Well I'm going to finish that design just for the sports of it (it's nice if you can bandy around a plot saying 0.0005% @1kHz 100W). I would dread listening to it though.IVX said:Oops..4000 damping factor equal a lot of feedback, on the THD/Freq plot easy to seen slope with pole around 300hz. It's remind my UcD experience with additional loop on the opamp based integrator, but [email protected] was <0.002%, but the sound lost any of the UcD advantages. JP, if i remember correctly, posted here THD/Freq plot of the such UcD modification also with very sexual [email protected], however (AFAIK) they do implement this approach only for subamps, probably JP/Bruno can comment it?
The distortion results are also similar. I don't consider a bandwidth (let alone a power bandwidth) of 150kHz or more necessary or even wise. Either you have to use a very high switching frequency or you'll get serious intermodulation artefacts when feeding the amplifier supersonic noise (like playing an SACD).
Having a BW extending to 100kHz is only a way to improve phase response inside the audio band. Choosing a Bessel (Thomson) lowpass response for the amp is equally valid, and allows for a much reduced bandwidth.phase_accurate said:Some People (Mrs Manger for instance ) are hard to convinve for having bandwidths lower than 100 kHz.
But apart from that, there is something else that makes me curious: How does the UcD react to supersonic noise ?
My feeling says that it is less susceptible than a carrier-based class-d on the one hand - OTOH its small signal bandwidth is larger than the bandwidth of the output filter, making it more susceptible to supersonic noise (i.e. TIM). I am of course aware of the advantages of what you do with this filter dimensionig (better carrier suppression for a given overall cutoff frequency, better load independancy ....).
Choosing a Bessel (Thomson) lowpass response for the amp is equally valid, and allows for a much reduced bandwidth.
phase_accurate said:I agree on that. Unfortunately there are a lot of other unknown lowpass poles in a chain usually. This is a thing that could be optimised in the (reproduction part of it at least) chain without much cost and just some work effort, i.e. including preamp and tweeter response as well in order to achieve an overall Bessel response.