Unbalanced vs Pseudo-balanced interconnects (inc. DIY)
Posted by: Huge on 18 May 2016
Richard - if you think it should be in the Padded Cell, then please move it there.
I recently decided that I would benefit from improving the interconnect cable from my ND5 XS to my Nait XS 2. So I looked at the alternatives from the HiLine down. Unfortunately lack of space behind my rack makes the HiLine a risky proposition (risk of breakage on trying to positioning it!). I looked at reviews of Flashback Sales, Witch Hat and AR Sound, and concluded the latter made most sense for me. Then I read about DIY interconnects using Mogami Neglex cables. As I’m quite used to using a soldering iron I decided that was the way to go – cue more research.
The question arising was how to configure the cable – the classic approach is a conventional unbalanced coax pair, but that doesn’t work well for DIN – DIN cables. Next option was unbalanced twinax, OK, so Mogami W2549. But there’s another option – to use Mogami W2534 (screened 4 core) wired as a twin (stereo) pseudo-balanced pair. So I decided to model them using electronic circuit simulation software, so see if I could glean any information other than the maxim “Unbalanced Coax for single ended inputs, screened twinax for balanced inputs”… and that’s when chaos commenced!
I started with a basic model of just the main components: The output impedance of the source, the input impedance of the amp and a set of L-C-L cells for the cable. I put interchangeable voltage or current sources into the screen to represent the energy of the RFI impinging on it. I made sure I could switch between current and voltage as the induced current in an antenna isn’t amenable to analysis by classical Thévenin equivalence.
Initial results indicated a specific weakness of the pseudo-balanced connection at higher RF. This was as expected; what wasn’t expected was that the pseudo balanced configuration was orders of magnitude better below 100kHz. So now the hunt was on for a more complete model to see if this was real or an artefact of my simplification.
To the model I added the terminal impedances (R & L) of the DIN connectors, the parallel capacitance of the source and load impedances, the earth connections and stray capacitance from the load amps 0V ground planes to the case (at earth potential). Then it struck me… in normal configuration with a Naim source component, Naim amplifiers don’t have single ended inputs as they don’t connect their zero volts line to earth – they’re actually unbalanced differential amps.
So: Run all the models again in this configuration. Now, even at high RF, the models indicate that a pseudo-balanced connection performs almost as well as coax. Adding a bypass capacitor (2n2) to this, linking the receive end screen to 0V, improves this further. This now seems to be better than an unbalanced coax cable for all frequencies of interference.
The Mogami cables (W2549 and W2534) to make the twinax and capacitor bypassed pseudo-balanced interconnect should arrive this week, so the proof of all this modelling will be in the listening!
I don’t propose to post all the circuit schematics and graphs (and certainly not the tables) from the modelling, but I’ll consider a link to it if people are really that interested (and if so that will definitely be Padded Cell material!).
Huge posted:
The fact that a ferrite on the interconnect makes a significant difference strongly suggests that I have significant levels of radiated RFI as well.
Hi Huge - I don't follow your logic. You might have radiated RF issues, but a choke on an interconnect will impede common mode RF current flowing along the cable.
I use them with my radio equipment on feeds and control cables - if they impeded radiated RF as opposed to common mode RF near my active antennas they would probably catch fire ! 
But Huge - they can get warm - and in the cited example they could get very warm because of the eddy current induced in the lossy ferrite material from its inductance - i.e. there is a real R - now I know you knew that 
Yes if they were perfect inductors there would be no heat - not sure mankind has discovered those yet
Correct common mode applies to both conductors. If you put a choke around only one conductor you will have created an inductive filter.
The RF can be created from the device or passed via the mains into the device and it can conduct. RF can get anywhere and everywhere. Also remember beers of the nature of RF you have nodes of current flow and no current flow and therefore current can flow in a way you don't expect between devices - even if both connect to the same mains supply that is acting as the RF source.
Simon-in-Suffolk posted:But Huge - they can get warm - and in the cited example they could get very warm because of the eddy current induced in the lossy ferrite material from its inductance - i.e. there is a real R - now I know you knew that
Correct common mode applies to both conductors. If you put a choke around only one conductor you will have created an inductive filter.
The RF can be created from the device or passed via the mains into the device and it can conduct. RF can get anywhere and everywhere. Also remember beers of the nature of RF you have nodes of current flow and no current flow and therefore current can flow in a way you don't expect between devices - even if both connect to the same mains supply that is acting as the RF source.
Hi Simon I did acknowledge the heating in real world inductors ("only the inductors' resistance and magnetic loss are in the real plane"). Indeed this is one of the reasons why my mains DM inductor is air cored (and so d*** big!).
With interconnect cables the ferrite is almost always common mode as it's clipped round the shield and all the cores, rather than disassembling the structure of the cable; and this is certainly the case for mine.
Th point about the resonant nature of RFI causing nodes and antinodes is valid and well taken, and as you say, it makes things rather unpredictable. The inductors (inc ferrites) in/on my mains supply system are there to ameliorate the overall effects of conducted RFI (and any local pickup), acknowledging that it can't be eliminated, only reduced. Short of completely covering a cable in ferrite, inside and out, there doesn't seem much one can do about this, except perhaps using a log-periodic distribution to achieve selectivity to a narrow band (the less you open the window, the less the muck flies in!). I think the side effects of the cure may be worse than the disease.