going back to this topic, if you think "digital' as in the forms of "on" and "off", I would argue that the bulk of the amplification in a typical class D amp, whether analog or digital input, is done with a bunch of "on" and "off"s. so that qualifies it the 'digital' title.
so it is fair to call them digital. Of course, there are different degrees of "digitization" among those digital amplifiers but that's just a fact of life and doesn't foundamentally change the picture.
Again I have to disagree here. "Digitizing" is taking a signal, and "rounding" it's value into the nearest discrete "number" you wish to put on it. You do this a given amount of times per second.
May it be 1 bit at a megahertz sample interval (DSD) or 16 bits every 44.000 of a seconds, yes, then you are digitizing.
However, the self-oscillating class-D (for example UcD) isn't measuring in intervals, it's doing so continously, and can turn the output transistor to 0 or 1 at any given time and later recreating the input wave *also* in an analog way, by putting the outputted power into a coil. This switching couldn't be stored "digitally" as it can switch at any given discrete moment in time- if you'd have to store it you'd have to slash it up in intervals again.
Put in other words, you can't represent into a list of numbers how the transistors in a UcD design have switched in a one second time frame. Sure, you can give me a row of zero's and one's, but you cannot exactly say when these zero's and one's occured. You lack the exact moments in time in your "number list" - in a PCM stream this is clocked, and then it is a digital representation.
This is a good case *against* calling this type of class-D digital - this is analog.
Of course, if you see "analog" as the most intense and ultimate form of digital quantization (on an atomic level) then I have to give you the advantage. But then class A or AB is also digital, as you are smacking around a finite number of electrons back and forward...