Per:
My reply will be off subject a bit but I use the Cygnal Chips and they have both of the attributes that you question about the new Philips devices. 1) Onboard R/C oscillator and 2) Brownout Detector.

I find the R/C oscilllator to be very handy because the first boards come right up even if the XTAL circuit is not working (as a matter of fact the R/C oscillator is what you run on out of rest anyway). But the tolerance of the oscillator just is not good enough for accurate baud rates and/or time period measurement. One Philips part I looked at had the ability to trim the R/C oscillator into speed but Im not sure this is a temperature compensated calibration. Also consider that each PC board must be calibrated and that may very well cost you in labor what a crystal costs you in additional materials.

I always use an external supervisor chip. My experience over many years has shown that the guys and gals that design sophisticated microcontrollers just do not have the experience to design a real good onboard reset and voltage level detector. The tolerances of such circuits generally are so broad as to render the circuits nearly useless. Keep also in mind that as time goes by more and more special peripheral circuits are appearing onboard these chips and a hard RESET is the only really known relaiable way to ensure proper initialization of these circuits. I suspect that many of the companies that are plastering the likes of CAN controllers, Ethernet controllers, USB engines, etc etc are not even designing these circuits themselves but instead buying IP that they not fully familair with. (Are those FULLY tested? Probably not!!) Do you want to wait till your product is in the field to see if an even somewhat marginal on-board reset circuit misbehaves just a little bit and fails to ensure that one of these complex chips is not fully and certainly reset? Not me!!

Michael Karas