OK, so I’m putting together a new autopilot. My last iteration I had down to ~1" square and it used the LPC2138. This one will have a more sensors, and will be a little larger. The processor is also larger.
Anyways my point is that a lot of work goes into the component selection and the final artwork to get it this small. Anything superfluous is left off.
Is there any reason to use an external oscillator? It looks like I can use the PLL with the internal clock to get the full 100MHz. Are external oscillators used because they can be more accurate? Lower power? Less noise?
Possibly more accurate… a lot of the time, the internal oscillator is implemented as a LC oscillator, which will tend to drift with temperature changes. I could be wrong with the LPC2138 on this aspect, but I know this to be the case on MSP430.
If this does not matter for your application however, you should be okay.
Internal oscillators are RC, not LC.
A properly designed oscillator using a crystal should be just as accurate as an oscillator module, unless the latter is an expensive unit with an oven for temperature stability.
Leon
Just to be clear, this is the 1768. I used a 20 mHz crystal on the 2138 with a x3 PLL to get to the full 60Mhz capability.
I’ll do more reading about the temperature drift. This is a real time system that is using time steps to perform calculus operations, so a second needs to be a second - to a fault. I can’t quote an exact percentage where it would be unacceptable.
leon_heller: My apologies… knew it was involving a capacitor rather than a crystal or ceramic resonator (which are often modeled as a LC circuit). 
LooseTooth: Some MCUs have two oscillator inputs, one for the actual system clock, the other is a 32.768kHz crystal for this purpose – easily divided to get exactly the timebase needed. Maybe this is an option?