What makes One Music Server Sound Better than Another?


So this week my Mojo Audio DejaVu music server that I have used for the past 2-3 years crapped out. Benjamin at Mojo was more than helpful and the DejaVu is on its way to Mojo Audio where it will make a full recovery.

Thankfully, I still have my Antipodes DX2 Gen 3 (their former flagship) music server so I hooked it up. After wrestling with Roon protocols, transfers, and set-up menus, I was able to get it going so I have music. The DX and my Sonore Sig Rendu SE opt. are both connected to my network so the DX (like the DejaVu), is only being used as a Roon core and the Sig Rendu SE serves as the Roon endpoint for streaming Tidal and Qobuz, with a direct USB connection to my DAC.

The point of this thread is to ask, how come I perceive the the DejaVu server as sounding better than the Antipdes DX? In fairness, the differences I perceive are not great but it seems the DejaVu is fuller sounding, more tonally rich, and bolder. Is this why some here spend $10K+ on a Grimm, Taiko or something else?

If a server is basically a computer, sending digital information to a streamer/endpoint and, assuming that digital information is transmitted asynchronously and reclocked by the DAC’s master clock, and assuming noise is not the issue (i.e., both units are quiet and there is an optical break between the network and both the server and endpoint) then what are the technical reasons one should sound better than the other? It is not that I want to spend $10K+ on a music server with a lifespan of maybe 5 years before becoming obsolete, but I would like to understand what more you are getting for your money. So far, the best I can come up with is lower internal noise as the major factor.

As a side note to the above, when I thought things looked hopeless for getting set up, I scheduled a support session with Antipodes and, although I lucked into the solution before the meeting time, Mark Cole responded ready to help. Setting up the session was super easy and reminded me of the superior level of support I had come to enjoy from Antipodes during the time that the DX was my primary server, including multiple updates and 2 or 3 hardware upgrades, which prolonged the service life of the DX. Good products and good company.

 

mitch2

Showing 18 responses by jji666

I added ferrite cores to my ethernet cables as a wild guess at reducing radiated noise but no idea how effective that is.

This is not my area of knowledge, so just passing on what I've read.  You may want to Google the efficacy of ferrite chokes on network cables. I have read that it can actually introduce errors.  Presumably the network components can correct for that as I and others have described above, but the ferrite core may actually be more of a concern than anything a fancy server can resolve.  

To be clear, I am not expressing an opinion or disagreeing with anyone. Just suggesting you do some reading and confirm you want to take this approach. 

The arrogant puffery here is funny for its sheer lack of social inhibition.

Boy, this is a game of whack-a-mole, that’s for sure. But what really gets me is that when someone has nothing substantive to offer, they have to attack the messenger. Sorry to disturb your pro-fancy-server echo chamber.

The fact is that one poster told me to "get out there and try some servers" (paraphrased) and another used that old construct of "my system is more resolving so I must be correct." So I described my experience, both with servers and resolving systems, in order to respond to their assumption that I was speaking without experience.

I advocate keeping an open mind. But those that state beliefs contrary to science and engineering are the ones that need to back up their assertions with evidence.

arrogant puffery

It’s not puffery if it’s true. In terms of arrogant, I suggest you check the tone of your own post.

OP: I don’t know exactly how those two different servers run Roon - it’s possible they’re adding some DSP or otherwise doing non--Roon-standard manipulations. But if not - same Roon version, same configuration, across a network, bit perfect is bit perfect. Doesn’t matter the server source.

For clarity, my post is not about hooking a DAC up to a server directly, where at least there is a form of direct electrical connection. It’s about server streaming to endpoint over a network.

If some of the posters’ suggestions here were true, when you go to an ATM to withdraw cash sometimes you get a few dollars less and sometimes a few dollars more. When you check your bank account online the numbers will be different every time. When you email someone, the network will introduce typos. When you send a word document, text will change.

IP network transmission simply does not work that way. As long as you have bit perfect music input at the server end you’re going to have the same bit perfect output, no matter what else is at the server end, at the streaming end. [Admit the distant possibility of streamers sounding different].

If you want to explore this, I suggest you take a computer and use that as a Roon server to your same endpoint and do those comparisons. As long as you’re not hearing dropouts, and each is configured for identical DSP, volume, etc., whatever processing you have configured is identical (and there’s sufficient horsepower!), it’s gonna sound the same. I have 3 roon servers here on different platforms for different reasons, but it’s not SQ.

"It’s about server streaming to endpoint over a network."

That is exactly what I am doing, along with having optical fiber isolation of both the server and the endpoint from the network.

Sorry, should have been more clear - the reason I added that paragraph is that someone always posts about concerns only related to having the DAC connected to the server.  All of this about noise inside the server running down the USB cable.  I wanted to be clear that concern cannot apply to your situation or what I was discussing. 

 

My plan is to continue using a reliable, well-built, server, somewhere south of $5K,

Part of the point is you don't need to do that, albeit I am not criticizing the choice.  It's more of a matter of convenience if you want auto CD ripping and things like that.  Roon works with any PC with sufficient horsepower to run whatever DSP etc. you want to use.

You won't get any sound bump from any choice unless you hear hiccups or dropouts caused by an overtaxed server, which you can easily determine with the processing speed indicator in Roon.  I have a fairly powerful machine currently running at 60x, meaning I can theoretically support 60 streams before any issue.

Not everyone wants to build their own PCs, I get that.  But it's actually not rocket science and you can build a well overpowered Roon server for $1000 maybe less if you have a PC with some of the parts already that you can salvage (power supply, case, the stuff that gets obsolete less quickly).

My personal view is people buy these pre-built audio server computers because they look cool.  Which they do.  But they're sold with stories of superior audio quality when you get to identical quality with the above plus your chosen streamer.  [Repeat acknowledgment of theoretical possibility of streamers sounding different].

Anyway, I just say this so that when newbies search "what music server should I buy to use Roon" there will be some balance of opinion against those expensive, fancy servers that get obsolete within a few years due to ever increasing CPU, memory, and hard drive speed, when you can upgrade a standard computer for ~$500 instead of spending 4 to 5 figures to replace the whole fancy prebuilt server.   

I'm talking about the SQ of a server when using an endpoint over a network.  It is the same set of bits at the receiving end, no matter the transmitter. That exact configuration is all I'm referring to.  

As I said in my post above, if that wasn't true, then emailing word files would introduce typos and your bank account balance would be different every time you check it.

That isn't to say that fancy servers don't have their advantages in terms of other functions and features, support, being well built, and looking cool.  But over a network, the packets don't know if they are coming from a supercomputer or a networked doorbell.  

it sounds better.  it does.

Not sure if this is jest.  But if it sounds better, which empirical result I am open to, then it is because there is a different configuration. The server is not sending the same set of bits, which necessarily means that the server has something different configured within the software.

Because it cannot sound anything but exactly the same unless the bits it sends are different. The receiver simply doesn't know whether those bits are coming from a fancy server or a Raspberry Pi. 

Imagine someone throwing a floating candy in a stream of water, and then 1 mile downstream you pick the candy out of the stream and eat it. A low grade computer is a contaminated sewer infested stream, with industrial waste, dark brown in color. A high end music server dedicated to audio the stream is pristine, perfectly clean, pure water. You get the candy with either stream, it is the same piece of candy either way, but which one would you eat. 

I think you just nullified your own argument?  I'm not saying that to get into a fight - but you say it is the same piece of candy either way.  Yes it is the same, identical, exactly the same, not different, not dirtier, piece of candy.

In an IP network, the candy is broken down and re-integrated at the receiving end, exactly as it was before.  Other than the lag it would cause, it could go around the world, into space through a satellite, and back down and be the same.

What you are saying is that a dirtier computer sends different bits.  It does not, or it would not be bit perfect.

It just doesn't and can't work the way you describe.  Sewer infested stream, brown water, those are analog concepts if they can be applied at all.  What comes out of the network card, over the cable, and to the switch is identical regardless of your pollution analogy.  The network doesn't care about the brown water in the server.  It just doesn't.  

I am not saying that you should not purchase fancy servers.  But don't purchase them for sound quality unless you insist on plugging the DAC directly into it without using a streamer, which isn't a best practice according to Roon.  

Have you ever tried to catch a chicken?  

Imagine someone throwing a floating candy in a stream of water, and then 1 mile downstream you pick the candy out of the stream

Read the link I posted previously from Antipodes

OK, so we have chicken, word salad from Antipodes, and candy. At least we won't go hungry.  But the SQ remains the same.  Digital logic, especially in IP networks, simply isn't amenable to analog logic, much less analog metaphors.

What can account for significant sonic differences despite being bit perfect has to do with the timing of the data stream as well as embedded noise and jitter. 

This *could* apply to the digital transmission of data from a computer or other digital device via USB to a DAC, which is a kind of a stream.  Technology has addressed it a long time ago, but theoretically at least your concern above could apply.

Doesn't apply to IP networks.  The packets are reassembled, anything missing is requested and reintegrated, before that info goes anywhere.  The buffer assures there is sufficient data for continuous music.  If that failed, you'd hear a dropout, not less depth and bloom.  

 

I think you just need to go a good audio dealer and ask them to play you your favourite track with a macbook and then a high end streamer to prove it. 

I have actually done quite a bit more than that.  I've been building all variations of PCs from fanless quiet nothings to super beasts. So I've had my hands on a lot of different variations. 

Some folks just can’t get out of their own way. They rather choose to butt heads and throw around their expertise irrelevant to the matter on hand 😊

Now now, no need for ad hominem attacks because someone posts a contrasting viewpoint.  That reflects much more poorly on the attacker and just means you don't have any substantive point to add.  

If I were to be snarky in kind, I'd agree that my expertise is irrelevant to the matter at hand because the matter at hand is purchase decisions irrelevant to the actual objective of sound quality. 

But I won't.  Instead I will just say that I'd rather have a positive "relationship" (such as it is with a forum) with someone who loves audio even if we cannot see eye to eye on these matters than get into a flame war. The point being there are so few of us audio lovers why do we have to battle. But if you're going to flame, I'm not going to just be a target either. 

It is absolutely true that I have never caught a chicken.  Not unless it came from a drive-through window.  But I do remember when my relatives in Sicily taught my sisters how to pluck one, thinking it would be a useful skill for them in the US.  

 

Read the link I posted previously from Antipodes. 

I had scanned it before, but I have now read it closely.  Actually, I think it supports the position that the server end of a server-network-streamer configuration does not contribute to SQ, assuming sufficient processing resources to use DSP and avoid dropouts.

Here is what Antipodes says about their servers: 

Step 1 runs the Server app and does the best job if it is a computer with relatively high power (but not too high), and with a lot of RAM. It is the heavy-lifting stage and achieves a lot, but the power needed makes it relatively noisy (in an electronic perspective, not an acoustic one). In a car cleaning analogy, this stage is where you water-blast the big bits of dirt off, but the result is not yet acceptable.

So Antipodes does what Roon suggests: it has a high powered, electronically noisy server that decompresses the FLAC to PCM and applies DSP.  This is an acknowledgement that no matter the electro-noise at the server side, the network will clean it off for them.  

Here's what they say about their "player" side (streamer in our lingo):

Step 2 runs the Player app and this is where the essential neutrality of the sound is determined. To achieve this we use only a moderately powerful computer because we need to get the electronic noise interference levels down to much lower levels than what comes out of Step 1. 

Right.  the noise at Step 1 doesn't matter - it is "cleaned up" later.  While Antipodes allows their streamer to take credit for getting the level of electronic noise down from what the server sent, it is actually just the network that does it.  However, fine, good to have an electrically quiet streamer just to be sure.  

The remainder of the Antipodes steps are not relevant to the conversation.  The underlying point is that Antipodes itself acknowledges that it is not cleaning up the audio quality at the server end.  It is simply doing the same thing that ANY sufficiently powered computer will do.  Except, and I don't know this part, maybe there is a proprietary protocol in communicating with their streamers so that only the Antipodes server can talk to the Antipodes streamer so that you have to buy both even though the server part doesn't need to be quiet or clean or anything but a computer.  

I will say those are beautiful looking pieces and I appreciate how they relabel the backplane for audio purposes so that those who aren't comfortable with computers can use them.  However, I'd bet if you plugged a ripper into the "disk" port  or a disk into the "Ripper" port, they would work either way because they're just standard USB ports.  

Look, these guys make beautiful high end computers for use with audio.  There is nothing wrong with buying one of those.  I would if I had that kind of income.  And they cannot just come out and say "you don't need our servers to feed a streamer and get the same quality" albeit that is the natural result of what they do say.  

But, please, just don't buy into the idea that you HAVE TO have one of these to optimize sound quality.  It's a shortcut.  You know it's quality gear. But the server side simply isn't relevant to SQ.  Not as long as the server can keep up with the stream it needs to send. 

 take your server/player to somewhere where their is a high-end "audiophile" server/player, setup with appropriate highly revealing other components and cables and compare.

I believe I've already done this without having to take it anywhere.  I've built all level of servers, from the AudiophileStyle's totally quiet fanless model with no moving parts and using the various OS tweaks that are/were supposed to bring OS operations to the quietest levels possible, to powerful beasts that may clearly have noisy electro-stuff.  I mean, I have probably built 15 Roon servers.  All flavors. The transmitter in a network configuration just cannot generate a different sound.

I've been designing and building media computers for over 30 years - I'm not just guessing here.  What I don't have are the fancy milled aluminum parts to make those beautiful chassis.  Please, don't tell me the chassis affects SQ!

In terms of resolving gear, I have Magnepan, PSB, KEF, Monitor Audio, B&W, Krell, Balanced Audio, Focal, Levinson, Classe, Bryston, Parasound, Wyred4Sound, Auralic, Schiit, and Mytek.  I'm good on resolving.  

You can talk about how best to catch a chicken but until you have experience catching chickens, what you have to say is mostly meaningless.  

See above.  By your analogy, I have caught a lot of chickens. I have been obtaining, churning, buying, listening, tweaking, all manner of resolving enough gear for a long, long time. 

 Music, like all art is personal.

Music is personal, for sure. What isn't personal is how IP networks function.  What isn't music is what goes from the server to the streamer. It's packets of IP data, the same regardless of the sender.  

This "my system is bigger than yours" method of winning arguments, well, just try to read it objectively. 

Well jji666, you have a contra point to every point.  That is called being pedantic.  

Or, maybe, just correct?  As above, ad hominem attacks really just affect the credibility of the attacker. 

I wish you happiness enjoying this hobby in your own way. 

And absolutely the same to you.  As I said above, I would very much prefer to focus on our shared passions rather than differences in belief.  

And to wrap it up (hopefully), compared to many of these debates, we've gone back to our corners relatively peacefully.  I think we should both be commended for nothing worse than a few swipes. 

The K41 server has two ethernet ports- one purposed for input and one purposed for output to the player.

This is a cool configuration.  I happened upon something similar by accident when I was trying to shield my very old, but still super quiet and super useful, streaming Windows 7 machine from the internet.  I just connected it to the second LAN port on my Roon Server.

What isn't personal is how IP networks function.  What isn't music is what goes from the server to the streamer. 

OK, I can see how that comes off as pedantic.  It wasn't intended to be so - I was excited to point out something I'd read another poster say over at the Roon forum, which is that it's better not to think of what your Roon server transmits as data, not music.  Anyway, please, continue on the other topics and let's move on from the core debate...

 

I disagree, you seem quite closed minded.

Relying on science and engineering isn’t being closed minded. I have honestly and sincerely considered the points put forward by those who believe two servers with identical software configurations but different hardware would transmit two different "bit perfect" audio signals over an IP network. It is just my belief, which is backed by actual network and computer engineers (who do not sell the stuff - therein lies the difference from the marketing pieces) that this is impossible, and if computers and networks worked as you believe, they wouldn’t actually work at all for things much more mission critical than music.

Mark Jenkins:

“The motherboards are sourced from the world’s best supplier and they cost around 6 times what some of the competitors are using. We tune the motherboards to shift the frequency peaks of the noise generated by each component in order to eliminate noise nodes, so the mainboards start as an off-the-shelf board and then are customized for our use.”

Antipodes also places a lot of emphasis on the quality of the power supply, which they manufacture entirely in-house.

Mark Jenkins:

“What we did with the new power supply was to test the injection of noise into the motherboard at various frequencies to see which frequencies did the least damage to the sound quality, and then we designed the power supply board in such a way that the noise component was in the benign frequencies. This has a similar effect as a zero noise power supply.”

Lovely. I’m sure the stuff is as solid as the Rock of Gibraltar. There may be many reasons why this represents an upgrade over a standard PC running Roon: reliability, utility with DACs directly connected, a bunch of stuff.

But none of that stuff includes sending "quieter bits" over an IP network. There’s no such thing. They’re not dirty candies; they are not elusive chickens; they are the same packet of data as would be if the computer was built of space shuttle parts and encased in lead.

I am not criticizing anyone for purchasing one of these. They look fantastic. Nice bit of kit...nice bling for the system. But from an IP network perspective they can’t send cleaner packets. It’s all cleaned up by the time it gets to the streamer.

Call me all the names you want. It doesn’t change the science or the engineering. And what I meant by "need" is that if you want to properly back up your position, when it’s contrary to established science, the onus is on you. No, you don’t have to. But then don’t criticize someone trying to engage with you from a science and logical perspective.

It’s downright tribal. Am reminded of Lord of the Flies

It does seem to be. It’s just depressing that we can’t engage in a civil manner and that when someone runs out of points they turn to insults.

I am very familiar with Synergistic Research digital cables.

Your input is appreciated.  Just to point out, those are not network cables and what we are specifically discussing is if some electro-noise at the server end can effect what is received by a streamer over a network, presumably through one or several switches. 

LOL

With respect, that is not LOL-worthy.  At least if I understand your proposed logic, that because a USB cable and a coaxial digital interconnect sound different to you, that the inner construction of a computer can effect what is received from it at the other end of an IP network.  Apples and oranges. 

One more thing I would like people to consider:  fiber optic network cable cannot possibly carry any electrical charge or interference. Many suffering from audiophilia nervosa will use this technique to rule out any electrical interference traveling from server to streamer.

Fiber optic network gear can be had for well, well below $1000.  That will categorically eliminate any possibility of the relevance of the electrical situation within the server. 

So, even though I don't believe in the difference in the first place, it can easily be eliminated for well under the cost of these servers.  There is no need to do it at the server end. It's just a different way of looking at IP network functionality - whether it be twisted pair or fiber, the noise, if there is any, doesn't make it to the streamer.