Hallo Per,

I must confess that P89LPC932 seems to be a highly interesting and innovative microcontroller, although I do not like philips very much, from other reasons...

You should interpret built-in RC-oscillator and built-in brown-out detector of 'LPC932 as added features, which you CAN use, but need not to use. If you need the precision of quartz oscillator, then you will not use RC-oscillator, and if you can profit from saved PCB area and reduced costs of omitted quartz then you will probably use this RC-oscillator. It always depends on your needs...

Brown-out detector is the same. If you have a stand-alone application without any bus system, without RAM banks, EEPROMs etc., why not use built-in brown-out detector?
But if your application is NOT a stand-alone application consisting of only this micro plus 'two resistors and a switch' then problems can arise:

1. What is a good threshold level of brown-out detection for your 'LPC932 need not to be a good level for other external devices!! 'LPC932 can accept a wide range of supply voltage, other devices possibly not...

2. Stand-alone applications do rarely suffer from those levels of interference, spikes and dips on supply voltage and ground bounce like extended microcontroller applications!! Don't make a mistake, that's just another league, believe me... And what is a good brown-out detector for a clean and tiny stand-alone application can result in total desaster with bigger applications.
Keep in mind, that only due to the increased capactive load of a bigger application (use of extended address bus and data bus!), supply voltage spikes and ground bounce internally of the micro can be so high, that even the best decoupling measure could not prevent the built-in brown-out detector from failing. Don't forget, that decoupling measures do only sit OUTSIDE the micro. And that you can keep supply voltage quiet at the supply voltage pins of micro, does NOT mean, that internal supply voltage points directly on die are also quiet. NOT AT ALL!!
So, built-in brown-out detector will always suffer from this impossibility of decoupling supply voltage at die level.
How do manufacturers solve this problem? Well, they can't eliminate supply voltage fluctuations. The only way seems to be to make the brown-out detector less sensitive...
Here comes the big advantage of external supervisory chip: It can be decoupled separately to any needed extent. It does not suffer from what is going on directly on the micro's die, 'far' away from any decoupling capacitor!!

3. In opposition to what manufacturers promise, built-in brown-out detectors are very often NOT foolproof, even in typical stand-alone applications. Also, performance of built-in brown-out detectors are often very poorly documented in datasheet. Sometimes you need some erratas to understand how they actually work. E.g. many brown-out detectors need a minimum time for the decrease of supply voltage between two specified levels. This minimum duration can be so long, that battery powered applications can show malfunction, when Vcc is simply switched-off by a switch and not enough decoupling capacitance (better: storage capacitance) is provided on the board. Then, micro sometimes looses its code, and there seems to be no explaination...

Concluding: I'm a big fan of EXTERNAL brown-out detection. So, for me the use of external supervisory chip is essential, even in stand-alone applications....

Kai