It's fun watching discussion that have no ending.
Instead of focusing building an amp around 5532 using differential config and non-inverting.
The topic now sway towards equipment that are not design properly causing unbalance to be worse than balance.
"No Ending" just means we have a profound circle of confusion here ;-)
Using unbalanced may or may not create hum/buzz problem in the real world, sorted by importance (IME):
- When conncting gear that both has audio GND connected to mains earth, either directly (worst case) or the "GND isolator" circuit often seen, a resistor shunted with a small capacitor shunted with anti-series diodes (which must withstand the rated short-circuit current for the circuit breaker / fuse in you house's distribution panel). The larger the GND difference the bigger the issue, either from using different outlets or from I*R drop along the PE conductor that carries mains filter deflection current. "Antenna" feeds for cable TV etc also establish such a polluted GND scenario.
- When one of the components is earth-grounded is connected to a class-II (2-prong) "isolated/floating" device whose supply has significant mains coupling. SMPS are most offending as the couple the recitified mains, whereas 50/60 transformers couple the "pure" mains only (50/60Hz + some harmonics + noise)
- When both devices are class-II "isolated", the system GND being fully "floating," and have medium to high coupling to the mains, again SMPS being most problematic. The worst case -- in the US -- would be to connect one device on one phase of 110-0-110 biphase net, the other on the other. Leakage doubled. In 3-phase networks, using differernt phases.
The problem simply scales with the shield resistance of the unbalanced cable, the higher this resistance and the longer the cable the more error voltage is developped. The bottom level is set by the internal routing of audio GND in the device itself and some devices have, as noted, issues here and some even generate their own noise from internal currents sharing a GND path between inptus/output connectors.
As for the solution without modding the devices and trying to establish a system-wide common signal ground grid/plane (which is a lot of effort), the simple 4-resistor subtractor recevier is good enough to reduce the error signal to be insignificant in most cases. 20...30dB reduction is a lot here.
For precision/measurement applications, you will want something better, depending on the needs:
- transformer for best LF CMRR and more importantly, to avoid CM clipping. RF must be taken care of seperately.
- balanced CM- and RF-bootstrapped high-Z instrumentation input, Whitlock style. Pin 1 connected via direct RF path to chassis and with a series R for the DC/LF path, to avoid large balancing current for above case 1.
- use excellent cabling especially when the cable is expected to carry shield current.