This is not a 1KHz wide spectrum but over 1,2MHz wide showing inaudible switching frequency, you are continuing spreading BS and trolling on this 3d.
Some more trolling BS:
The noise of the Class AB amp consists of low-freq ~1/f noise and the flat, white noise shelf. Due to the wide-band feedback nature, the level of this shelf is determined mostly by the input cascade. The shelf lasts ... into MHz range.
If the amp is measured on its own, the thermal input impedance (10kOhm "standard") noise dominates. For "standard" 30dB and 20kHz bandwidth it becomes ~ 50uV, which is about -112dB re full-scale signal 100W 4Ohm amp. Of course, it is audible from a 90+dBW tweeter (well, grammar corrector did not screw me up, finally) from 0cm distance to a median listener. 90dBW loudspeaker can (potentially) produce 110dB at 100W at 1m. 0dBSPL at 1kHz is audible to 50% of listeners in headphones, thus about -10dBSPL shall be audible in free air. However, almost no one complains if the listener is at >=1m distance unless it's anechoic room and the ear has already adapted.
If the amplifier is measured with the input shorted, the noise is quite a bit lower. Now it's defined by the voltage noise density of the input cascade. You can often assume self-noise to be > 10dB lower than for the open amp.
The noise of the class D amp consists of the same ~1/f + white noise of the input cascade, same as for AB, plus Class D - specific high-frequency rising noise, ~f (or even steeper, ~f^2). One of the reasons I already mentioned: feedback takes a derivative of the output cascade's noise and distortions, hence the ~f nature of preemphasis, the aforementioned "hiss". It's even higher for "digital" coarsely-quantized PWM Class D amps using noise shaping explicitly. The detailed explanation is not for the uneducated. I looked through Bruno Putzey's publications - he definitely knows the theory and plots the right figures, but he did not go into an explanation I could refer to. It requires the reader knowledge of math and feedback control theory at PhD / MSc in automatic control which I can't assume as existing after reading some of your illiterate comments.
I've been through many episodes clearly displaying that claims of understanding of the theory of feedback control have nothing to do with reality. Once in my experience, while skiing in the mountains, a friend of mine told me that an optical system [he worked on for a few years] had been frequently going out of control. That optical system was an optical last mile, with a pair of transceivers A and B, each having a laser and receiving photodiode. The initial working point was set by a technician. The system was designed to remain in the photodiode’s working zone in any conditions by adjusting the laser power, driving it lower when it’s clear, and higher when it’s raining or foggy. As the fog affects both transceivers, the indication of atmospheric losses was read locally, from A transceiver‘s photodiode to affect the same A transceiver’s laser’s average power (same on B). However, a way too often transceivers were jumping out of phase: A was transmitting on the highest power and B on the lowest (or vice versa). To me, it was immediately obvious that while the feedback was negative for common mode, it was positive for a differential mode. Of course, they had to use “A” photodiode to adjust “B” laser - but that was not possible within their architecture.
To them, it was not obvious at all: the company fought with the problem for at least a couple of years, with nearly 200 engineers and endless consultants and professors. All of these engineers and scientists learned control theory in Universities, everyone passed the exams, but no one could apply the theory to practice (BTW, eventually, the company went belly up). If you are like those engineers and incapable of understanding what I wrote - feel free to call it BS, I really don't care.
Of course, depending on the conditions of measurement (open/close), the total noise spectrum may look different. The upward ~f slope may be clearly present or maybe somewhat masked by the white noise of the preamp. In any case, it does exist for any Class D topology I am aware of.
On that, I'd like to wish you all the best!
