This is interesting. Assuming Qobuz is the source, does it make a difference if you download your playlist to your local hard drive for playback vs streaming it?
Figured I’d ask if you’ve tried this on the IRS system.
Hi Ted,
Thankyou for the explanation. This makes sence. Ethernet seems to be the biggest problem when used in a High End Audio system. I haven’t read all the articles on this forum, but I started a test with a galvanic isolator directly into the dac. This helped a bit. The transformer in the isolator is built for a certain bandwide. The noise outside this bandwide became in trouble. Earthloops due to the ethernet connection were ended. Later I also added a second isolator on the other side of the cable. I thought in a closed loop noise due to interference could not longer be received, because the way to earth was blocked both ways. This was a second “major” step a bit charched.
After this both way galvanic isolation I started to isolate al the audio related cables with isolators on both ends. This realy opened my eyes, not knowing how big the problem of ethernet noise realy was. I think I only reached a loss of 20%.
After that I started to build linear powersupply’s for al my equipment. From version 3 a Mosfet regulated powersupply with ultra low impedance capacitors I hit the jackpot… again a bit charched. But with the linear powersupply’s with above 130db SNR or S/N the impact on low frequencies and the mid and high became better and better. With adding 4 powersupply’s I think I reached another loss of 20% noise and a profit of 20% real music.
After this step I bought an AQvox switch and replaced my old 8 port switch. This was again an eye opener…
Than I looked into a set optical converters. The specs for these optical converters were singlemode, because the stream could be dedicated. Singlemode has thinner fibers, so the light could travel in a more straight line then with multimode. I selected a maximum speed of 100mbit, because every bit gets more attention… more time… whereby high frequency noise wouldn’t get a change due to the low speed. The next parameter was low power. So the converters itself wouldn’t generate new noise on the clean side. The powersupply must be 5V so I could combine the AQvox switch with the clean side converter and use very short ethernet cables to connect them and another short cable from the AQvox switch to the dac with a galvanic isolator. The optical solution didn’t brought much with there standard switching powersupply’s, but on a good linear powersupply I reached a level I never thought it could be possible. The optical solution in combination with the AQvox switch and a high end 5 volt Farad power supply with 140db SNR or S/N brought another 60% profit. This Farad is a commercial powersupply… ultra high end.
Realy phenomenal how much can be reached by cleaning the ethernet. Tests already comfirmed that far more expensive dac’s could not deliver better sound not even when they are connected to this same clean stream. Also I was not aware how much better a far more expensive dac could be…
Hmmm, now I’m concerned and confused. I have Blue Jeans Cable (BJC) CAT 6A cables in my house; one 50’ run from the PC upstairs to the router and three 30’ runs from the router to the stereo and TV. When I ordered them I read all the information on the BJC site and there was nothing indicating I would be better off with CAT 5e or 6 for my application. In fact, the way the BJC site is worded 6A is the best performing, so I thought, why go with the lesser cable? What isn’t stated on the BJC site is if the cables are or are not shielded, so maybe there’s hope. Regardless, should I replace these cables or just be happy I have certified cables?
The 10GXS12 and 13 CAT 6Ause UNSHIELDED connectors, but utilize a special galvanically isolated iso-shield™ that is tuned to the EMI range to absorb the interference. This reduces the cost of $$$ RJ connectivity and reduces the size of the cable.
6A is CAT6 INTERNAL electricals with the ADDED 6A Alien NEXT (ANEXT) requirements and the extended bandwidth to 500 MHz allows full duplex for 10G Ethernet. You need to use special NIC cards for 10G.
If you uses a standard 1GBaseT NIC card, it will act like CAT6, and be a 2.4 GB/sec link, like 2400 series cable. It will “still” be shielded but the added ANEXT won’t be utilized.
6A confuses some because it balances INTERNAL and EXTERNAL cross talk with a 500 MHz bandwidth to run 10G. It will exceed standard CAT6 as it is still shielded, but with standard NIC cards will run 1GBaseT. The core cables specs are double the required BW, you’re just fine and future proofed too.
If you DITCH 10G capability you can use UTP 3600, 4800 and DT600e to allow extreme overhead in 1G BaseT, but with no ANEXT built-in to the design, you can’t run 6A to a full 100 meter guaranteed link. Some far shorter links will run 6A, but the added external noise limits the reach, so not a recommended application cable for 6A.
Using shields reduces the INETRNAL NEXT about 6dB, but the shields allow critical ANEXT reduction, the balanced approach and double the BW gets you 6A with iso-shields.
Galen Gareis
1 Like
@cudfoo. Do you know of anybody that has tried this set up into the Bridge II on a DS Sr?
Show me the spectral Ethernet energy plot of the data before, and after the DAC. With proper Ethernet it will be ZERO difference in BER. ZERO. Ethernet is designed to suppress the effects of noise up to 100 meters. 10G Ethernet can pull a signal that is is BELOW the zero dB ACR level! Perfectly.
NEVER use single mode fibers for short jumps as the design won’t cancel reflections off the termination points. If you do, use a mode stripper to cancel reflections. 50 micron is more than good enough. Attenuation loss is completely meaningless within the optical power budget. The optical to voltage stage will work exactly the same again, because the system is designed to ignore noise. And yes, optical devices have “noise”, interpolation too.
Fiber is SLOW. ANY fiber size, the core size does not change the Vp of the light. Light travels at the speed of the medium it is in, doped silica glass. This Iimits the speed to about 60% Vp. Copper is faster, up to 90 percent Vp. The wavelength of the light defines the BW, not the speed in the medium. The higher up you go in the electromagnetic frequency (we use wavelength for defining fiber optics since we can’t see the color) spectrum the more BW you can carry.
Removing modal dispersion in such short lengths is meaningless to bandwidths that are in the many Ghz-km for SM, theoretically infinite BW, and over 500 Mhz- km on even garden variety MM. Modern VCSEL MM and SM laser technology significantly reduces chromatic dispersion, too.
Going from 100 dB to 130 dB of (supposed) SN isolation is factor of 1000 and very, very, very few test units can test at this level. Show me the data.
If it can sound better, it has to realistically test, measure or calculate better. I see a lot of well meaining talk but with scant evidence of repeatable and verifiable data to support any of it. Better is indeed better, and I see no reason to limit that, but it DOES have to be proven with accepted scientific method.
What circuits are measurably Being impacted by The noise? How? And, optical states are also full of noise, galvanically isolated or not…they throw off noise. Is adding a galvanic circuit WORSE than not using it? I see no spectral noise plots defining the advantages, and even WHERE it is an advantage. If we can’t fully define the circuit that is most non linear to noise, how do we even know we are doing anytning?
To make better components we need to witness the walk, not listen to the talk. In God we trust, all else being data. The things we can measure have to be factually correct or we are punching holes in vodoo dolls.
The numbers don’t lie. The impact those numbers have on performance will audibly vary person to person, absolutely. But the better numbers will remain the same, and be verifiably the same forever. THAT is true improvement.
Galen Gareis
Hi Galen Gareis,
I wrote my conclusions and test results on the forum to help people and not to prove a point. I think ethernet noise influences the sound of audio systems more then one would expect. I think and experienced it’s the most important step one could make to improve the sound of an audiosystem.
Wijnand
We need to get past thinking and to real world data. I want to think a lot of things seem “plausible” but with no data to back it up? There is no conclusion other than the politics of who’s side you’re.
I don’t do sides. I look for REAL answers and repeatable tests that show true improvements, THEN and only then can we expect real world advances. If we hang all sorts of unproven artifacts all over our equipment what exactly are we REALLY doing? More “I think” components aren’t doing DAMAGE over “I think” it’s working? Why do we side with unproven aspects actually working verses doing damage? Collectively you can be worse off.
I fully understand wanting stuff to work. But DOES it work?
Galen
Sorry I don’t have to get past thinking. I already reached so much with thinking. I like to discuss the options I reached and some already implemented in their systems. The improvements were dramaticly. The change was so dramaticly that is was not in the league of a change of a Dac or Pre-amp, but more like a change of loudspeakers.
It’s just a tip! Which maybe you or someone else could use.
What changed where? Really?
Galen
The changes in SQ due to better network with lower noise levels. Due to the optical converters and other galvanic isolaters and taking out every switching powersupply, by replacing it for high quality linear powersupply’s. Off course there will be always switching electronic’s like the optical converters, but the noise that will be generated by this optical converter is nothing compared by a Nass, Router or PC
I agree/don’t disagree with either of you based on my limited experience and even more limited technical knowledge in this area.
But…what a difficult thing to parse!
This is a thing that comes up (at least in my thinking) more and more often hereabouts.
Yes you are right. It is very difficult indeed. It’s realy not to make a point, but to start a discussion that can help people improve their systems. For me it’s a fact for others to believe or to pass. Both are good 
But we don’t KNOW what got better, or what parts really changed the SQ. It isn’t defined. This is like making spaghetti sauce where you know all the ingredients we throw in, but don’t have a clue which ones are effecting the taste, and which ones are immaterial.
To make a TRUE changes, it has to be defined in and of itself to cause and effect. A measurable difference, somehow. The overall impact on SQ, we all know is going to be listener dependent, but the measured CHANGE, that stays the same.
Digital does NOT have the ability to CHANGE the PCM encoding, that stays fixed at the front end. To add coloration, which is what SQ is, it has to be a CONSISTENT and UNIFORM real-time change to the ANOLOG signal. This distortion is either when it is being fed into the system, or added as analog distortion after it goes analog pulling the signal out of the system.
Ethernet is super good at moving data from A to B with no errors. Noise is ignored, not so with analog. The filters we use for AD and DA are where we make approximations and can add external superimposed SQ colorations.
NOISE is not steady state, so the problem is dynamic over time, if it is indeed changing SQ. We should be able to create a worst case condition that shows that the DA is more stable making the approximations if we do “this”, and this does “that” to account for what we hear.
Hearing a difference doesn’t mean you KNOW anything. Copper sounds different, but I don’t KNOW exactly WHY, for example. The buck stops when we have an answer why, not the effect.
Adding FO links adds EMI/RFI noise of it’s own. Why is it not considered WORSE than a straight copper interface? I see no proof of the trade-off. Saying we have a galvanic isolation isn’t an improvement is SQ, it is just a physical electrical separation between devices at a DC level. RF still remains.
Galen
I can go left or right and arrive In NY city from Ohio. One is a guess, the other is KNOWING how to get there. In audio we have an awful lot of heading west from Ohio and once we arrive in NY feel we have the answer when we have not a clue.
I’m all for improvements but I have seen ZERO evidence of measured technical superiority of anything offered yet. I’m ignoring the ability to hear the changes, which if repeatable and better, remain so.
Galen
1 Like
Thanks for your reaction! I can explain If you like. I went left, right, above and under it actualy was a journey of 10 years. Now I have a solution and knowledge that I can share.
Wijnand
And yet people sell cables with different types of copper… evidence be damned.
1 Like
That exactly why we give you the choice to not buy it. There is no proof outside of resistivity that can be electromagnetically verified on copper.
The WHY is what advances the art, as it gives you more design and control decisions.
Galen
Not necessarily. I have tested at least 25 different types of cables. Copper, silver, shielded, unshielded, both sides shield connection, single side shield connection, powered shields etc, etc. At the end I went back to standard cat6 cabling, because many high grade cables are not neutral. It’s better to isolate a cable at both sides, but make sure the signal does not get weaken to much. You can prevent this by good powersupply’s, switches and converters. Now the cable is isolated there isn’t a way for noise to travel. Off course the noise in a certain bandwide that travels already with the stream and came past the galvanic isolators (transients) can be reduced a bit by better cabling, but the cable won’t be influenced much by magnetic fields interference anylonger, because there’s no source or ground, noise can travel to. Only the closed loop between isolating transformers.
Still cabling influences the sound by a nuance. I would say for 5% maximum. The powersupply’s for 50% the glass converters for 35% and the isolators together for 10%.
I get the message stop with this discussion and I will. Thank you for reading!