why does streaming from local storage sound better than the same album, at same resolution streamed from Qobuz?

[Blackmorec]

In order to move a digital file along a piece of wire it needs to be converted into a voltage.  The voltage you create when listening to a ‘stored file’ is basically created in the server, using the server’s DC power supply, then shipped out to the DAC via an interface along a single cable. The DC power supply adds its own noise and ripple to the voltage representation of 1s and 0s and the cable adds any radiated emi voltages it may pick up.  Servers very often use high quality, low noise power supplies optimised for audio. So the resulting noise can be fairly minimal. 

 

When listening to remote files from Qobuz for example, the data bit stream is synthesised several times along the network, with each power supply adding its own noise and ripple to the stream. Power supplies in commercial networking products are usually $5-$7 affairs made in China from the cheapest components possible and are not optimised for audio applications. Then there are several cables along the way to pick up stray EMI. Noise from the multiple plover supplies interacts and create harmonics, so the final noise spectrum CAN be immensely complex.

 

So what ends up entering the DAC is either noise from a single, audio optimised PS plus a single cable’s worth of EMI in the case of a server stored file vs noise from a variety of super-cheap power supplies, plus harmonics, plus EMI from several cables, often long in length for remotely streamed files. 

 

[Blackmorec]

4 hours ago, plissken said:

 

The data, as it passes through routers is completely reconstructed and forwarded. Switches and routers routinely perform store and forward. Literally STORE data and then FORWARD.

 

When we are talking streaming content, ripple isn't a concern because you aren't going over copper but optical on these transits. There is no AC ripple on optical connections.

 

 

I'm curious where your information comes from. For my Aruba builds my power supplies have a cost starting at $350 for the lower wattage models. I'm pretty sure the BoM is way beyond $5-7. You also get a life time warranty *Lifetime for the product cycle*

 

Or did you mean consumer?

 

 

Huh. The data doesn't support you in this though. All these devices convert AC to DC and by virtue of conversion everything besides 50/60hz fundamentals are all sorted on properly designed gear like the Meanwell Supply with the ground strap.

 

 

I hope you aren't out there giving advice.

Here’s a picture of your average power supply in consumer grade network electronics. The size is roughly 100%  ie. correct size. 

When a data stream is ‘reconstructed’ (I used the word resynthesized) for transmission on a standard consumer network this is typically the power supply on which the data stream is based. Also bear in mind that the incoming power to this supply is 230V AC, so there’s also an AC to DC rectification going on, with all the associated vibration. By removing this supply and powering the circuit from a remote LPS like a DC3, the upgrade is SQ is night and day, like you’ve upgraded an amp or speakers.

 

The Meanwell supply in one very well known and hugely expensive Audiophile network switch can be had for $9 bucks, $7 if you buy more than a hundred units. 

 

In my original post I’m talking about a standard ethernet network based on typical electronic retailer supplied consumer networking products….not hand built, enthusiast designed servers. 

 

 

 

 

 

 

[Blackmorec]

3 hours ago, MarkusBarkus said:

...yeah, the power and data models may be a bit out of date, @Blackmorec. For example, I have about three feet of copper in my network. The rest is fiber. All components, including network components, are on good quality LPSs. For example.

And with this set up do you still hear big differences between playback of locally stored and remotely streamed files? I’m talking about the cause of why local file playback sounds better than playing remote files over a typical LAN built with consumer grade electronics. What you are talking about is usually not the cause, rather the fix…ie steps taken to improve how remotely streamed files sound. 

[Blackmorec]

12 hours ago, davide256 said:

Sounds like you are not a STEM graduate, haven't trained in digital circuits and networks. Impactful noise can only be originated from the router/switch itself to a

connected source/endpoint. Do worry about that. In between any noise passed from a next hop  device is far, far below the noise level of the connected router/switch

itself. Commercial networks no linger use wire connection other than for local hop, too slow, everything is muxed over fiber between switches and routers inside a commercial network.

 

More troublesome is product quality on packet/frame processing, cheaper gear behaves more erratically than a Cisco or Juniper product, Don't cheap out on your network

hardware. Intel ports are good but beware of the "comes with" Ethernet port of other manufacturers. Historically endpoint devices behave oddly when the network behaves oddly

 

The line about "data bit stream is synthesised ", thats not how it works. Data is transmitted in IP packets encapsulated inside  Ethernet frames, read into and out of buffers as its

received and transmitted. The data operations that occur are on frame and packet headers in order to switch/route traffic, data payload  is untouched between source

and destination.

Hi there, i used the word ‘commercial networking products’ when more accurately i should have said ‘consumer networking products’. I’m talking about the stuff that the average consumer is given by their ISP or buys from their local Branch of Curry’s or PC World.

 

Also you are talking high-level Network Standards jargon, while I’m talking about what actually happens physically when a voltage stream from a cable is stored on or read from a local buffer in a network device like a router, switch or network bridge.  

[Blackmorec]

What I learned by studying the networking standards and theory is that whenever files are moved across a network, as long as the final file is a ‘bit perfect’ copy of the original file, all should be good.  But is that true for audio? In audio, we take that bit perfect file and transpose it into an analog voltage stream which we amplify then transpose into sound pressure waves, which we then detect using our ears and transpose twice more, once into nerve impulses then again into conscious music in our brain.  And it seems that when the final measure (detection device) is our subconscious limbic system, the ‘bit perfect’ strategy no longer seems to hold up given that otherwise identical files (from a bit structure point of view) sound different….sometimes markedly so…one generating real pleasure and happiness and the other seeming to grate and irritate. Hmmm. So something is afoot, but what? At this stage we need the measurement guys to step in and tell us what’s changing.  But so far all they tell us, using standard measurements is that nothing has changed but that can hardly be true, given that we perceive large differences. 

My feeling is that they are looking in the wrong places and measuring the wrong things. What we are looking for is very, very small changes in amplitude, phase and frequency in the differential between the 2 signals reaching our ears. If those change, even by minuscule amounts, what we perceive changes. So once we find out why we hear 2 different versions of an identical bit stream, we can begin to understand what’s generating those changes and come up with strategies to ameliorate those shifts. 

Of course, the above contradiction of the theory is based entirely on perception, and to fully understand what’s going on, we need to thoroughly investigate the whole process by creatively applying the most sensitive measurement tools, of which we have lots. Given my background in highly sensitive analytical systems, I’m fairly sure that the key is in creatively applying measurements rather than just slavishly following standard test procedures, which to date, to the best of my knowledge have revealed nothing. We need to look at the basis by which our brains detect/identify the differences and work out how to model those same measures using existing tools. 

Experienced audiophiles with highly resolving systems have learned that by applying better cables, routers, clocks, power supplies, DC cables etc. to a network, perceived sound quality improves. We can correlate those improvements to 

less jitter

less noise and ripple

less EMI

less vibration 

fewer cable reflections

less network traffic

less CPU interrupts etc. 

But what we haven’t yet done is qualify exactly what those effects are having on the final music we listen to. 

I have spent the past 3 years improving all the above aspects of my network and digital streaming in general and my system has reached a point that I never believed possible from digital in general.  My observations predict that if were to install yet better USB and Ethernet cables for example I would hear further improvements, similarly if I were to further improve my already top class power supplies, I could still gain substantial benefit. So far, having performed dozens of similar upgrades I haven’t been disappointed, even once.  I would however be really happy to find out precisely why these improvements are having the effect they do but the uniform answer I always get is ‘expectation bias’, which is frankly BS, given the cumulative effects.   What we need is some really good EE or CS PhD student to undertake the research of how network quality impacts sound quality in audio network streaming. Such research would have considerable commercial applicability and I’m damned sure some quite decent funding could be put together via a consortium of interested companies. 

The reason locally stored files sound better than their remotely streamed counterparts is entirely down to network quality. We can restore the balance with carefully selected network components, but we still can’t explain in scientific and engineering terms exactly what doing so accomplished, beyond the improved perception. 

[Blackmorec]

2 minutes ago, hopkins said:

 

Using a DAC with an electrical connection to the rest of your system is asking for trouble. This is the case with USB, SPDIF, AES, I2S or Ethernet inputs. These connections are in fact like highways (with several Ghz bandwidth) for noise to enter the DAC. 

 

Lowering noise in your "source" is not something that can be done with any reliability - you would have to measure exactly the spectrum of noise, and that is not possible as any small change is going to matter and every single configuration is going to be different.

 

The best approach, and the least costly, is to improve things on the DAC side. Whether all this matters or not, and if it does, how "far" can you go using the type of solutions you describe, is anyone's guess. 

I follow your logic but here’s the problem. Take any one of the best DACs in the World (you decide which, because it doesn’t really matter), connect it up however you want and have a listen to a remotely streamed file. Mmmmmm wonderful, I’m sure. 

~Now go ahead and upgrade the power supply on your pre-server switch from the standard, cheap-as-chips SMPS to say a DC4. And what sounded wonderful previously, now sounds doubly wonderful and that’s because with every piece of hi-fi equipment I’ve ever encountered in over 45 years of very actively pursuing this hobby, they all follow the same rule….the better the input, the better the output. I not aware of a single piece of kit that fails to respond to this precept.  I am aware of one very talented manufacturer who is developing a system that by design produces great music independently of the network input quality but even that system responds positively to replaying a file streamed to RAM then disconnecting the network.   

 

As to least costly…..there you may be right, given that I haven’t compared a lot of DACs on a straight switch - server - ethernet patch cords vs my DAC fed by my optimised network, so I can’t really comment. 

[Blackmorec]

11 minutes ago, hopkins said:

 

I don't see how any of this is any different than what I said. 

A better DAC still has the same problem. ~It may sound better but it’s still equally sensitive to the network quality. So your solution doesn’t address the actual problem, given that your better DAC will still exhibit large differences between local and remote file playback as long the remote playback is via a poorly optimized LAN 

[Blackmorec]

17 hours ago, plissken said:

 

This is easily tested. I'm game for testing this if others are. I'll help put in the effort if others will sign on to participate. But @Archimago already did something like this...

 

Doesn't mean it can't be done again.

Hi Plissken 

As you say, its extremely easy to test. Just throw an ethernet cable over my bannister rail and connect between router and server, thereby bypassing all my optimisation measures. And the result is a massive downgrade, like I’d downgraded from some Magico M2s to some very much lesser speaker….ridiculously easy to hear the difference.  A complete collapse of the fully immersive, 3 dimension musicians playing instruments to a much more 2 dimensional presentation.  I actually tried this manoeuvre before spending more money on further improvements just to make sure the effects I was hearing were indeed coming from the network. 

 

One thing I would freely admit…..that what changes with these network optimisations is the presentation of the music….for example the main effect of adding better power supplies could be heard in the pace, rhythm and timing and in dynamics and micro dynamics of the music, while with better cables and vibration control, the effects were mainly on detail recovery, air and atmosphere. By continually improving the network, the sound took on an entirely different nature, becoming holographically 3 dimensional and completely immersive, like you are sitting in the middle of a huge sphere of music, where the venue and its musicians are replayed in a most believable way (assuming the recording has those qualities obviously) The whole upgrade strategy brings predictable and reproducible improvements.

[Blackmorec]

13 hours ago, firedog said:

Correlating doesn't actually show that anything is happening that improves or degrades SQ.

Can you actually show: a) cause and effect - any of your changes, and "less jitter" etc?

b) if you show changes in (a) does that cause any change at the output of the DAC?

c) Is that change audible? Changes at -115 db and more don't count. No one can hear them. And I'm being generous. 

Hi there firedog,

I suggest you read the very next paragraph I wrote following those ‘correlates’ 

 

“But what we haven’t yet done is qualify exactly what these  effects are having on the final music we listen to” 

 

Pretty much encapsulates your above comments, I would have thought…..at least that was my intent 

 

Also this arbitrary -115dB…..have a look at the amplitude difference between an LHS sourced 85dB @ 1m signal at a 4meter distant listening position at the left vs the right ear. The differences are very small….in the order of 0.4dB  on a 73dB signal and we have absolutely no problem in hearing that….no problem = the difference is HUGE because that’s how we assign directionality to the sound source.  The question is not “what absolute minimum level can we hear” ….the question should be “what minimal differential amplitude can we detect between our 2 ears”. I think you’ll find that the answers are VERY different in that we are FAR more sensitive to small differentials in amplitude than we are to overall signal amplitude.  And what seems to be changing via these network optimisations are indeed these differentials.  

[Blackmorec]

19 hours ago, davide256 said:

But you aren't... please  understand that 1 reason why we do digital networks instead of analog network signal transmission is that it eliminates additive noise transmission.

 

No, thats not the case. This is purely a function of software and endpoint hardware optimization. Qobuz legally has to do content protection to prevent piracy, which adds overhead to streaming  caching and causes background network activity demand on the CPU. An offline local file eliminates the network activity demand on CPU and a purchased file allows you to bypass the Qobuz app entirely for your preferred/optimal player.

So why does changing from ethernet to fibre optic make a huge difference in terms of SQ? Why can you hear large improvements when replacing cheap network power supplies with much better linear supplies? Why can you literally transform the sound quality by upgrading network components without touching software or end point hardware?

 

and analog noise is an entirely different kettle of fish….nothing to do with what we’re talking about here. 

[Blackmorec]

3 hours ago, davide256 said:

If its directly attached to server or endpoint it matters for wired noise transmission which is analog voltage noise.. if its a device in between endpoint switches/routers

only the routing/switching efficiency  and jitter comes into play.

 

Digital noise is what you get from D/A conversion. IME D/A conversion is vulnerable to PS voltage noise interference and the most important place in the audio chain for

prevention of PS bus voltage noise affecting USB sender/receiver circuits. 

 

Haven't had a need to use the FE port on my Etherregen since I switched from the off brand Ethernet ports on an AMD board to Intel brand ports on a Z390 board.

That said I also no longer have any jitter/throughput anomalies by isolating all audio components to 1 Etherregen.

 

All you are doing by upgrading power supplies on gear in between is reducing internal network delay/ error correction overhead... but networks don't come with a "buy better gear"

light to tell you that.

Interesting!  Can you explain the basic physics behind how a better power supply reduces internal network delay and error correction overhead in terms of what’s causing what?

 

[Blackmorec]

4 hours ago, plissken said:

 

No, this can be done over your existing. It's called tunneling or VPN. It's a logical overlay over your physical infrastructure as it sits.

Too many potential unknown variables.  With a cable swap the only variable is the current streaming network vs a direct cable. Easy and simple 

[Blackmorec]

7 hours ago, plissken said:

 

ZERO variables. Nothing would change with your setup.

Errrr…I’m having some difficulty with the scientific logic here. If there are no variables, there is no comparison. There has to be at least one variable. The whole point is to make a comparison between two things, so what’s the other thing I’m comparing my network with. Surely its an alternate network…..at least that’s how I understand VPN tunnelling?

[Blackmorec]

8 hours ago, plissken said:

 

ZERO variables. Nothing would change with your setup.

Hi Plissken,

I added a lot to my previous post, but the edits were too late and rejected, so in addition to my last post…….

 

Errrr…I’m having some difficulty with the scientific logic here. If there are no variables, there is no comparison. There has to be at least one variable. The whole point is to make a comparison between two things, so what’s the other thing I’m comparing my network with. Surely its an alternate network…..at least that’s how I understand VPN tunnelling? Anyway, let me know the process and I’ll take a look. I really am keen to understand why upgrading network components like PSs and cables has such an impact. I’d like to know what those changes do exactly to the network stream and why those changes have such an impact on sound quality. 

You may not trust my ears, but I do, implicitly. For example, when I changed 2 DC cables on a switch and wi-fi bridge  from Neotech to Mundorf Silver/Gold I first heard an improvement, in that the system hinted at being more 3 dimensional, even holographic….but after a couple of days it was sounding worse, then quite a lot worse as I needed to up the volume by 2dB (no idea why).  The system then would sound better for a day, worse for several days, then better, then worse…until finally it suddenly sounded stunningly better….at which point it stabilised and remained completely stable in this new altered state. Along the way treble presentation changed, bass presentation changed and most of all, the 3 dimensional presentation.  I didn’t analyze the sound to hear those changes….I simply responded with feelings and emotions….loving it one day, finding it irritating the next. In terms of expectations, there were several occasions where it had sounded great and I thought it was finished running in, so my expectation on sitting down for a listen was that it was going to sound great, only for it to sound irritating with all the magic missing. But in the end, I was so impressed with the final sound using those Mundorf DC cables (very kindly made by Nenon BTW) that I changed the rest of my DC cables and all the internal PS cabling to the M silver/gold. The result was truly stunning but the running in was equally long and irritating. I recommended the cables to a friend, who obtained a pair and documented his listening impressions and he suffered exactly the same running in rollercoaster as I’d heard.  Over 45 years of building and refining hi-fi systems I’ve learned that true analytical listening is really quite a skilled process that requires a lot patience, careful comparisons and a really good, stable reference with which to compare, so instead of using the conscious, analytical part of my brain, which is prone to all sorts of conscious biases, I’ve learned to depend on the limbic part of my brain and simply monitor how the system is making me feel,  Joyous, happy, elated, excited etc or disappointed, unmoved, uninvolved, disinterested. I don’t need to know what specifically changed, just whether my system is more or less magical to listen to.  During listening I’ll write down adjectives that pop into my mind to describe what I’m hearing and how its making me feel, so in the end I can write a description of what I heard. 

With a science (but not IT) background Id very like to understand what’s going on, but I do get really irritated when someone quotes a lot a networking standards jargon and claims that changes can’t happen, because what’s actually going on is that theory and practice don’t match, when the measurement criteria is musical quality. The interesting question is therefor why?

[Blackmorec]

14 hours ago, davide256 said:

A device with a noisier/min-spec power supply is more likely to error invoking  fault protection/error correction solutions processor overhead. Heat is also an enemy, 

 

If an Ethernet frame is bad the frame is dropped and the receiver  end at the IP layer should detect that a packet was lost and can either request data re-transmission

from the source  or for audio dither the missing data if it can't wait for re-transmission. IP layer expects packets to arrive out of sequence and if one is missing that is not

an instant re-transmit request, there is some timer that has to expire before a re-transmission request. Some audiophile player solutions

sound poorer if you don't maintain very low network jitter and errors, even with endpoint device song buffers, no certainty why. 

Thanks Davide, a very concise answer to my question…..but here’s the gotcha, and I’m honestly not trying to catch you out, just looking for an answer. 

Lets say that I replace the cheap-as-chips SMTP on the final switch before the server with a DC3, a double regulated and still one of the finest LPSs available.  Ive done this and the result is an almighty leap in sound quality. Great. So now let’s say I replace that DC3 LPS with the new kid on the block, the double regulated DC4. I’ve done this too, and the result was another mighty leap in sound quality….so do you think that the same mechanism you describe can still be responsible, given the close-to-SoTA quality of the DC3, or is there something else going on, whereby you can CLEARLY hear the quality of the LPS in the final presentation. I’d be quite happy to find out that the high level of sound improvement has nothing to do with the ACTUAL data stream, but what is going on that I can very clearly hear major improvements in presentation between different power supplies, cables, addition of anti-vibration measures etc. When I went into digital streaming, I believed that as long as the stream wasn’t error prone, things like cables and power supplies, that make major differences in analog, would make no difference in digital, yet in reality the replacement of a DC cable between the LPS and switch is clearly audible. Transformatory even.  I would love to be able to demonstrate these effects to you, so you can hear what I’m hearing. What’s more, all these affects are additive, such that a network where every component is optimized sounds way, way superior to your ordinary patch cord, an improvement that in analog would need more than one very major component upgrade to achieve.  And here’s what’s even wierder. With the very simple patch cord LAN, there’s a major difference between local and remote streamed files, but as you improve the quality of the network that gap closes, Significantly, to the point its virtually undetectable. But with ALL the upgrades and consequential SQ improvements, the remote  file SQ should, at least theoretically, overtake the local file replay, but that never happens. Why? Because as the remote file replay improves, so does the local. That a part I really don't understand but I’m far from the only person to have found this, which makes me wonder if the network improvements have nothing to do with data per-se and are more to do with other co-generated, co-transmitted effects that ripple through the network to disturb some key audio processes. 

[Blackmorec]

3 hours ago, plissken said:

 

I responded to the context of what you considered changing variables in your setup as far as hardware change up (that you would run a cable?) and that is what I meant by ZERO change to your variables in you physical layout.

 

I'm only saying we can introduce a VPN tunnel to the mix with the same copy of a song stored locally and one hosted.

 

Does that clear it up?

Ahhh yes, now I see what you meant. Sorry I got hold of the wrong end of the stick.     

[Blackmorec]

On 9/26/2021 at 5:25 PM, hopkins said:

 

 It's easy to verify that what you call "network improvements" have nothing to do with the data by doing a bit perfect test. 

I would imagine that the bits themselves are identical. Its the analog ‘fabric’ that the bits are made of that changes. From example, produce a bit stream over a network based on cheap consumer switched mode power supplies then compare it with exactly the same bit perfect stream produced using high quality linear power supplies and if your system is any good, you should hear major changes in the presentation of the music.  

Bear in mind that in a bit stream, there’s no such thing as a ‘1’or a ‘0’. All there is is an analog representation; a time bound voltage that either changes polarity (1) or doesn’t (0)