The more intense the coupling between channels the more severe the distortion glitches at loss and recover of sync, which happens when instantaneous signal levels become too dissimilar for the intensity of coupling. The distortion depends on the type of coupling. Fast square pulses from gate drive di/dt coupled into carrier residual are the worst (distortion spectrum similar to square wave), followed by square wave from output inductors di/dt coupled into carrier (distortion spectrum similar to clipped sine wave).
The solution adopted by well known brands is not cleaner. Shifting idle self oscillation frequencies of channels by 20khz to 40khz moves the main peaks in noise floor out of audio band, but side-bands are still present (due to intermittent interference, like premature switching in 1 out of 2, or 1 out of 3 events), the channels technically idle "battling", the noise floor has more than noise in it.
Class BD has similar pseudo-crossover and sideband distortion issues at low level, in my opinion it is good for battery operated equipment. In mains powered equipment the extra few watts to make it low distortion are always available, in this case the trick being adopted by various designers for low level listening is to act on power supply to reduce rails, until high signal is detected, a 30% reduction already cuts loss by 50%.
There is a superior solution for clean noise floor in self-oscillating multi-channel, but I will leave to you the work of figuring it out. It is called S__f T____ing.
The solution adopted by well known brands is not cleaner. Shifting idle self oscillation frequencies of channels by 20khz to 40khz moves the main peaks in noise floor out of audio band, but side-bands are still present (due to intermittent interference, like premature switching in 1 out of 2, or 1 out of 3 events), the channels technically idle "battling", the noise floor has more than noise in it.
Class BD has similar pseudo-crossover and sideband distortion issues at low level, in my opinion it is good for battery operated equipment. In mains powered equipment the extra few watts to make it low distortion are always available, in this case the trick being adopted by various designers for low level listening is to act on power supply to reduce rails, until high signal is detected, a 30% reduction already cuts loss by 50%.
There is a superior solution for clean noise floor in self-oscillating multi-channel, but I will leave to you the work of figuring it out. It is called S__f T____ing.
The fixed-synchronized-clock approach (quartz or not w.t.f. haha) still used by some brands is no miracle either, it exhibits low level interaction when PWM transitions of various channels happen in close time proximity, in other words when channels exhibit similar instantenous output voltage, this interaction is worse than in self-oscillating, micro-dead-bands as from dead time are introduced in transfer function. In self-oscillating low level interaction introduces "1-out-of-n-period" micro-glitches. Thus the difference is that in fixed-synchronized-clock the error due to micro-cross-talk is static! (although shifting clock phase from channel to channel can move the micro-dead-bands to high signal levels), while in self-oscillating the error is dynamic, as the switching frequency either becomes different in next cycle (if enough mismatch) or becomes locked synchronized due to crostalk (like a "sticky" behavior between Fsw of channels).
Last edited:
Its a long way from 2015th year till now.I made exactly what i was talking about with noise floor buried very deeply where no one can find it 😀 .Its possible to make all up to perfection....it took me years but its possible.Eva you are right here....BD has injected error on micro power level....can be moved somewhere but not eliminated.I have figured out many things with very hard work and i eliminated this error completely.I am working on few more things right now and once i finish all i will publish results.This was great journey but not finished yet 🙂
- Status
- Not open for further replies.