Hi guys, curious which input you guys prefer to use on your upsampler.
Tonight I spent some time comparing the network input via Mosaic vs USB to my PC vs an optimized USB -> isolator/reclocker-> spdif input using tidal.
Surprisingly I’m pretty sure the spdif was the best of the three. The direct USB was using a cheap standard USB cable and the network was using standard cat6 but I was surprised that Mosaic didn’t win.
Has anyone else had a similar experience?
IMHO Ethernet.
Ditto from NorCal.
Likewise, Mosaic/Ethernet, by a proverbial mile 
. No clocking issues to worry about as compared to an independent S/PDIF input.
Have you guys done anything special with your ethernet? I’m a bit skeptical that audiophile switches/LAN cables will make much difference, but interested to hear your experiences.
I spent a fair bit of effort optimizing that SPDIF path for my previous DAC (an Esoteric K-01) and it was clearly superior to the USB input on that device, so I’m not surprised it outperforms the direct USB input to the Vivaldi (currently using a free USB cable I found in a box).
My SPDIF does go through a D to D converter that isolates it from the USB input and reclocks the signal with a decent OCXO clock. It’s also using excellent USB and SPDIF cables so the jitter performance should be quite good.
This path basically leaves the Vivaldi Clock by the roadside. An SPDIF input syncs to the clock of the SPDIF Signal.
Ethernet and USB signal streams are “pulled” in sync with the clock of the receiving device.
I have a Rossini, not a Vivaldi. The Rossini also sounds best (by a mile) with Ethernet. I have been experimenting with a Silent Angel Bonn N8 Ethernet switch and I hear a clear difference. Also betwen different power supplies for the Bonn N8. Will try Etherregen and Meicord Ethernet cables later this month.
See here for Etherregen review
Oh, one thing I forgot to mention is that I don’t have a Vivaldi clock! I’m using the DAC in universal master mode to the upsampler.
Is it true that the SPDIF inputs are not reclocked in the upsampler if it’s run in W1 with the DAC in universal master mode?
Well, I have the Vivaldi DAC, Upsampler, and Clock, and a Cybershaft OP21A reference clock. My apologies for the length of this reply. The TL;DR version is in bold below. I’ve done 4 things to the network ahead of the Upsampler:
1. Standardized on unshielded and certified Cat6 cable (per dCS recommendations). I’ve tried a few “audiophile” Ethernet cables and concluded that my life expectancy did not justify that rabbit hole any more. 
Further, while I believe in the old aphorism “everything matters,” and that includes cables in the digital realm, I believe the “everything” that matters in Ethernet cables is that certification on which one can rely and expect the cable to perform as intended, and as all applicable “participants” [from the designers of the spec to makers of the equipment we use] expect it to behave. I believe a poorly designed or made cable can affect the intended performance, but I do not believe that a properly designed/made/certified cable can act as a tone control or a device to recover additional data about the recording (e.g., soundstage, imaging, depth, etc.) that is otherwise lost by another properly designed/made certified/cable. In theory, an improperly designed/made cable could add something to the signal. (e.g., a variety of different types of noise, but I am still struggling with how an Ethernet cable could deliver more bass or less treble as I have read in some (IMHO only) frankly nonsensical reviews. So, having tried a number of different cables to check against my own expectation bias, I have consistently returned to my standard from Blue Jeans Cable. (I use fiber for a couple of long runs between my primary switch and the switch I am about to describe.)
2. For my speaker and headphone systems, I use a Cisco 3560 switch. These are enterprise switches, with much better PSUs, and audibly lower noise than the (not cheap) SG300 switches that I had previously. I learned about these switches, IIRC, over at the Naim and Audiophile Style forums (along with the 2960 series). Their performance is not based on audiphoo. They aren’t cheap new, but they’re readily available refurbed for very reasonable prices. I have tried so-called “audiophile switches,” and have heard zero audible improvement over the Cisco 3560. In fact, with one exception, they generally sound worse (albeit only by small degree).
3. Between each switch and the DAC or Upsampler, I use a DJM GigaFoil4 filter. What I like about the GF4 is that it essentially incorporates in one box what otherwise requires two Fiber Media Converters, two PSUs, and a short length of fiber, which is what I used to have between each DAC and switch. Even though the GF4 is supplied with an LPS, I power each GF4 with the 5V/1A output of a Keces P8 LPS. [BTW, props to Lumin for including a fiber input on their X1; I hope Vivaldi’s successor is similarly endowed.]
4. Finally, I am currently re-testing the UpTone Audio EtherREGEN. When I first tried the ER, I did not think it afforded any improvement compared to the GF4. Still, I did not return it to UpTone (which has a very reasonable return policy), thinking that I might try it again at some point. I so stated in more than one forum, including here. I also expressed a fair degree of circumspection about the validity of utility of clocking the ER. Back in late April or early May, Alex Crespi of UpTone called me at home, and asked if I was willing to discuss these assessments. I was more than happy to. We spoke for over an hour. I agreed to re-read the “white paper” on their site (which I had previously read), and to re-listen to my system with the ER in it and clocked by the Cybershaft. I’ve been doing that for about three weeks now. And I am beginning to wonder if the damn thing doesn’t make a difference.
I am well aware of the shortcomings of sighted testing, and I know the arguments about the various types of listening comparisons, expectation bias, and confirmation bias, as well as “brain-burn” (which I find more persuasive than a lot of the hooey I read). What I have found works for me is extended listening (much more fun that frequent swapping of stuff like cables or network devices), keeping notes, and occasionally asking my non-audiosilly wife how the music sounds to her. Once in a while, a change will be immediately apparent (clocking or Torus power supply). Other times, it will take more time to perceive that something has changed (Cisco switch or Iconoclast Gen 2 cables). I will continue to listen to the EtherREGEN. I am well aware that objecvtivists tend to be dismissive of it, because of the lack of verifiable measurements in support of some of its claimed effects. The ER is a fully functional switch, and it costs much less than other so-called “audiophile switches,” though it has fewer ports; but I would like to see something that explains why it should/does sound different, or contributes to my Upsampler/DAC sounding different. And I am hopeful that UpTone’s John Swenson will in fact publish measurement data to back up the claims (I like to understand the engineering, when I can; ergo BJC and Iconoclast, for example); their white paper has a lot of words, but UpTone acknowledges that it’s short on measurements. Until then, I will keep listening. The ER does no harm in my system, and I am open to the possibility that my system sounds incrementally better with the ER in it.
Greg very interesting.
I had both a GigaFoil 4 and a Keces P3 LPS in my set-up and could not get it to work without repeated dropouts. My head was spinning with all suggestions I was getting from my (GigaFoil) dealer, Small Green Computer, dCS and Netgear Orbi support. I finally gave up and returned the GigaFoil and Keces LPS. Shortly thereafter I purchased a TrendNet switch with a fiber optic connection and ran a ~40’ run of fiber cable to a opticalModule I purchased from SGC which I placed just in front of my Vivaldi Upsampler. Since then Roon has been rock solid without any issues at all.
Jim, I recall you posting about your GF4 difficulty. Obviously, you tried to assess/fix it, but your experience illustrates one very important thing I have learned: you must try tuning/tweaks in your own system on your own time. Reading reviews might give you guidance and insight. But there is no substitute for trying something in your system. Glad that (Sonore?) opticalModule is working out for you. To me, it looks like a nice package
Greg, please keep us posted on your findings with the EtherRegen. I’m currently using a Netgear GS105 that I bought back in 2007 running off a linear power supply that I happened to have on hand.
The EtherRegen could be a nice replacement if it indeed offers some benefits.
An alternative approach would be the Cisco switch you’re using, but I’m not clear if that would offer any benefit over my current setup, which is already running off clean power.
Jeff: I’ve actually got one of those Netgear 105s. I bought it when I lived elsewhere and used it to extend some Ethernet into a different part of the house. They work, but they can be improved. If those 5 ports are enough for you, I would encourage you to give a try to a Cisco 2960 or the EtherREGEN. The ER is expensive compared to a refurbed 2960, but it’s returnable if you hear no benefit. And you can pick up a 2960 for $50. This one doesn’t have PoE, but it doesn’t sound like you need it. That’s just a steal. It’s all gigabit, and over at the Naim forum, there’s been some suggestion that 100mbps might actually be preferred, but I don’t know that this assertion has been proven. (BTW, that thread link is to one of the short threads, not the multi-thousand post thread.) I know the isolated port on the EtherREGEN is 100, and that is more than fast enough for me. And I believe Andrew or James has mentioned that 100mbps is more than adequate for streaming.
You may hear no difference with either of these devices, but I know I am hearing a difference between them and my previous Cisco SG300 switches.
P.S. Sheesh, here’s the 10/100 version for $35.
Thanks for the info Greg! I went to the local computer mall today and found a shop selling used networking equipment. There was a 2960G in the window and it’s now mine along with a pair of Cisco SFP modules 
Did you ever compare ethernet vs. running fiber from the Cisco to a media converter with a clean power supply? I couldn’t find the appropriate converter today but I have one on the way. Seems worth a try, perhaps for a similar effect to the EtherRegen or GigaFoil4.
If I can get most of the advantages of those solutions with this setup it will be a great value. Total outlay for the switch, SFP modules, fiber cable and media converter was around $130!
Thanks,
Jeff
Sounds like you got a great deal Jeff; props! A 2960G in the window; that’s amazing. I just received a 3560CX yesterday; it was sold as used for $199, but it arrived in a Cisco sealed box, and actually appears to be new. My guess is that it was part of a corporate inventory somewhere, so technically second-hand, but never deployed. Luck of the draw.
To your question, yes I have done exactly that. I have done FMCs with power supplies from UpTone and Teddy Pardo. But I disliked having all those extra boxes around that I needed to hide. And there is some theoretical possibility of the fiber converters generating RFI that can get back into the line. I suppose a way to check for that would be to try both shielded and unshielded Ethernet to see if there is any perceived difference. I do think our dCS boxes are very well designed and engineered by people who know a thing or two about noise and ground planes. Still, no harm in giving their design the best possible input, is there?
Anyway, that experiment convinced me of the benefits of some sort of fiber isolation somewhere in the chain, preferably as close to the DAC as possible. Also, I liked running fiber, though the apparent fragility of the cable always makes me nervous.The Gigafoil4 is an excellent answer to all that: no real fiber to run and very robust. Easy to power with a single PSU [their supplied linear PSU is perfectly competent, but I have mine running out of a Keces P8 that’s already next to the cabinet]. Obviously, it’s not worked for everyone, but mine have been great. It’s just not real cheap. Neither is the EitherREGEN, though both are much less expensive than a lot of other foolishness being bandied about. And they do offer some true network noise isolation. But I think that’s all they do. As @Anupc has demonstrated before, and is continuing to work on, even the cheapest of switches and compliant Ethernet cable will deliver bit-perfect files. Any changes in sound have to be caused by or related to something other than asynchronous data over Ethernet.
As I may have mentioned elsewhere, on a properly set up system, I am not at all certain I can tell the difference between copper and fiber—and one side of my brain tells me I should not be able to tell any difference. But in the back of my mind, I know I have optimized the network for the best signal into my equipment. And I think that offers me some psychoacoustic peace of mind. I’m a “belt & suspenders” kind of guy, and besides, I like tinkering with this stuff.
Hi Greg, sounds like you got a great deal too! Curious, why do you go for the 3560CX over the 2960? Does it offer a feature you need that’s missing from the 2960? It’s certainly nicer looking
I considered an EtherRegen as well, but the additional cost doesn’t seem warranted if I can acheive a similar result with this fiber setup and a linear supply I already own. I enjoy tinkering as well, so this setup is a bit more fun to put together!
Primarily the Layer 3 option. Right now, my other 3560 [which I snagged for a silly low price] is PoE powered, and I wanted to try a swap with a unit that is AC powered. I have a place for each of them; it’s just a question of whether there is any real advantage to one or the other being in the audio system. I like how these things are built.
Also, I think your assessment of rough equivalence between your fiber kit and ER is probably correct
I am not sure you really gain a lot by having the DAC end of a fiber link connected to the DAC via ethernet. Looks to me (without having tried it) that Problems of PSU and other nasties getting into the DAC would be no different than connecting a switch to the DAC via ethernet. Now if the DAC had an SPF cage, so the fiber could be connected directly to the DAC (like in the Lumin X1 case), things might be different.
I have tried it, and it can be different. It depends on how one does it and, to your point, the quality of the pieces and PSUs. In a network full of noise and noise sources, isolating that last little leg, and using good power, can make a difference. If one doesn’t have a noisy network, it might not make much difference at all.
Greg and Rudolf, you may both be interested in following this link to a piece by the English reviewer Andrew Everard from as long ago as 2015 when he changed over to fibre:
Thanks Pete. Interesting to read. 2014-15 was when I first began putting fiber into my network. Lack of networking diagrams makes it a little difficult to understand how everything is routed in his system, but between AirPort Extreme and old Airport extenders, it sounds like he has a mess of a network. But I confess I am dubious that a 3m swap of copper to fiber should have had the results he describes, given his first insertion of fiber 10 months earlier, unless he has something making an awful lot of noise before his Naim. And that’s possible given some of the descriptions of his network.
I find his dismissal of expectation bias amusing, especially now that he’s done the fiber swap twice. I think the process he describes is a textbook illustration of expectation bias.