16/02/28 19:37:09.51 6lPo5eJl.net
HeadFiのDaveスレッドのポスト197に書いてあるよ。
Component count is very important for transparency. Doubling the number of parts in the direct
signal path does degrade depth perception and detail resolution.
But there is another problem with balanced operation. Imagine a balanced differential in,
differential out amplifier. The input stage is normally a differential pair (maybe cascoded)
with a constant current source. Now the input stage is free to move up and down to accommodate
the common mode voltage - but the input stage common mode impedance is non linear,
and if the common mode voltage has a signal component (it always will have due to component tolerances)
then this will create a signal dependent error current, thereby generating distortion. Unfortunately,
the negative feedback loop of the amplifier can't correct for this distortion as it can't see the error
on the summing nodes. So there will always be a limit to the performance. With SE operation, this problem
does not occur, as the differential input stage is clamped to ground.