"** Then kindly explain how the PWM signal is being generated. "

Relaxation oscillator to generate a sawtooth wave, comparator to chop it up into PWM. It's not that hard really. The chips that do it are just easier to use than to design it with discrete components. After tha you got your PWM square waves, then you design the drive circuit for the outputs. What's more, even though it is BTL, you only need one, it is simply inverted for the other phase.

Looking at the picture, there are plenty enough transistors and whatnot on the board to do it.

Of ocurse the question is why. Why didn't they just use some off the shelf chip for that ? To answer that we need to know things we are not going to find out. Amybe some parameter was too hard to get out of the chip, maybe their chopping frequency is lower than the chips like to operate at. Maybe certain effects were easier to implement, though I'm having a hard time fathoming any at the moment.

Or it may be attributable to audiophoolery. Some people like discrete components, myself included. I know damn well thaat what's in the chip is the same as what's in the transistor package, but...

Like if you bought a Marantz that uses an STKXXX instead of a Luxman that uses the same STKXXX, what did you get different ? Maybe us nuts just like to make the engineers work harder.

But really, among musicians you will find a healthy dose of audiophoolery, so it may be discrete for marketing purposes.

Of course in class D it doesn't mean a thing as far as I can tell. But then if ONEE golden ears can prove he hears the difference, the market will beat a path, to a certain door.