Stereo crosstalk gets worse with frequency?

[xiedidan]

Hi, Guys

I'm debugging on my TPA6120 board and tested its RMAA with E-MU 1212M. For now, the results look ok except stereo crosstalk:

-90db is not so good, so I looked into the detailed graph:

While the frequency goes higher, the stereo crosstalk gets worse?...

My question is quite straightforward then:
Why could this happen?
And how to solve this problem?

Thanks

 

[jcx]

its not a audible problem - crosstalk is one of many numbers that quit correlating with hearing at gross levels by what electronics achieve
anything better than ~40 dB is fine as long as it is linear, not too strongly peaking in frequency

the chip is capable of better by the data sheet - so you may have compromises in the board, gnd layout or are just seeing TRS common impedance and cable wire mutual coupling - at -60 dB debugging your test setup is critical too

 

[MarcelvdG]

Presumably either capacitive or inductive crosstalk between the channels, or both. Like jcx already wrote, if you want to improve it, check for common wires or wires of one channel that are close to wires of the other channel. High impedance nodes are particularly sensitive to capacitive crosstalk, low impedance loops to inductive crosstalk.