Wow lads.
I've read that project page over and over a while back as I wanted to make a low voltage relatively high quality (and powerful) amp for monitor speakers. The power involved, and the fact the output is bipolar, really does mean a transformer is required (as opposed to simple boost converters). Although given the switching topology, its winding ratio can vary.
Another restriction (a BIG one) is its use for audio, so one must push the frequency up well out of the audio band, and even then beat frequencies betwen parts of the circuit can creep into the audio chain. The EMI from that circuit should be quite low, as relatively low conducted noise for a switched mode power supply. A flyback or boost would require so much filtering it could esily double the cost/size of the unit. Constant frequency PWM is mandatory for low noise applications, to ease filtering, at the cost of possibly most complex parts, and lower effiency at low power.
So, could it be built cheaper? if it was simply a quick and dirty 12V to +/-35V supply, at that power, its possible, but it wouldn't be any good for audio. That page is MADE for audio, perhaps not the finer points of SMPS, but I'm confident that schematic will work jstu fine for an amp. Ubergeek, if you feel you can knock up a schematic which the same noise output, efficiency, that uses less parts and is cheaper, go ahead! Its not the only way to do it, but I feel the designer, given all the variables (including parts availability, since the PWM chip is common) did an excellent job
Only ***** about SMPS, is winding your own transformer, getting any core gapping right etc.. really does require an oscilloscope.