A quick immediate test - there is a line amplifier with 30dB gain (called Amp 30dB) on my workbench. Its input is shorted and its SE output is connected via shielded twist-pair cable RCA-XLR to the balanced measuring input. There is no other connection than USB from the soundcard to the PC notebook that is class II instrument.
Now, to the same 230V/50Hz mains line there is a SMPS 24V/5A power supply connected which I switch on and off. The DC output of this power supply goes to AIYIMA A07 amplifier which I also switch on and off.
This A07 amplifier is neither connected to the 30dB amp, nor to the soundcard. The only interaction is via mains 230Vac power, which is used to supply the Amp30dB, PC and SMPS.
1) output noise of Amp30dB, SMPS off, A07 off
View attachment 150116
2) output noise of Amp30dB, SMPS on, A07off
View attachment 150117
see the spike above 500Hz
3) output noise of Amp30dB, SMPS on, A07 on
View attachment 150118
Again additional spikes
These interfering spikes are very low in level, however in the audio band. The only coupling is via 230V mains network. The line above 500 Hz is most probably an intermodulation product reflected in the audio band. Additional filter at the SMPS mains line would with highest probability change the measured plots.
Only 50 cm of pseudobalanced RCA-XLR cable was used at the Amp30dB output. Longer cable or 2-wire SE cable would for almost sure worsen the results. The additional spikes have origin in mains interference voltage and also result in cable shield capacitive currents.