assuming you use a different drive for each). The probable explanation for this is that an OS drive is constantly churning (and drawing current) while a storage drive becomes idle once the music file has been buffered into RAM. While a hard drive sounds better than SSD, it's not good to have a spinning drive in your server due to the vibrations that it creates and so I have avoided all internal hard drives for this reason. What I found to sound best is compact flash and so I have 2TB of compact flash that I am using for storage but to use a compact flash drive for OS duty presents many challenges. What I found to be the best solution as an OS drive is to use an older SATA II SSD that utilizes SLC memory and the smaller capacity, the better the SQ. These drives consume only about 100mA of current, considerably less than SATA III SSDs which can consume as much as 1.2A in bursts. They also generate much less HF noise (which is in the 6GHz range). I was able to find an NOS 64GB Intel X25E SLC SSD on Ebay for only $80 and SLC SSDs will likely outlive many of us because they have very long life cycles. Coupled with a well-insulated OCC copper SATA cable by Pachanko cables (here we go with cables again) and an SOtM SATA filter, I have been able to get this SSD to sound better than a spinning hard drive and equivalent to a compact flash drive. Moving on, the Antipodes DX also utilizes SOtM's very well regarded tX-USBhubIN which is a specialized USB output card that re-renders the original signal to an even cleaner one as it powered independently by a separate rail from its PSU, incorporates a low noise clock and is almost completely devoid of any noisy switching regulators. No matter how clean the original signal, I have found that re-rendering it with an Iso Regen or a tX-USBultra results in even better SQ and so in my setup, I am using both.
Regarding the OS, this has everything to do with latency resulting in software errors that definitely impact SQ. Inherently, I find MacOS to sound better than Windows 10 with Linux sounding best of all, however, with optimization using such tools like Audiophile Optimizer, Windows Server 2016 is even better yet. The advantage of Windows is you get to use Chord's ASIO driver which has functional advantages over the drivers used by MacOS and Linux. Furthermore, Windows has the broadest compatibility with regards to software players beyond Roon. Spotify HiFi (lossless) and Pandora lossless are reportedly in the works. Who knows when they get released to Linux but for sure, you will see them in Windows immediately. The advantage of Roon beyond SQ is that you never have to see your OS. You can control everything via Roon's GUI with any tablet or smart phone.
While the above sounds overly complicated, if you know what to do, it's very easy to do this yourself and it doesn't have to be expensive, depending on how far you wish to take it. For example, we are talking about a motherboard that costs less than $200 and RAM that sells for $30. As previously stated, my OS drive cost $80. Compact flash isn't overly expensive these days. You can stick with your NAS but unless you replace the clocks on your router and power it cleanly, it won't sound as good as direct storage is what I have found. For those that wish to avoid the aggravation, then there are such turnkey devices like the Antipodes DX (or Zenith), however, having heard the DX, I don't believe its PSU is in the same class as Paul Hynes' SR7 because even without full replacement of clocks, my server sounds better.
What is truly icing on the cake is the replacement of the noisy clocks within the server but also within the renderer and thus far, I don't know of any one who has been insane enough to do what I have done to the extent that I have done it but after hearing the results, I believe it can serve as a blueprint for perhaps the ultimate server of the future. With clock replacement, this is not a jitter issue but a noise issue. These stock clocks don't provide timing to the signal (ie 44 or 48kHz), they provide timing to allow proper functioning of certain components and are very noisily powered. This noise is directly imparted upon the signal and as I have gradually replaced these clocks, I am astounded by how much more open and airy the soundstage becomes and how details are cleaner and easier to glean as if veils upon veils have been removed but perhaps what is unique is this "buttery smoothness" that I have previously used as a descriptor. I simply have never heard smoothness (without compromise of detail) like this before, even from Blu2, and I can assure you, DAVE reveals these changes brilliantly. What I previously thought were bad recordings that were unlistenable for more than a few seconds suddenly have become very palatable recordings.
For those looking to build something on their own, the following are my key ingredients to success:
1. Paul Hynes SR7. It is unlikely this PSU will be surpassed by another PSU anytime soon. For me, this is a statement PSU that is as end game as any component I am aware of. As many know, Paul is based in Scotland. I am impressed by how many of the very best products that I admire seem to be coming from the UK these days. (MQA might be the exception, lol).
2. SOtM sCLK-EX. This is a clock board that can be used to replace any clock on any device. This single board can be used to replace up to 4 clocks. As an example, most routers utilize 2 clocks, a network switch utilizes 1 clock, a motherboard utilizes a single system clock that then serves as a reference for many subclocks (DPLL). Most input and output cards (USB, Ethernet) utilize 1 clock. Most endpoints (Iso Regen, microRendu, ultraRendu, tX-USBultra, sMS-200, etc) utilize 1 or 2 clocks. In my particular chain, I replaced 8 clocks. Each sCLK-EX board costs about $1100 with 4 clocks activated and so I own 2 of these boards.
3. Mutec REF10 master clock. This master clock generator is used to synchronize all the clocks that have been replaced with the sCLK-EX to its own 10MHz OCXO clock. This is reportedly the finest OCXO clock there is today with respect to phase noise (which is the most important quality for audio) and easily surpasses the phase noise of atomic clocks (rubidium). It is not likely to be bettered anytime soon. At about 3,200 Euros, this device is expensive but considering what it has the potential of doing (ie elevating your server to "finest in the world" status), I consider it a bargain.
My current server is undergoing final modifications (hopefully, the last) and I should receive it back soon from Korea. Once I receive it, I will report back on how it impacts BluDAVE but having just heard the impact of swapping an inexpensive Pangea USB cable with SOtM's latest USB cable with my Blu2, I have no doubt it will add to what BluDAVE provides.
