Just want to give you guys a heads up of the new trials I am doing the keep electrical things apart from galvanic point of view. This make sence. I made a mistake to sell some other stuff recently, tried to restore it but simply could not find any to a decent price.
On top of it all, that would just wind me back a bit and instead I thought, OK, why not try what has been benficial for the network; optics.
So bought this: USB 3.0 over Fiber Extender compatible with USB 2.0/1.1 – Transwan
Tried it with somewhat poor PSU’s but is surprised already from the positive point of view. This is however not as cheap as other things I try to use. Rendered me even a small customs and VAT add-on (Swedish level). But already without doing all to make DC supply as good as possible, I see some great potential in this concept, just as it turned out for the dedicated LAN optic network. The typical “extra-all” from a computer USB out is gone, just as I appreciated from the Matrix Element H USB controller card. I am adding some other things as well, and think I keep on updating this subject as it progresses, for you who are as big a nerd as I am. It seems a computer is impossible to make shut-up. Even though good PSU’s is implemented, the MB operation is full of local DC/DC converters.
Hi Torben. Nice to see others barking up the same tree…
However there is a thing that your connections are doing that may not be the intended effect. If you use common mode chokes there is a physical theory and a practical outcome. To work theoritically correct, ie being absolutely invisible to any signal being an exact copy on both leads, it means that it have to route in perfect magnetical zero point through the common mode choke.
That is like star ground something that is benign on paper and virtually impossible to realize. When routing threw a common mode choke you will inevitable inflict some differential mode inductance as well. And implemented on digital signal, that may prove an increase of jitter.
A square wave is according to Fourier, an infinite amount of sine waves. The higher the frequency the larger the damping of inductance. The highest frequencies are the one’s that build the most vertical part in raise and fall of the square wave, why adding inductance consequently as a part differential mode, may distort the time factor on rise and fall of the square wave, hence adding jitter.
The practical effect is distorting the correct specified impedance in the USB interconnect. And as the USB Audio is not a hand shake protocol, it is as fragile as SPDIF or AES3 when it comes to matching impedance. The USB cable is a bit tricky to balance, it contains both a DC power supply, that in some cases is not needed and others need to be clean. Here any filtration is benign. But the other part is a high speed square wave, that only will degrade with added inductance. If the USB cable you are showing is only signal and no dc supply is required, I would skip the inductance. IMHO.
Rather than doing a usb to sfp to usb conversion
This is my current setup - With a fitlet3 as a NAA, with Sfp connection to my Cisco SG350 switch.
USB out from the fitlet3 to an intona to dac.
Not saying one way is better than the other since I never tried yours, but it at least skips the conversion from usb to sfp to usb?
My NAA is connected to a switch from LHY Audio (SW-6). The port for the NAA is a “super-isolated” single RJ45, also disconnecting ground. The other port for the incoming LAN is a SFP module. So the last playback devices, NAA and DAC is totally isolated between network, NAA and DAC. It was earlier too, by an isolation chip on the Matrix Element H USB controller, but not the cable shield as it was an USB cable. I think the use of optocouplers on cards and disrupt the long ground connections acting well as large dipole antennas is crucial. If not optocouplers enterprise switches with ground to PE is essential to keep shields and ground as clean as possible. Not that is will stop the computer communication from working, it is made immune to these interferences, but since all roads lead to Rome (DAC), I find it necessary to do some actions like the ones we hopefully will see here. Thanks for sharing.
Some good solutions:
Torben
Yup I have a JDS synapse 3D printer version on order for my portable hqplayer setup to be used on the road. Will report back how well it performs compared to intona.
A bit of an update. The JDS synapse arrived today for my on the road hqplayer setup. Love the color and the
small form factor.
It does work fine, plug and play and immediately you can hear the difference vs directly plugged in to my dongle dac amp.
My only complaint so far it’s being too light in weight I can not manage to make it laid flat but I suppose standing up sideways is fine for now!!
Update on using the Transwan USB fiber optic extender; improvement! However not as I originally planned it, I bought a USB PCIe controller card with possibility to inject better external DC, but that card just was catastrophic. Continuous frequent dropuots and lousy SQ. But instead put the first USB dongle in a standard NAA USB3.0 MB connector on the I/O panel and viola’, everything worked great. First dongle does not appear to that be sensitive to MB USB 5 V DC. I will continue my trials here. Still second dongle, needs an external 5 V DC feed. Use the included smps and a quite new DC filter called Hades (aliexpress), consist of Li batteries and super caps at several F capacitance and very low RSE. The filter effect is substantial. Now the consolidated isolation is better than my Matrix card, resolution is slightly improved, tonal balance is as usual when removing interference, better bass control. Hence a small tonal tilt to the brighter side, but nothing bad. 3D perception in acoustic music is greatly improved. The impact and sheer size from a large concert piano is overwhelming. Over-post-processed music is what it is, hardly anything to get better except for keeping high mids in control and possible to listen to without 1 minute fatigue and shut down. There should be DNR info in meta-data from Qobuz, so that one could choose the better masters, instead of just bit rate and depth as the only and quite un-interesting info. It is quite possible to ruin music whatever container you put it in.
So far so good …
Update: JDS Synapse also works with rpi4 or rpi5 so there is no compatibility issue like certain intona models with rpi
I do not agree to this conclusion or should I say statement. Adding a ferrite bead is an unpredictable impedance match as the common mode rejection also become some (hard to specify) differential mode. Either the designer make sure the added ferrite bead stop band is far from signal pair payload (then the sentence of optimal impedance is not even applicable as it will not interfere with signal) or one end up with amplitude damping and time shift (jitter). Look at Jenving Tech’s famous USB and HDMI cables. Their HDMI was a long time almost the only one to be able to sync long distance 4K transmissions remaining solo in electrical domain. Today it is often hybrid cables with some signal transfers as optic fibre. The reason is impedance match is hard to achieve and when frequency goes up, so does potential jitter. Both capacitance and inductance should be kept very low (in terms of well within the interface tech spec) as they have profound negative effects on signal integrity, especially square waves.
But jitter does not survive in buffert stages. And with a correct implementation of buffert and local re-clock in close perimeter to DAC chip, to avoid re-enter of jitter, the signal can be fully restored and match original package check sum. There is a problem however, error correction, as the more it has to work (computing power), the more the electrical noise will raise (the basic reason for output from NAA compared to main playback and resample pc). The lower the error correction the cleaner the signal. I prefer a galvanic separation, preferably with some sort of opto-coupling, but gladly with double galvanic isolation. The shield of the USB cable is my firm beleif, must be lifted, as it inevitably with send noise between units ground. Here the ferrite bead could improve, but I’d rather disconnect entirely.
This is my personal findings besides all the pc and software stuff involved in our niche … 