UpTone Audio EtherREGEN Review and Comparison

UpTone Audio EtherREGEN Review and Comparison

Rajiv Arora

Introduction

The UpTone Audio EtherREGEN (shortened to eR for the rest of this review) is a diminutive Ethernet switch with a bold value proposition. UpTone claims it “...is capable of producing surprisingly audible sonic improvements in fine music systems.”  In this review, I’ll evaluate this claim of sonic improvements, and pit the eR against a couple of competing “audiophile” switches. But first, some background.

Background

For many, the very idea that an Ethernet switch can make a difference in sound quality (SQ) is anathema. “Where’s the proof,” they cry. “Show me the measurements.” At an even more fundamental level, their question is: “How could it make a difference?” An Ethernet switch’s job is simply to link Ethernet devices together by relaying Ethernet frames between the devices connected to the switch. As long as it does this without error, how could it affect SQ? After all, audio components that use Ethernet - servers, streamers, DACs, etc. - typically buffer the network data stream before further processing. As long as the data arrives correctly - bit-perfectly - how could SQ be affected?

As someone with both an academic and industry background in networking, I can sympathize with this skepticism. Not that many years ago, I too scoffed at the notion that any part of the chain of digital components upstream of a DAC - not just switches - could have any impact on SQ. It was only when I actually took the plunge and started listening to gear that I realized the upstream chain could profoundly affect - both positively or negatively - the SQ of the system. On the networking side, this  includes switches, routers, and even Ethernet cables, which I admit I found very disconcerting. I did my own experiments using OS monitoring tools to look for errors, retransmissions, or any other indications of functional misbehavior, and found none. All I could conclude was that we were hearing some orthogonal phenomenon - unrelated to the actual digital function of the device - which was not yet understood.

My first foray into the world of audiophile switches was almost three years ago. An Audiophile Style forum member, whose ears I trust, posted about the improvements he heard after getting his chain modded  by Korean manufacturer SOtM, to use their then-new sCLK-EX clock board. Based on his enthusiastic report, I took the plunge and had my gear modified as well. This was the famous SOtM trifecta - a modded switch, a modded sMS-200 endpoint, and a tX-USBultra USB regenerator. The clock board in the tX-USBultra regenerator provided the system (25MHz), USB (24MHz), and Ethernet (25MHz) clocks for the entire chain. I posted about my impressions here, here, and here. Quoting myself: While we expected the tX to have an impact, we were just astounded by the effect of the switch.

So while I have experienced the SQ impact of a switch, I do still wonder: how does a switch impact SQ? To be fair to the skeptics, manufacturers have not really answered this question. The first wave of products, like Paul Pang’s modded switches, the SOtM-modded switches, and The Linear Solution OCXO switch, were all consumer switches that had been modded to:

• Replace some components with better quality, such as capacitors and regulators
• Replace the ultra-cheap Ethernet clock (typically 25MHz) oscillator with a TCXO or OCXO oscillator, or modify the board to accept an external clock
• Allow operation with high-quality external linear power supplies.

All of these modifications were found to dramatically improve SQ to a varying degree, which was enough for many audiophiles like myself to adopt them, and for manufacturers to offer them. But in terms of an actual explanation of the mechanisms at work, manufacturers offered only vague statements about reduced noise generation, resistance to RFI/EMI, and lower clock phase noise.

Technology Background

To his credit, UpTone’s design engineer, John Swenson provides more than just generic correlation evidence to support the rationale and design of the eR. In a white paper entitled Understanding how perturbations on digital signals can affect sound quality without changing bits, and how these issues are addressed by the UpTone EtherREGEN, Swenson enumerates the mechanisms that degrade SQ, and how the eR addresses them. I look forward to the measurements data to back up the claims, which Swenson says he will publish soon. I encourage all of you who are interested to read the white paper at the link above.

The claims are certainly intriguing. While the phenomenon of leakage currents is something Swenson has written about extensively on this forum, what I found particularly intriguing was the claim that the high phase noise of a low-quality upstream clock induces ground plane noise on the downstream device’s receiver circuits. This ground plane noise in turn induces additional jitter (i.e. adds to the phase noise) on this device’s clock. He claims this effect is additive, so phase noise continues to degrade as data moves down the chain. The rubber hits the road when the receiving device is the DAC. This is where the accumulation of induced jitter has a sonic impact on SQ.

Is this proof? Certainly not until Swenson’s measurements are published, and even then I am sure there will be a robust debate about these theories! I find these claims intriguing because they do correlate with my empirical findings:

• The sonic effects of switches etc. are not related to their functional correctness.
• Sonic improvements accrue by replacing SMPSes with linear PSUs, and grow with increasing LPSU quality.
• Sonic improvements accrue by improving the system clocks of all upstream devices, and SQ improves as clock phase noise is lowered.

How does the eR address these SQ-sapping gremlins? Swenson describes a multi-pronged approach:

1. Use special high-speed digital isolators to block leakage currents,
2. Use fully differential circuitry to block upstream phase noise, and
3. Reclock outgoing data with an ultra-low jitter local clock.

For more details, I refer readers to the white paper.

Product Summary

Like any Ethernet switch, the eR has multiple ports, and provides connectivity between devices connected to these ports. The interesting aspects of the eR are:

• It has 4 x RJ-45 gigabit ports and one gigabit SFP (Small Form-factor Pluggable) cage (typically to accommodate optical transceivers) on the ‘A’ side of the unit
• On the ‘B’ side is a single RJ-45 100Mbps port
• The ‘A’ and ‘B’ side are separated by a so-called Active Differential Isolation Moat (ADIM™)
• UpTone claims the best SQ occurs when the DAC-attached machine and the upstream network are on opposite sides of the ADIM™
• No direction is preferred, and SQ benefits are claimed to be identical whether going ‘A’ to ‘B’ or vice versa.
• The eR supports an external 10MHz reference clock on its BNC input. When connected, there is a slider switch to select this mode of operation.
• External power supplies are supported in the range 7-12V DC, with current requirements of 7V/1.4A, 9V/1.0A, 12V/0.8A.
• Finally, a ground screw is provided.

Key Questions

I always find it useful - both for myself and the  reader - to lay out the key questions that my review is trying to answer. Proceeding from that, I then develop an evaluation plan, or testing strategy, to answer these questions. 

The key questions with the EtherREGEN are:

1. Does it work? Does its presence in the chain deliver an improvement in SQ?
2. How does SQ scale (improve) with power supply units (PSUs) of increasing quality?
3. Does SQ improve with an external 10MHz reference clock?
4. Is SQ better when crossing the ADIM™ moat: A <-> B?
   • How much of this is due to the B side being 100Mbps?
5. Does grounding the eR improve SQ?
6. Does fiber sound better than copper?
7. Does SQ vary with optical transceivers and do attenuators help?
8. How does the eR compare with other audiophile switches?
9. Is the moat impregnable? Are upstream or downstream improvements no longer necessary?
10. What happens when you go down the rabbithole of chaining switches?

Review System Topology

To aid the reader’s understanding of the various configurations tested in this review - which can quickly become confusing with all the possible combinations  - I’ve shown below a pictorial representation of the audio systems I used in my review. 

Primary System

As I described, when it comes to audiophile switches, this isn’t my first rodeo. In fact, my current system employs a chain of two such switches in series - the TLS (The Linear Solution) OCXO and the SOtM NH-10G switches.

I’d like to call the reader’s attention to the fact that my system uses an all-in-one music computer. This is a topology choice, just like others use separate computers to run the music server software and the endpoint/renderer. There are pros and cons around these topologies, which are not relevant to this review. The key point to note is that UpTone recommend the eR go immediately upstream of the DAC-attached machine. In my case, this is the music computer.

For this review of the eR, I held the rest of the topology constant, while varying the network connectivity in the grey box as shown.

Music Data Flow

In my system, music data resides on my NAS. This is a choice I’ve made. Many users will have music reside on a storage device (SSD, HDD etc.) on the music computer. My choice of NAS implies that during music playback, data flows across the switch, regardless of whether I am streaming local files, or from Qobuz or other streaming services.

Second system

I always like to test in multiple systems, so I also evaluated the eR in my friend M’s system. M recently sought my advice to adapt his home theater setup to enable high-quality 2-channel listening. He’s been on, and very happy with, the topology below for a few months, and was curious to hear how the eR would fare in his system. M doesn’t have any special power conditioning, and uses stock or entry-level cables across the system, except where shown. His music lives on his NAS, and he uses DLNA to stream from minimServer on the NAS directly to the DLNA renderer in his iFi Pro iDSD DAC.

Function and Operation

There isn’t much to say about the eR functionally. It’s an Ethernet switch. It works. 

Here are some operational considerations:

1. Heat: The case does run hot. In my setup, in a well-ventilated spot, with a single fiber input on the A side and the B side connected to my music computer, the unit was at 44ºC. This isn’t a problem, but adequate ventilation would be a consideration when placing the eR.
2. Stability: Due to its diminutive size, the eR can be lifted off its feet by the torque from heavy cables, as I found with the SOtM dCBL-Cat7 Ethernet and dCBL-BNC-75 clock cables I was using. To some extent, you can mitigate this by using the same cables on both the A and B side, or consider a weight of some kind that doesn’t impede heat dissipation.
3. 100Mbps: Assuming you cross the ADIM, as UpTone recommends, you’re limited to a throughput of 100Mbps. This isn’t an issue while streaming music, but it does increase the time for some operations. In my case, the DAC-attached machine is a music computer running Euphony OS (a variant of Linux) and the Stylus music playback software. Here are the operations where you notice the 100Mbps limit:
   • OS updates take significantly longer than with a Gb switch. Obviously, this is moot if your ISP plan is less than 100Mbps. Mine is 300Mbps, so I did notice. Of course, I could temporarily move the connection from the B to the A side of the switch, but that’s inconvenient.
   • Stylus has a “Buffer queue to RAM” feature that results in a notable SQ improvement. However, if buffering a whole album, especially at higher resolutions like DSD and DXD, this operation can consume several minutes at 100Mbps. Here again, this depends on your setup. If the music is on local storage, then this is not an issue. In my case, the music is stored on an external NAS, so music file I/O flows through the switch and incurs this limit.

With all that said, these operational considerations were minor and did not detract from the sonic performance of the eR. 

Review Playlist

EtherREGEN 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 some of the 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.

Listening Impressions

I’ve made it a habit to burn in review gear for a minimum of 200 hours before even attempting to evaluate its sound quality. In the case of the eR, I was very glad I did, because it most definitely needs burn-in. The bulk of the change happens in the first 24-48 hours, but my unit continued to improve for several more days.

Speaker-based Listening

Since my system is already tricked out with existing switches and high-quality power supplies, I’ll first describe the eR in M’s system, which is much more “typical!” We baselined our listening with no switch in the path - just a direct generic Ethernet cable from his router to his Pro iDSD DAC. Refer to the system topology shown earlier.

Bowspirit, from Austin instrumental ensemble Balmorhea’s excellent album Constellations (Western Vinyl, 16/44.1), is also the opening theme of the quirky indie show, Rectify. It’s a richly layered track with violins, banjo, acoustic guitar, and drums, that defies pigeonholing into a genre.

Step 1: Add eR with stock SMPS into the chain. Router on ‘A,’ DAC on ‘B’

Since I’ve heard this kind of effect before, I was watching for M’s reaction, and he didn’t disappoint. “What just happened,” he exclaimed! There was a big increase in the soundstage size, especially in the image depth. The violins and the banjo were more precisely positioned, and there was more texture and detail all around.

Step 2: Replace stock SMPS with Farad Super 3 LPS

This was another significant step up in SQ. Not only was the bass deeper, but it became much tighter and well-articulated. The wallops of the drum went from somewhat woolly thumps to emphatic thwacks. The violin and banjo lost their slight stridency and sounded more tonally refined. Both image size and fine details went up a couple of notches.

Step 3: Externally supply a 10MHz reference clock with the Mutec Ref-10 

M already had a Ref-10 in his system, providing a reference clock to his iFi Pro iDSD DAC. So it was a simple matter to use another clock output to reference-clock the eR. Despite the already impressive SQ gains from steps 1 and 2, this was yet another big uptick in SQ. This step was all about palpability and realism. The effect of the ultra-low phase-noise reference clock  is akin to swapping in a superior lens in a camera. Everything in the soundstage snaps into greater focus. Instruments take on a 3-dimensional realism, instrument placement in the image is even more precise.

All together now…

In a subsequent session, we swapped periodically between the baseline (no switch) and the full monty of switch+LPS+clock. 

This time we listened to a delightfully fresh interpretation of Carmina Burana, by Anima Eterna Brugge/Jos van Immerseel (Zig-Zag Territoires, 24/96), played with a smaller ensemble on 1930’s instruments. The difference between no switch and the full monty was both stark and startling. On In taberna quando sumus, the full monty switch configuration opened up the soundstage dramatically. The individual sections of the chorus were more precise, with more depth. Individual voices in the chorus were easier to distinguish. Brass instruments - trumpets, trombones, and tuba - in particular were much more realistic, with more texture and air volume.

M texted me later that evening to say he had placed an order for the eR.

Headphone Listening on my Primary System

Please note that the switch configurations shown in this and subsequent sections fit into my overall system topology (shown earlier) by replacing the grey box pictured on the left.

For the first round of tests, I listened to The Mandalorian Theme - Epic Version by L’Orchestra Cinematique (HM Chapel Music, 16/44.1). While I wasn’t at all impressed by the series, this epic theme makes for a great demo track. 

Since I was impatient, I compared the “no switch” configuration to the full monty. The two configurations are pictured below. 

   vs.   

Even in the “no switch” configuration, the system sounded great, with satisfyingly deep bass in the opening notes, and great air and dynamics on the triumphal trumpet theme. But with the full monty configuration, there was just so much more! The image size was now huge, and the bass was deeper and fuller. On the drum beats, you could effortlessly distinguish the different percussion instruments collectively being struck. Perhaps because my system is much more resolving, and because I was using an even better PSU and cables, this difference between no-switch and full-monty was much greater on my system than it had been on M’s system.

I observed a similar difference whilst listening to this excellent new release of Beethoven Symphonies 5 & 7 by Andrew Manze, NDR Radiophilharmonie (Pentatone, 24/96). On the second Allegretto movement of symphony no. 7, the ambience and tension, as this movement builds from a pianissimo exposition of the strings to the crescendo with the full orchestra, is much more palpable with the full-monty switch configuration. The soundstage is larger, the strings have more dimension and texture, and the tympani strokes are fuller, deeper, and again, more textured.

This was a very promising start. Even before the head-to-head comparisons, the eR seemed every bit as capable as my previous switches in delivering a big boost in SQ. 

I next explored how the eR scaled with PSUs and an external clock, and how SQ varied in different operational modes. My aim was to tune SQ to the best achievable, before getting into comparisons.

Comparison of PSUs

SInce I had already tested the stock SMPS in M’s system, I omitted it from this comparison. I gathered up all the PSUs I had available, and did a little shootout. The switch configuration was the same full-monty setup as before, as shown in the diagram on the left. I then varied the PSUs, with all of them set at 12V.

There were indeed big differences between PSUs. As the PSUs got better, the SQ improved along the expected lines: deeper tighter bass, better air and separation, bigger soundstage, and smoother tonality.

Cutting to the chase, here are the PSUs in increasing order of SQ. All of them are a big step up from the stock SMPS.

A. SOtM sPS-500 with 7N UPOCC silver DC cable

Even though technically an SMPS, the sPS-500 did sound significantly better than the stock SMPS, thanks presumably to the noise cancellation and filtration that is SOtM’s forte. The SQ was remarkably clean and smooth, with good clarity and resolution. However, it did sound smaller and more closed-in compared to the rest of the field.

B. Farad Super 3 with Audio Sensibility Signature Silver DC cable

The Farad delivered an even blacker background, with a more expansive soundstage. There was more density and texture, and it did wonders to the bass. This PSU really allowed the eR to shine, and showed you what it’s capable of.

C. Paul Hynes SR-4 with Paul’s DC3FSXLR silver cable

As I’ve found in other shootouts, the SR-4 and the Farad are very comparable in SQ. In this particular comparison, the SR-4 won out by a slight margin due to its slightly larger soundstage. Transients also seemed just a bit more well defined. Both the Farad and SR-4 are excellent companion PSUs to the eR!

D. Paul Hynes SR-7 SR (single regulation) with Paul’s DC6FSXLR silver cable

These last 2 PSUs fall into the category of overkill, since they are substantially more expensive than the eR, and less likely to be chosen as companions for the eR. Although certainly, for someone who already has a unit (like me) with an unused rail, this would be an excellent option.

First up was the “standard” Paul Hynes SR-7, which has just the single custom regulation stage. Good as the SR-4 and Farad were, the SR-7 SR raised the SQ of the eR even further. This is as much a compliment to the eR as it is the SR-7, as it takes two to tango, and a component’s design has to be of sufficient quality to even benefit from a PSU of this caliber.

With the SR7 SR, the bass got even tighter and deeper, and the music took on a pleasing increase in density and solidity.

E. Paul Hynes SR-7 DR (dual regulation) with Paul’s DC6FSXL silver cable

It never ceases to amaze me how much the addition of a second regulation stage to the  already-excellent SR-7 improves SQ. Yet it does, and the eR rewards you for providing it. The SR-7 DR (dual regulation) rail seemed to reveal another layer of hidden detail, while further enhancing the bass, articulation, and soundstage. If there is a better external PSU than the SR-7 DR, I dearly want to hear it in my system.

As it stands, the eR improved with every stepwise increase in PSU quality all the way to the best in my stable. Very impressive indeed.

Effect of Ref-10 reference clock

Since I’ve already described the effect on M’s system, I’ll only add that I observed an even bigger jump in mine. I described it as a lens snapping into focus, and the same analogy held here. On the Beethoven 7th, in the pianissimo exposition, the individual violas and cellos in the string sections became discernible, and in the crescendo, when the wind instruments kick in, here again, it was much easier to distinguish the french horn from the trumpets.

SQ difference due to ADIM™

Was there an SQ difference if the music computer in my system was connected to another port on the A side, vs. connected to the B side, where data would cross the ADIM™ isolation moat? Yes there was. While the A side connection still gave a sonic benefit, there was an additional boost when crossing the moat. This SQ boost manifested as an increased calmness, a lowering of fatigue, and a smoother tonality, while retaining the clarity and resolution.

But could this just be the difference between 100Mbps and 1Gbps speeds? To test this, I used the Euphony OS advanced settings on my music computer to force the JCAT Net Card port to a speed of 100Mbps. Would I achieve the same improvement in this case, by using the A side of eR, but forcing the port down to 100Mbps? In a word, no. I heard very little difference when the port was connected to the A side, whether it was auto negotiating to 1Gbps, or when speed was set to 100Mbps. 

The sonic benefit I was hearing was indeed a result of crossing the A <-> B boundary, through the isolation moat. For the rest of my tests, I used the B side to connect to my music computer.

Effect of grounding

UpTone’s claim is that attaching a ground wire from the eR grounding post to an AC ground will remove “high source-impedance” leakage current from the SMPSes powering upstream devices connected to the network, by shunting these currents to the AC ground. This can raise SQ, although to what extent is highly system-dependent.

In my network, I had already cleansed the path from my ISP ingress to my audio system of SMPSes. My upstream cable modem, router and switches were all running on LPSes. Perhaps for this reason, I heard no benefit from grounding the switch. I left the ground wire connected in any case.

Effect of music location relative to switch 

As I noted earlier, my local files reside on my NAS, so they traverse the switch during playback. However, the Stylus music player I use has an optimization where music can be cached on the boot drive prior to playback. Indeed, as a further optimization, the tracks in the queue can be buffered into memory prior to playback. A similar situation occurs in the case where a user’s music files reside on local storage downstream of the switch.

In his white paper, Swenson notes: 

“A very large buffer where the input completely shuts while all music is playing can eliminate the phase-noise overlay of upstream sources.” 

Does this imply that the switch has less of an impact when the music files, either due to storage location, or due to the effect of pre-caching, are not flowing through the switch during playback? 

To test this, I ran two player software configurations on my standalone music computer:

1. Roon Server: Roon Core consumes track data from the storage device in a flow-controlled manner during playback. This represented the case where music data from my NAS was flowing through the switch during playback.
2. Stylus: I configured Stylus with both cache and “buffer queue to RAM” optimizations. This represented the case where no music data was flowing through the switch during playback.

For each configuration, I compared the “no switch” with the “full monty” configuration. Result: both configurations benefited greatly from the full monty switch in the path. Even with no music data flowing in the buffered case, the switch still makes a difference. Moreover, the magnitude of the difference wasn’t significantly greater in the Roon Server case.

This finding - of the switch making a positive impact on SQ even when no music data is flowing through it - also held when I compared the eR with its competitor switches. 

Is the eR a perfect moat? 

Any time a device claims to isolate with words like moat, it gets our hopes up that perhaps this device renders further optimizations upstream (or downstream) moot. To test this, I examined both sides of the eR.

Effect of further downstream optimization

If the eR upstream of the DAC-attached machine is a perfect moat, then is the reclocking JCAT Net Card Femto, with external SR-4 PSU, still necessary on my DAC-attached music computer, compared to a standard Ethernet port on the motherboard?

I think you can guess the answer! Yes, it still makes a difference, even with the eR upstream of the machine. For all the good the eR does to SQ, it is not a perfect barrier. In my experience so far, no such device has provided perfect isolation. This does not detract from the value and benefit of the eR, but it just means it doesn’t remove the need for end-to-end optimization.

Effect of further upstream optimization

So if the eR is not a perfect moat, can SQ be further improved by adding an additional optimization upstream of the eR? To test this, I considered 2 scenarios, as shown below:

In one case, I introduced my existing TLS OCXO switch upstream of the eR, connected to the eR via the usual copper Ethernet cable connection. In the second case, I introduced a Sonore opticalModule (oM) FMC (fiber media converter), connected to the eR via a fiber optic cable. The latter path has the advantage of galvanic isolation due to the use of fiber. Additionally, the oM is an audio grade FMC, using ultra low noise linear regulators, and a high-quality ultra low jitter FEMTO oscillator. The use of the oM also required the use of fiber transceivers and a fiber optic cable. For this test, I used the transceivers supplied by Sonore, which were TP-Link TL-SM311LM 850nm/550m transceivers.

Both configurations further improved SQ over just the eR in isolation! I know this is the kind of improvement that, while appreciated, also induces a groan and a shake of the head. Where does this end?! Hold that thought - I’ll come back to that in a later section.

Which configuration sounded better? While you might expect the air-gapped fiber path would hold the advantage, in fact these two configurations were very similar in terms of the additional SQ they added to the eR. The TLS OCXO switch configuration was stronger in enhancing the size and depth of the soundstage, while the oM configuration added more density, but was a little more closed in. 

At this stage, I became aware of experiments being reported on several forums with different transceivers. Beware! This is another rabbit hole of endless experimentation, but I did dip my toe in the pool. After some experiments, I settled on a pair of the Startech SFP1000ZXST 1550nm/80km transceivers, with a 2m cable. Since these transceivers are designed for long  distances (note the 80km rating), I used 10dB attenuators, which also improved SQ. This combination was notably better sounding than the TP-Link transceivers. With this configuration, the oM leapfrogged the TLS OCXO switch. The oM added even more dimensionality, more meat on the bone, while also opening up the soundstage.

I don’t want to suggest this kind of chaining is essential. The majority of the SQ gain comes from the eR upstream of the DAC-attached machine. The addition of the oM is like the icing on the cake. 

Ethernet cables

One axis I did not explore was to vary Ethernet cables. Since I’ve been experimenting with switches for some years now, I have tried many different cables in the past. Yes, cables do sound different in my system. Some time ago, I had picked the SOtM dCBL-Cat7 cables as the best fit for my system, for their rich and balanced character. For this review, I held these cables constant in all the comparisons.

Comparisons with other audiophile switches

I consider this section - the comparison with competitor products - to be the most important part of my reviews. How a component sounds in and of itself is interesting at best, but how it sounds relative to its competition is one of the most important pieces of information a potential buyer wants to have.

I already had on hand my own units of the TLS OCXO switch and the SOtM sHN-10G. Full disclosure - at my request, @The Computer Audiophile requested review samples of the Aqvox SE and the Melco S100 switches. Neither were provided. Luckily, as is often the case, the community here came to my aid. A kind AS’er from Hong Kong loaned me his personal unit of the S100.

I limited my comparisons to the SOtM and Melco switches. I didn’t bother with head-to-head comparisons with the TLS OCXO for a couple of reasons. First, there have been disturbing reports of quality issues and dubious business ethics by this company reported here on this forum. More tellingly, in my previous testing, the TLS OCXO switch was no sonic match for the sNH-10G.

Without further ado, let’s see how the eR fared against these switches. 

Comparison with the SOtM sNH-10G

The SOtM sNH-10G (US MSRP $1700)  has been the primary switch in my system for some time now. It is a product SOtM designed from the ground up with audio quality in mind. In addition to Evox capacitors and linear regulators, the sNH-10G features their low phase noise, sCLK-EX clock board that can optionally be driven by an external 10MHz reference clock. This switch has 8 copper RJ-45 ports, and 2 SFP cages, all of which can operate at 1 Gbps.

 I compared the following configurations:

        vs.            

For this comparison, I took a trip down memory lane and resurrected an album, Birds of Fire (Columbia - Legacy, 16/44,1) by one of my favorite bands, the Mahavishnu Orchestra. The track Open Country Joy is a joyous jam session between John McLaughlin, Jerry Goodman, Jan Hammer, Billy Cobham, and Rick Laird. Yeah - how could that not be amazing! This is by no means an audiophile recording, but it still revealed important differences. 

Using the sNH-10G as my baseline, the eR sounded a bit cooler, which I felt was more tonally correct, while delivering  very similar soundstage size and resolution. At 1:11 in this track, when the violin, electric guitar, and drums kick in with a crashing crescendo, the SOtM came across a bit hot in the treble, while the eR sounded less fatiguing. The SOtM’s thinner sound contrasted with the eR’s more fuller signature.

Across a wide range of tracks, I found these two switches had distinct signatures, but competed very closely with each other. In many cases, the sNH-10G had slightly better resolution, while the eR won on tonality and density.

Overall, I preferred the SQ with the eR by a small margin. Once you factor in cost though, the eR won the price/performance battle handily. Given that the sNH-10G has been my reference switch for some time, this was an impressive showing by the eR.

Comparison with the Melco/Dela S100

The Melco S100 is a recently-released switch from Melco Syncrets Inc., and is sold under the brand names Melco in the UK and Europe (UK MSRP £1999) and Dela in Japan. It does not appear to be available in the US yet. The evaluation unit I was loaned was a Dela S100.

Melco do not reveal much about what aspects of their design make this an audiophile switch. The 2-page datasheet claims that “Audiophile techniques are used in the power supply implementation with a bank of audio-grade capacitors…” but there are  no other details, including the quality of the clock. Effectively, this is just a black box with some claims of audiophile sound quality.

The S100 has 2 banks of 4 RJ-45 copper ports. One bank operates at 1Gbps, and the other at 100Mbps. Additionally, there are 2 SFP cages for fiber transceivers. Melco claims that music streamers and players sound best when connected to the 100Mbps ports. In my listening tests, I did find this to be the case, although the SQ difference between 100Mbps and 1Gbps was not as large as the difference observed with the eR. Of course, in the eR’s case, data was also crossing the isolation moat. For this comparison, I used the 100Mbps on the S100, to compare the best possible sounding configurations of each switch.

        vs.            

My preferences lean toward orchestral and symphonic music, so when I purchased this album, José, Albéniz & Manén: Guitar Works, Vojin Kocić (Naxos, 24/96)  as an impulse buy during a Christmas sale, it came as a bit of a surprise to find how much I liked it. I used the Cuba movement of Albéniz’s Suite española No. 1 to compare the two switch configurations.

By this time, I had had the eR in my system for a couple of months, so I used it as my baseline configuration. I didn’t quite know what to expect of the Melco and had low expectations, so I was rather pleasantly surprised to find how good it sounded. Similar to the eR, it had a smooth, non-fatiguing tonal character. Soundstage size and resolution were also on par with the eR. The S100 surpassed the eR by a modest amount in its transient response of the guitar plucks, and the guitar sounds more solid and 3-dimensional. 

Listening on other pieces, this advantage persisted, with the S100 portraying quicker transients with more density and saturation of the soundstage. This difference, while notable, was not huge. Still, in terms of absolute sonic performance, I would give the nod to the S100.

How would I describe the choice between these 2 switches? If absolute performance is your goal, and cost is not a big consideration, then the S100 is an attractive option. However, the eR allows you to achieve similar SQ with a more incremental approach. Start with the $640 base switch, add an LPSU when funds permit, consider a reference clock down the road, perhaps even a clock upgrade after that. Finally, many users will value the information and transparency that UpTone provides, versus the black-box Melco approach. As with everything, it’s a choice, with no right or wrong answer.

Chaining Experiments (down the rabbit hole)

Finally, it was time to get into mad-scientist mode, and go chasing down the rabbit hole. Yes, I’m talking about chaining switches. I talked about this briefly in an earlier section, when I described the SQ increase by preceding the eR with the Sonore oM or the TLS OCXO switch. That experiment revealed that the eR was still amenable to upstream optimizations.

Let me first preface this entire discussion to say - none of this is necessary, and I would hazard a guess that 90+% of eR users would never contemplate this! So read this section for vicarious entertainment, unless you are looking for the ultimate SQ and have the deep pockets to afford it.

The premise here is simple.

• If one eR improves SQ as we found, would 2 in a chain improve further? 
• And by how much? 
• What about a chain of 3 eR switches?

Well, I can’t answer these questions, since Alex didn’t supply me with 3 eR’s - it was hard enough to pry one review sample out of his hands! However, I did have 2 other switches of similar quality on hand, so why not use those.

My musical choice for this section was this new release of Beethoven Piano Concertos Nos. 2 & 5, Kristian Bezuidenhout, Pablo Heras-Casado, Freiburger Barockorchester (Harmonia Mundi, 24/96) came online on Qobuz as I was starting this experiment. Do we need yet another recording of the Emperor Concerto? Yes, if it’s this good! Bezuidenhout plays on a 1824 Conrad Graf fortepiano, which has some of the woody tone of early pianos, as well as the grander dynamics of the modern instrument.

Effect of a 2nd switch

I compared the single switch eR with a chain of SoTM > eR, as shown below.

        vs.            

Sure enough, the 2-switch configuration sounded even better. What I heard was an amalgam of the best attributes of the sNH-10G and the eR. The excellent resolution and detail of the SOtM switch was evident, but teamed with the density and tonal richness of the eR. 

Just for grins, I flipped the order of the switches, and was surprised to find it sounded … different. Mostly, there was a more forward, thinner sound I didn’t quite like as much. I found I much preferred the SoTM > eR chain. It appeared the sonic signature of the last switch in the chain was the dominant one, which is perhaps not surprising.

Copper vs. Fiber

      vs.     

Since I had had a good experience with the  Startech SFP1000ZXST transceivers, with 10dB attenuators, I was curious to see how this fiber optic path would fare in the 2 switch configuration. The answer is - very well! Comparing the copper path to the fiber path, I found I greatly preferred the fiber path, and indeed this difference was even greater than what I had heard comparing the oM and the TLS OCXO switch. Note that in this deployment, everything else was constant, including PSUs and clock, so the only change was between copper or fiber between the switches. The fiber path had cleaner transients, and sounded more coherent. And with the Startech transceivers, there was more density and saturation to the soundstage. I could have gone down a secondary rabbithole of trying more transceivers, but decided to leave that to others! As I finish this review, there appears to be another promising prospect from Planet Technology being reported on the forums, which I will try in due course.

So did the second switch deliver the same value for money? Well, no. The bulk of the improvement comes from the first switch. The second switch is definitely a step up, but the incremental gain is a fraction of the first. To those looking for absolute performance, this is very worthwhile, while those looking for value will likely not consider this a good return on investment.

Effect of a 3rd switch

    vs.    

Why not go for broke and chain all 3 switches? Why not indeed! Fortunately, I had the spare PSU rails to try this.

What can I say? Incredibly, the 3rd switch does in fact add a small dollop of additional SQ. Do I recommend anyone do this? Not really, although some fanatics might. For me, it really comes down to value for money. With a 2-switch chain in place, wouldn’t the cost of the 3rd switch and a PSU (rail) be better invested elsewhere in the system for a bigger sonic impact? I’d certainly want to explore that.

Reflections

If you’re exhausted at this point, consider my predicament! Luckily, I enjoy experimentation, so this was fun for me. Still, I had to be disciplined and follow a written plan to make sure I covered as much as I did. And with all that, there are so many configurations I couldn’t test.

In this section, I want to put all the findings in this review into some context, as key take-aways. One of the dangers of describing SQ improvements in relative terms is that it is easy to lose sight of the actual magnitude of the improvements. So here are some reflections, based on my own subjective experiences:

1. The SQ gains from Ethernet improvements in aggregate are significant, and comparable to USB improvements (regenerators, cables, etc). In my system, the total magnitude of USB optimizations has tended to yield greater aggregate SQ, but I consider both areas of network and USB to be fertile ground.
2. The SQ improvement from the combination of switch, PSU, and clock is comparable to making a component upgrade (for example, a DAC or amp) of similar price. I know this a provocative statement, and may not always hold, but it highlights my own experience that upstream improvements deserve appropriate investment, and deliver similar value
3. SQ improvements from chaining switches are additive, but asymptotic. By this I mean additional switch(es) provide diminishing returns, so consider value for money carefully before taking this approach.
4. SQ improvements in the chain, especially in the network path, are very system dependent. Take my findings as data points, but always verify for yourself if the improvements manifest in your system, and if what you hear is worth the cost. This is where the satisfaction guarantee of products, like the 30-day money-back policy for the eR, is incredibly important. 

Summary

The UpTone EtherREGEN is one of those rare products that delivers high-end audio performance at an affordable price. Not only does it deliver “surprisingly audible sonic improvements” as claimed, it holds its own against the best-sounding competing products out there, that cost 3x its price or more.

Furthermore, unlike any of its competitors, UpTone have attempted to demystify the technical basis for the product, and are to be commended for that. Whether or not it convinces the skeptics, I for one applaud their transparency in the matter.

For a modest investment, users can start with the sonic benefits out of the box, and extend these benefits further over time with upgrades to the PSU and clock. With this stepwise approach, users can achieve similar absolute performance as the best of the EtherREGEN’s competition.

 I don’t use the word “bargain” lightly, but in this case it’s true. The EtherREGEN is one of high-end audio’s bargains. If your audio system uses Ethernet for streaming audio, you owe it to yourself to try the EtherREGEN.

Product Information:

• Product:  UpTone Audio EtherREGEN
• Price: US MSRP $640
• Documentation:  EtherREGEN User Guide

Associated Equipment

 Primary System

• Music Computer:          Custom computer: H370M-ITX/ac, i7-8700T, 8GB Apacer RAM, HDPlex H3 case,  32GB M10 Optane SSD for                                              OS, JCAT Net Card Femto, running Euphony OS with Stylus or Roon+StylusEP music software
• Music Storage:               Synology NAS DS916+ 4-bay, attached to router via Ethernet
• Headphone Amplifier:    Cavalli Liquid Gold
• Headphones:                 Sennheiser HD800 (SD Mod), Meze Empyrean, Abyss AB-1266 CC
• USB Regenerator:         SOtM tX-USBultra SE
• Reference Clock:           Mutec Ref 10 10MHz clock driving the tX-USBultra and switch
• Power supplies:             Paul Hynes SR-7 DR (dual regulation)  for switch & tX-USBultra, Paul Hynes SR-4 for JCAT Net Card                                                             Femto, HDPlex 400W ATX LPS for music computer motherboard
• Power Details:               Dedicated 30A 6 AWG AC circuit, PS Audio P12 PerfectWave Regenerator (under review)
• Power Cables:               PS Audio AC-12 (wall to P12), Cardas Clear Beyond (Cavalli Amp), Cardas Clear for all other components
• USB cables:                   Intona Ultimate 0.5m, Intona Reference 0.5m
• Clock cables:                 Habst 5N Cryo Pure Silver 50Ω and SOtM dCBL-BNC 75Ω
• Ethernet cables:            SOtM dCBL-Cat7
• DC cables:                     Ghent Audio custom OCC JSSG360 ATX and EPS cables, Paul Hynes fine silver (SR-4, SR-7)
• Interconnects:               Cardas Clear XLR balanced
• Headphone cables:       Transparent Ultra cable system, Cardas Clear balanced and SE cables
• Accessories:                  Synergistic Research Tranquility Base XL UEF with Galileo MPC

Acknowledgments

Many thanks to the following companies for supplying cables and accessories to aid in this evaluation: 

• Cardas Audio, for a full loom of Cardas Clear cables.
• Transparent Audio, for the Transparent Ultra headphone cable with a full complement of headphones leads and source terminators.
• Intona EU, for the Reference and Ultimate USB cables.
• Sonore Audio, for the opticalModule, TP-Link fiber transceivers, and fiber optic cable.

About the Author

Rajiv Arora - a.k.a. @austinpop - is both a computer geek and a lifelong audiophile. He doesn’t work much, but when he does, it’s as a consultant in the computer industry. Having retired from a corporate career as a researcher, technologist and executive, he now combines his passion for music and audio gear with his computer skills and his love of writing to author reviews and articles about high-end audio.

He  has "a special set of skills" that help him bring technical perspective to the audio hobby. No, they do not involve kicking criminal ass in exotic foreign locales! Starting with his Ph.D. research on computer networks, and extending over his professional career, his area of expertise is the performance and scalability of distributed computing systems. Tuning and optimization are in his blood. He is guided by the scientific method and robust experimental design. That said, he trusts his ears, and how a system or component sounds is always the final determinant in his findings. He does not need every audio effect to be measurable, as long as it is consistently audible.

Finally, he believes in integrity, honesty, civility and community, and this is what he strives to bring to every interaction, both as an author and as a forum contributor.