I am very glad for the intense debate over this subject who just happens to be one of my main interests and focus in the HiFi-rig. Myself and two other guys, one happy do it yourself engineer, and one studiotechnician with several recordings on Tidal and also an M.Sc El. Eng. who consider all gear basivcally as crappy and there is no hardwae in his studio or listening room that has not been modified. He does not approve of many power supply arrangements at all, and is constantly pointing out insufficient signal and ground designs. After opening the dCS studio DAC, he filled an A4 papper with need to correct bullets after just a brief inspection.
The way the work is highly contagious and some years ago I started as well, mostly with outer devices to assist inner critical areas, but soon went on (in) and started to rebuild substandard voltage regulators, insufficient local buffert elcaps chaning to 105 degrees and zobel on each and every elcap. Low internal resistance, low inductans was highest priority, hence a must to cure the possible resonances.
OK, so I beleive gaining from shielding the card. The cover and backside of layered pcb forms a quite good seal for the airborna noise that semiconductive components like diodes and thyristors radiates at the ignition (opening). Further it gives transients in the leads. By sealing the USB card from airborn noise, by galvanically isolate the trafic I/O to the PCIe port, I hope to minimize noise superimposed to the bitstream and also switch of the 5 VDC if RME allows for it, even though it is from a decent linear supply, if it is not needed, why provide it? The internal switched PC psu does not even have to be connected to this specific card. The others seem to recommend it for some reason even if a linear is also connected. I am going to take it as far as I can to not compromize signal integrity to the DAC. Since the DSD Direct is right on to the DAC chip, this should make sense, and the price is not the end of the world, compared to buying devices in the high-end segment.
It is a bit strange that RME does not offer a pulse trafo input of the USB, bit it does not isolate the USB cable shield noise either in a easy to discover way, as direct chassis connection instead as the similar priced Teac DACās does. The cable shield noise does not belong on the pcb at all, in my oponion. I hope to soon have this installed and be able write a summary of my findings ā¦ 
I always figured those cost too much to be well designed. 
Audibility is not really an excuse, rather the ultimate measure. Yes, intermodulation distortion can make out of band noise audible (which is exactly why it matters), and it depends on the analog chain, but thereās no excuse for any DAC to leak noise in the MHz.
They are, but not perfect up to his standards. They are used all over the world as is, but there is always somethings that will increase dynamics, reduce dc-shift, remove hysterisis in a balanced output transformer or further enjoy the subtle nuances in his ears. He simply cannot withstand the challange. My closer friend, the DIY-guy, have reboilt the Yamaha NS-1000 monitors his playing, so there is only the outer appearence that resembles to the original. The completely revised cross-over, is among the finest crafted passive filter soundwise i have ever heard. Not to mention reinforcing the cabinet. After his adventures in this area, I am tempted to say, most speakers sound ok, despite their filters. When a rig becomes as silent as those e.g. of these guys, it is a joy to scritinize filters from Jussi. I have listened to hifi gear that could not play the difference. But the most important feature of the rig, silence when supposed to, is the one benefit that allows me to enjoy music the most. When listening to one-take orchestra recordings it is inevitable that some make some noise, cough, touch something making tiny noise. Those are the proofs of real musicians. Someone is alive in the studio. If those tiny irregularities does not sound, the playback system is not silent enough. When almost feeling the air vent system of the recording room, you starting to get somewhere ā¦
Yes, my point exactlyā¦
A more flexible option is USB over LAN (can connect to switches) by Icron
The REX unit (the unit that connects to the DAC) is powered by 5Vdc external PSU, so you can go fancy linear PSU or again use powerbank.
This allows you to by-pass NAA and just use HQP ALSA output. Not a huge advantage over UPBoard Gateway running NAA OS, other than never needing to do NAA OS updates ever again.
@dabassgoesboomboom
@jussi_laako
@anyone
I am working my way around the best ground practice with an arbitrary USB but off course most focus on USB card. It is not quite as clear how to isolate the inevitable shield noise pickup. I do not want it landed in the DAC. I can not see an efficient ground noise drainage system. In fact I see potential problems. The DAC have no ground connection. But it does not mean it will not load up ground noise and ripple oscillating in phase with the both shielded interconnect systems; USB and XLR analogue output. Since both of the shield connectors are soldered on to the pcb, there is possible stray in both digital and analogue section. Will have to make some re-wiring, I think, to spare the DAC un-necessary HF leakage in pcb ground and stray to the power and signal.
If the system overall has nothing grounded, you can ground the DAC via RCA connector for example, if needed.
Depends on overall system layout whether you need to. Each system will be different.
Please check the RME manual: there are several notes on grounding.
Seems you replied to my post but my system grounding is well sorted 
The manual is for cowards ā¦ 
I will go look where I put it and read. Now I at least know there is an AKM DAC, I do not know if there is a great difference between 4490 and 4493, but 4490EQ is what it is.
I am very glad to se that they have not connected the poor chassis leads to pin 1 internally in the connector. It is a ā ā ā ā connector as it is, but I guess I cannot growl to much given the DAC price. Anyhow, that means that pin 1 could be separated from the chassis ground, good. Next step will be workshop to see if RME use negative pole in psu feed to ground the chassi, or pcb ground. One could replace DC contact to a XLR in both linear PSU and RME, to utilize the shield for grounding the RME chassis, It is so close to the USB port, that I could isolate the USB shield from the connector and instead solder a lead from USB cable shield directly to new XLR DC connector ground. Hmmm Look like a few hours in the workshop again, happy me ā¦
Thereās quite a bit of difference, but my two ADI-2 Proās have 4490, so my measurement results should match with your device.
Mainly an issue with unbalanced RCA connections:
Frequently Asked Questions & Answers
Will be interesting if you see an balanced XLR measurement improvement with the D90SE
The thing with balanced XLR (analogue) is that it is a beautiful technical feature, that is almost 100% wrongly implemented.
Manufacturers/designers keep konnecting pin 1, together with shield, to the pcb inside uploading the shield noise.
The same goes for USB. It does not matter if the have a pulstransformer, since it isnāt the bitstream that at stake, it is the shield noise here too.
It seems that it is taken out of the equation to be very careful of the ground integrity to offer all circuitry a stable and clean 0 reference. The sh*t load must be dumped to chassis ground. And here I cannot understand why the use of external psu cannot include connection to ācoldā ground for the chassi.
As I have said numerous times in other frums, the computer build (230 VAC) with ground to chassis and the I/O plate connecting all port shields to chassis is a class example of how to do it, but in HiFi, the donāt. I have seen it in one rather inexpensive DAC, Teac UD-501. The port shield includs a flange do screw to the chassi packplate and the entire USN input is on a separate card, and finally, it is grounded through protective earth. All the right moves to keep USB silent.
USB doesnāt have transformers, but Ethernet does. And common issue among audiophiles is to use shielded ethernet cables instead of standard U/UTP cables. And that spoils the galvanic isolation it would otherwise have by connecting device grounds through the cable shieldā¦
Best option however is of course to use optical ethernet. One of the reasons for NAA to exist too. Because unlike Toslink, optical Ethernet doesnāt have bandwidth limitations for audio. You can easily do 10 Gbps over really long cable links.
Another possibility is of course WiFi, providing you with complete air gap. But keeping RF side clean from the analog side is fairly complex affair.
There is lot of audio equipment too that is not grounded, with two-pin mains connector. These shouldnāt be connected to ground through their connections to other equipment.
One good example from past was my living room system. Where thereās a TV connected to cable network, TV also has the RF cable shield connected to itās ground, floating power connection like all (?) TVās have with two pin mains connector. And similar AVR connected through HDMI (shield connects devices together). Connecting computer to this system was a disaster because that caused ground connection through HDMI and 3-pin mains connection. I measured 110V AC difference between the cable network shield and mains earth pin. Only solution to this system was to use ground breaker module in the cable network coax.
I have UD-501 too, and NT-503 as well.
But I prefer not have such approach, but I want USB totally isolated from the analog side. So you cannot connect both USB and analog side grounds to the chassis. Instead only analog side should be connected to chassis, and USB should be totally galvanically isolated from the actual converter side. There are number of DACs that do this, with opto-isolators between.
For DACs like ADI-2 and UD-501 devices like Intona should work fairly well, but it is not completely optimal since you still have another device connected.
It is not an issue if DAC is designed correctly so that thereās no galvanic connection between USB and rest of the converter.
I have a Matrix Audio Element H USB card. Doesnāt make a bit of difference to me. No better than the mother board USB output regardless of powering the card from internal power or from a battery. Mother board USB, Element H USB, raspi USB, S/PDIF all sounded equally good through my DAC.
I have the element H and while it wasnāt a huge upgrade. It did improve the soundstage and detail (inc the lower end). In saying that. I have since added a linear PSU for it and it then makes a bigger difference.
My setup is a W11 Pro PC, Tubulus Concentus USB cable to a Matrix X-SPDIF and then a Tubulus Concentus I2S cable into a PS Audio Directstream DAC.
Things Iāve noticed more so, vocal and instruments are more separated, lower bottom end. Really wide and open sound stage. Picking up breaths and violins sound more natural and āliveā if that makes sense.
For the price, the Element H is cheap as chips. But does need a linear PSU and quality cables. Prior to the Element H I got rid of AQ Coffee cables for the Tubulus cables and that made a HUGE difference in my system. I didnāt know the Element H existed until I checked to see if there was a newer X-SPDIF
Hope this helps
It does. Thank you. I will go to further excesses when it come to cabling and separating ground from entering both DAC and USB card, and the cable I have chosen is so exactly matching the USB characteristic impedance requirement under the USB standards, so if every other device i designed accordingly , there will be a match in heaven ā¦ 
Thanks for your reflections. I do not agree to some of them, though. And I donāt know if we may or should turn the thread into electrical design philosophies? I know for one, I startedā¦
The two pin psu HiFi devices that should not connect to ground will inevitably do so, irrespective if it should or should not. My DAC offer a two pin psu, but also a balanced XLR interface. The XLR interface is a paradox, because according to white book design, the shield should connect to chassis ground, and there is no chassis worth connecting to, since it is not grounded. But still the card itself is draining by screw fasteners and electric contact to chassis. That in my book is a reason for uploading not only shield noise, but chassis pickup, as there is no drainage. It would be far better if there is a ground loop separate from electronics, where starting at the USB card, the USB shield drain wire is free to connect to the computer outside of the chassis. The PC chassis is ground. In the other DAC end the USB drainwire is again free and connected on the outside to the chassis, giving a ground to the DAC not infecting circuitry. This applies also for the XLR balanced, there is no need to connect shield drain wire to pin 1, simply connect only to chassis. The pin 2 and 3 are balanced and require no ground, it is automatically harmonized, cleaned and centered. The only āifā is to check that the shield drain wire is not wrongly implementet in the power amp. If correct, the drainwire could connect only to chassis, and pin 1 be left empty. With this scheme there will be no landing of shield drainage noise to any fragile spot, in any device in the chain, one could even let the two pole psu connector be as is, since the DAC is drained to chassis potential two ways and none of them will incorporate the circuitry or pcb grounds. One could question if there is a chance of stray between shield and leads, but I guess the noise will carry the least and be happy with two cold ground points in either direction?