Higher End DACs

Djones there is no level matching when doing a digital to digital comparison.

All tests are done under the exact same conditions.

In the case of the Laptop vs the Innous server, 

1: Play Music track 1 min interval, mark passage duration on Roon
2: Disconnect or reconnect either the Laptop or Server.
3: Play the exact same track from the beginning.
4: Note the differences.
5: Repeat the same test several times to verify that the conclusions are indeed correct.

So far in every test of a non dedicated server, ie Mac Mini, Desktop computer or similar device the dedicated servers did indeed produce a far more analog like presentation.

Dave and Troy
Audio Doctor NJ dealer of a lot of digital

False dilemma fallacy, EITHER all these people hear it so the device really works OR it’s a conspiracy. There could be a third choice perhaps your tests are not controlled for biases, maybe you didn't match the sound level and louder was considered better. 

but when so many people hear the same thing then either there must be a huge server conspiracy or these devices really do affect the sound quality.

Hire that man!

This is an interesting discussion, that we have a tremendous amount of experience in. The entire argument, that "bits is bits" or that digital transport devices don’t or can’t make an audible difference because a dac has re-clocking or anti jitter circuitry, bears out that in most tests that digital transports can, and do, make a very audible and noticable difference in sound quality.

We have also seen these improvements not just in dacs but in streaming dacs that have a direct path to the Ethernet as well.
Our history goes back to using a standard computer via a USB cable running Jriver to play stored music into our dacs.
First we were using Mac Mini’s because of their small size and ability to run quietly and easily fit onto our racks.

We got our first dedicated server when the Aurender N100H hit the market in 2015. The Aurender was a fantastic product, well built, beautiful, easy to use, headless, and boy did it sound better than both the Mac Mini and modified Mac Minis that we tried.

We took our N100H over to clients houses and they easily heard a wider sound stage, greater dynamic range and an overall improvement in sound quality over the non dedicated servers.

Then we transitioned to a Baetis Reference server a few years later and that much more expensive server indeed did sound even better than the less expensive servers.

We played with the original Nova Physics Memory Player, Naim Unitisevers, the Nad and a few other devices.

The more expensive servers did really sound better. Than a few years later we were introduced to the Innous line of servers and the Zenith MK II started to give the Baetis a run for its money yet cost so much less.
Then the Innous Zenith MK III came out and it was far superior to the older Zenith.

Recently we did a shootout of the Innous Statement, vs Baetis Reference and the Laufer Techik Memory Player. In the same reference system and the difference between all three servers was again audible and each server did indeed have a sonic flavor, and a distinctly different set of musical qualities.

Two recent tests will also be of interest:

Test One: Innous Statement in a completely different room feeding a Krell K300i integrated amplifier over Ethernet vs Innous Zenith also feeding the integrated amplifier via both from a filtered Ethernet port from the Zenith and over a USB from the Zenith.

Long story short the much more expensive server in the other room sounded better than the direct Ethernet connection from the Zenith feeding it in its same room. So in this test both test data was feed to the dac via Ethernet one server was in a different room, but was a far more expensive server vs a direct Ethernet connection to the integrated amp, this proves that a better data stream does indeed sound better.

There are two possible explanations, one that remote server was outputting a cleaner data signal that got to the dac even over the Ethernet port and the resulting network connections, or the fact that the server was in another room made the improvement, however the direct Ethernet connection from the Zenith should have sounded better because the dac was getting a direct, filtered and cleaned up signal

Another test Naim NDX 2 server running Tidal directly from the network and its own Naim app, vs the filtered direct Ethernet port in the Zenith then playing Tidal via Roon into the Naim.

Again logic should dictate that the direct network connection of the Naim streamer should produce the best signal or at best there should be no audible differences between both of these scenarios, a Tidal stream should be identical right?

Again the difference in sound quality was very noticeable, the server produced a more relaxed sound, a wider and deeper sound stage, and in general the sound was more engaging through the server.
We recently compared a standard high end Sony laptop over USB into a Bricasti M3 dac and then played the same music over Roon laptop vs the Zenith server, again the sever produced a much more musical sound.

The client heard the difference, the wife heard the difference and we heard the difference.
For those that are still incredulous way too many people will attest that they can hear these differences, it is of course possible to deny that these differences do occur, but when so many people hear the same thing then either there must be a huge server conspiracy or these devices really do affect the sound quality.

We have seen these difference manifest themselves with many uber expensive dacs that all employ re clocking, femto clocks and other anti jitter circuits, unless you consider Emm Labs, Esoteric, Light Harmonic and T+A’s engineering staff to not know what they are doing?

Could it really be that there are issues that the data can be affected by noise, and other factors that occur with both direct network connections?

A regular PC or Mac are not designed for the sole purpose of making the purest data package, compare the build quality of regular PC to a dedicated server running a quiet CPU, custom software designed to keep processes from using CPU cycles, circuit boards made to minimize nose and RFI and EMI intrusion, custom sound cards,
re-clocked, filtered USB and Ethernet connections.

Could it be that garbage in and garbage out really does manifest itself with a data stream and perhaps the more pristine data really does enable the dac to work with a better data packet which results with cleaner sound?

All we can attest to is that we have many, many servers and dacs, and we have yet to find a setup where the server didn’t elevate the system’s sound quality dramatically.

Dave and Troy
Audiodoctor NJ

You must have a lousy internet provider soundtest to have enough latency to interfere with your music streaming. I can ping Baidu in China from the southeast US  from a wireless tablet between 79ms to 83ms. If your home network has enough Lan activities causing noise get yourself a $50 netgear managed switch and put your streamer on a VLAN giving it priority. I wouldn't worry to much with latency unless you start getting dropouts and yes it's good to have an open mind I try to learn as much as I can but as the saying goes try not to let your mind be so open your brains fall out. 

With a server, we get so many dingbats and ridicule for pulling the stops out, but we are not complaining as the Extreme can sing for itself ;-)

there are infinitely more so-called digital experts than analog ones. everyone has a computer. the odds of stumbling over one are strong. especially the one’s who already know everything.

it requires an open mind and open ears to move the reference forward.

I have to limit my comments to the discussion above to why the Extreme has 48 gigs of RAM. 

When one installs the maximum number of RAM sticks to utilize all of the available RAM data channels you can hear an audible boost in SQ.  We are pretty confident its the lowering of thread execution latency when all RAM data channels are being used that delivers the audible uptick.  The smallest size RAM stick you can buy today is 4GB.  The Extreme mobo has 12 RAM data channels, 6 for each CPU so that's why the Extreme ends up with 48 GB of RAM.

What's interesting is when turntable manufacturers go to extreme lengths to deliver improvements in sound, it's pretty much universal accolades

With a server, we get so many dingbats and ridicule for pulling the stops out, but we are not complaining as the Extreme can sing for itself ;-)

I'm going to start my own thread on RF isolation of DACs. It's a topic whose time has come. It good to be pragmatic but it's also good to accept new ideas. Understanding is the first step toward a full resolution. 

Glass is more transparent to RF but photons can travel through solid building structures, too. Like a knife through butter 🧈  Elevators (metal) not so much. That’s why reception improves as you move toward the window in many cases. Depends on where the signal is coming from, etc. Sometimes 24” can make all the difference.

Why does my cell phone work in a windowless room ????

I didn’t find where the White Paper discussed RF coming in through the windows. Maybe I missed it. But RF comes in from outside, all kinds of it, where it can easily migrate into unused wall outlets, unused input and output jacks of electronics, holes in electronics, and into unshielded cabling. The entire room is lit up like a Christmas tree. 🎄 As I stated earlier the first line of defense should be the windows. And the unused wall outlets the second line of defense. 

One does not normally start with a conclusion, they start with a premise, "The audible effects caused by RF noise are difficult to objectively measure yet are simultaneously subtle and obvious. Suffice it to say that they manifest as a reduction in musical transparency and increasedlistening fatigue. " 


Curious how you were able to listen to headphones/speakers which would transmit RF back to your DAC (you claim direct connection, no amplifier).

In your white paper, you use a log-periodic antenna rated for 1Ghz-18Ghz (or 1Ghz -2Ghz model dependent). That will be great to cover intentional radiators, but most RF energy from unintentional sources is well below that. Testing below 1GHz would be far more useful in most environments. Curiously you show results down to 100KHz for which your antenna would be poor. What you identify as standard radio communications and microwave radio communications in your "Rural Isolation Ambient Zero Baseline" are predominantly VHF and UHF television (UHF = 470-806). (p.s. Your sweep times are way too short for high quality measurements).


Of course, shielding the outside of the DAC doesn't negate likely the largest unintentional RF components (by intensity), namely components inside the DAC. Of course, most DACs do seem to have metal cases which are effectively Faraday cages, so sensitive circuitry with the exception of cable entry would be covered.

W.R.T. claims of "Sounds Better", there is no discussion of the test protocol, the number of testers, etc. so I guess we just "trust you"?  A future test is identified as "Double blind listening tests with a variety of subjects/music"   I would think proving a problem exists would be the first thing to do?

I just knew that box of PIM flanges I bought for -210 dbc performance would come in handy....

Sorry, I only got as far as the second sentence, “Magnetic transducers (loudspeakers) get perturbed to produce hums and rumbles in sympathy with RF frequencies in the audible range,”  before I stopped reading. RF frequencies are not in the audible range.  I bet you didn’t know this forum was peer reviewed, did you? 😬 Shut the cave door and back to pigmy country!

OK, I understand dmance now he's selling a $4000 box that eliminates RF that only bats could hear. 

The usual scenery from a prime seat on the doppelgänger express...

heaudio123

the link to my whitepaper

i'll have my isolation units out to influential users over the course of the year. more ears and more voices will change minds.

Just for the record the OP was asking about DACs and internet streaming since he said he didn't even have CDs. Why in the world anyone would need a computer that could run SQL server for a small company and yes I have personal experience is actually outside the point of the thread. I said the computer mentioned looks to be a great product but IMO is a bit much for simply streaming over the internet into a decent streamer/DAC. I imagine if he is interested in it he'll look into it. 

How do you know the sun is hot? Read it somewhere or have actual experience? 

We can't address an argument, we refuse to consider our own bias, so we will insult and not think. Perhaps our bias is at play, or perhaps we have ignored other variables that could create perceived differences?  That is easier than putting thought into what is really happening. It is easy to link to pop-science articles than it is to understand what they are discussing and/or research if there is validity to their claims. It is much easier to claim we hear a difference than it is to validate that claim. It is easy to claim a difference is because of A, then to validate is it A.

I don't read this stuff in a book, but I may write the book (or the articles), and they are based on unbiased testing, but you don't need "testing" to know best practices w.r.t. keeping jitter low, reducing RF emissions, reducing vibration sensitivity, or what impacts jitter most .... the hifi audio reproduction industry is not the bleeding edge of technology.  There are many industries where verified low jitter, verified low noise, verified signal quality are important.

Funny stuff. 😂 Is he.... gulp, Ethan Winer? 😳

interesting how many opinions we have here about what can’t have an effect on digital audio, servers, reading files, and streaming.

you read this stuff in a book? or actually personally tested these things?

just for the record.

Again agree with djones51, w.r.t. -200db, that no one can hear it, and when the light of day is shined on the claim (properly administered test), the claimed ability to detect this disappears.

dmance, the only paper i could see that you referenced was on human hearing and fourier uncertainty principle which was sort of sham, hence why I don't believe it was ever published in a peer reviewed journal:  https://nrc-publications.canada.ca/eng/view/accepted/?id=ca9f5091-2658-449e-b08e-b071712acbf0
That paper was a joke, but unfortunately there are too many science "pump" sites and audio sites are always desperate for content.

Were you referencing another paper? 

 -200dB ? Noone hears that. I'm not saying  the computer mentioned above isn't a good product IMO it's a bit much for simply streaming over the internet. For running roon or other DB's I am sure they don't  even come close to stressing that computer. 

@dmance "the magic is [going] away"  Any solutions for your  tone deafness? 

@cleeds 
I have a signal hound BB60 and have done the tests (see my whitepaper). I have done the checks for alternative explanations. Repeated the tests. It's quite clear that a DAC fully isolated (say runing off a battery with a source chain buried in a Faraday cage and only optical I/O) sounds better ...and by that I mean exposes increased transparency. Open the lid of the cage or get lazy with what could be a source of RF...and the magic goes away.

And I'll have to say again. If you are asking questions about measurements to point the finger at a causal issue ...no test gear will suffice to find those for you.. the effects are real but must be down below 200dB.  There is something remarkable going on with our ear/brain that lets us hear these things.

The problem is that everything is affected by RF. I’m not sure that’s saying anything new. But that shouldn’t be terribly surprising since the signal itself is RF, no?

dmance

I considered the problem as a DAC being an antenna. I analyzed the DAC outputs while the DAC and device-under-test was in a Faraday cage. So it’s quite clear that when a generator of RF is even just proximate to a DAC (no connection) the outputs of a DAC are affected. This means the energy gets inside the DAC ... None of this is measurable directly ... But we’re talking High End DACs here...where these effects are plainly audible.

How did you "analyze" the DAC output to conclude its outputs were "affected?" How do you account for the fact that you consider the affects "plainly audible," but can’t show that with measurements?

@dmance  Any reason why you are unable or unwilling to start a thread focused on your own product?  

There is value in what's been recently discussed. There's also value in having focused discussions that are on topic and are not disjointed because of thread drift. 

@geoffkait
I considered the problem as a DAC being an antenna. I analyzed the DAC outputs while the DAC and device-under-test was in a Faraday cage. So it’s quite clear that when a generator of RF is even just proximate to a DAC (no connection) the outputs of a DAC are affected. This means the energy gets inside the DAC. And I have the corollary that when the energy bis there (even a high MHz) the sound is less transparent. So the DAC’s designed performance is affected by:
- the xtal clocking the output is perturbed
- the multiple harmonics at MHz affect the noise floor in the audio band
- reference voltage levels disturbances

None of this is measurable directly. The audio band (to 21khz in my case) shows no added jitter or noise floor changes to 150dB. Even in my conversation with Rob Watts, he says he can see nothing to 190dB. But we’re talking High End DACs here...where these effects are plainly audible.

Or do what I do and stop the RF at it’s source (coming in through the windows from outside) and turn the entire building into one giant Faraday cage. Problem solved! 🤗 it’s not that difficult. Furthermore, while not widely known, tiny little bowl resonators are effective in normalizing pressure zones in the room for acoustic wavelengths AND RF wavelengths. They’re Two! Two mints in one! 👯‍♀️

Very interesting thread....Thanks...

RF is the elephant in the room. The invisible elephant that changes shape and size with every digital tweak ...but still remains to affect the DAC. It's all because we have clocked digital devices proximate to the highly sensitive D/A.

Ahhh...but what if you could reduce RF to the minimum. So only room ambient remains?  Easy: put your entire digital chain inside a RF shielded enclosure (like my RF-STOP box). Only optical out; only optical in; and well isolated AC input.
It means audiophiles have to reconsider how they deal with high end audio tweaks. No single point product fixes the RF issue. No single manufacturer is dealing with all products RF emissions.
But today we can put our whole works into a Faraday box, follow a few simple principles for I/O and the elephant goes away.

soundtest3

A major high end DAC manufacturer actually recommends removing their Renderer module for an audible sonic uptick ...

Which manufacturer is that?

The RF "sauce" is not latency influenced. How much RF is bandwidth influenced. If you are making the router/switch work hard (latency has little impact on that), then the RF signature goes up.

I am trying my best not to make my posts a sales pitch, but to provide insights about we have observed and what we think is going on.

re your first point, the two modern choices of getting data to the DAC are a) Renderer in the DAC as data entry channel and b ) via USB to a USB receiver 

of the top line DAC’s we have seen, and tried out and reports we have received, USB has a nose ahead of IP in the SQ race.

A major high end DAC manufacturer actually recommends removing their Renderer module for an audible sonic uptick

it is a trade off between noise coming from LAN activities against noise coming from the USB receiver

To your second point, we have found that latency maters both for OS activities in the CPU and for interestingly for the LAN network and the internet backbone behind it.

For a long long time, we could not understand why real time streaming sounded noticeably better in the Netherlands than in the US.  The SQ delta between onboard storage and streaming in the Netherlands was quite small, whereas the gap at some customer sites in the US was huge.  Well guess what, a lower the latency of the network connection to the streaming services server resulted in noticeably better sound.  In the Netherlands there has been for quite a while a glass backbone to the IP infrastructure with very low latency and that was the SQ edge In Holland

we can even hear the SQ delta between streaming and onboard storage get smaller at night as the latency of the connection reduces because of lower traffic.

This being said you can still get very good and enjoyable sound quality from streaming even when the difference is quite noticeable.  We don’t come across many installations where the network latency is so high that it results in a sound we don’t want to listen to and can’t enjoy

The zero and ones are exactly the same, but the RF sauce that comes with it Is latency influenced and sounds different

I hope the above is helpful in describing the RF landscape that effects Audiophile sound reproduction

What’s this a sales pitch for a over priced computer to stream Qobuz? If you really want to keep all that noise down have the server in a different room and connect through ethernet problem solved. CPU latency isn’t a problem streaming over the internet neither is execution time they use ARM chips for the most part not x86. 

To get to the specifics of the Extreme and why its very good at streaming, is the the optical SFP port which allows a somewhat noise cleaner LAN feed than copper.

With Qobuz the sonics are extremely enjoyable !

What works, what’s cost effective, and what is today’s best possible sound quality for lossless streaming spans a huge spectrum of hardware complexity, software optimizations, weight and cost. The zero’s and ones are exactly the same be it a system built around Raspberry Pie, a SOC, a i7 or dual scalable Xeon CPU’s with 20 cores.

What is different is the RF emissions collection, and how it effects the sound quality. Here are my takeaways, observed over the last 5 years, and independently verified by intense DIY professionals like Romaz.

It has long been observed that lower software execution latency helps improve SQ, and we find it is still very much the case today. Lower latency shortens the CPU active time, and CPU associated RF emissions. What high powered CPU’s bring to the table are faster thread execution times, less CPU busy time, and that noticeably increases SQ to levels that slower and less powerful CPU’s cannot match.

RF emissions in the Kilo to Gigahertz  frequency bands which are emitted by the various clocks in digital gear can be attenuated by good shielding and grounding but what still gets through and then infiltrates in to the analog EM signal amplification environment results in an easy to identify audible signature. When carefully managed and shaped, these RF emissions can be SQ enhancing.

Digital equipment is as sensitive to mechanical vibrations as analog equipment. All crystal clock oscillators are vibration sensitive to an order of magnitude greater than their steady state frequency drift. Clock oscillators’ output signal voltages are also quite sensitive to power supply fluctuations. Overbuilt power supplies make their sonic presence heard in digital equipment.

Playing Streamed files does not require much CPU power, but the quantity of IP traffic going up and down comes with its SQ unhelpful sonic signature and the reason why the sonics of streaming are behind well implemented PCIe storage and RAM playback.

So while the size, complexity and cost of a high powered music server might seem to be over the top, the level of sonics one of these servers can deliver is not snake oil                                                                                                                                                                                                                                                                                                        

Difficult to overstate the improvement in sound quality of the Extreme - as a server or streamer. I debated for awhile whether to invest in a higher end DAC or the Extreme with my current ‘mid priced’ DAC. I am glad I went with the latter. Most significant component upgrade I have experienced. 

@mikelavigne   Mike, thank you so much for your response and the details you shared. As always, your contributions are informative and helpful.

When I'm ready, I'll reach out to Emile to see if he can configure the SGM for Roon only functionality. I'm asking Jord Groen of Pink Faun the same. However, losing the Roon Rendering function that both have devoted part of their design / build efforts to is a concern for me. Those are tradeoffs I need to resolve for myself.

Meanwhile I have mid level servers from Wolf Audio and Innuos inbound. They should offer insight into the level of difference(s) versus my current Roon server. These comparisons will also inform whether a major step up is warranted.

Streaming from the internet is not a CPU or memory intense application it can be done with a raspberry pi. I stream high resolution and don't even use a computer simply a streaming device with a Qualcomm chip in it. The database with all the metadata needs the memory and CPU power. If you don't use a server there is not really a need for a computer and actually I could stream from a NAS with what I use but the NAS is a type of computer. 

i’m not in a position to debate the technical side of this. but my understanding is that when it comes to digital audio, local files, NAS sourced files, or streaming files, all have the same needs from a CPU. and our performance expectations should be the same too.

sure; you won’t find streaming 352-24 or 4xdsd files; but that is a small part of file listening anyway. my streaming includes plenty of high rez.

Streaming something like Tidal or Qobuz over the internet is one of the least CPU intense things you can do. Take the server away with the need for intense data base management and you could knock about 70 pounds off that 90 pound thing. It's  the same with roon if you have very few files for roon to manage you can run it on low end computer the database is what eats up memory and processor.