MSP430F5438A: What is the reason for the resistance value of the pull-up resistor and the value of the capacitor specified for the RST/NMI pin in the data sheet?

The most fundamenal E-row contains values like 1*10^x, 2.2*10^x, and 4.7*10^x - I assume that's why the 47k are selected. Smaller resistors require too high capacitors for a certain time constant and cause a high inrush-current. At large resistor values (e.g. 1MOhm) especially the pin-leakage-currents may cause problems in harsh environments (e.g. high umidity and/or high temp.) - but even more annoying is the fact that an oscilloscope probe then alters your effective resistance (not that critical for a reset-pin but critical in e.g. OpAmp-circuits where the resistors define gain, cutoff-frequencies, etc.).

So the question on the resistance-range should be more or less answered. Let's get back to the capacity:

MSP430F5438A has internal pull-ups with typ. 35k (20k to 50k over full temp. range @ 3V). Let's assume VCC=3V with 10nF and the internal pull-up: The time constant is ~350us; the RST-Schmitt-Trigger (at 3V) has VT+=1.5 ... 2.1V, so the reset period is effectively given by Vt+=3V*[1-exp(-Trst/350us)] => Trst=(243...421)us. That should be enough for most systems.

Too short reset-time constants are not useful because

• There is a minimum RST-delay of 2us for the MSP430F5438A
• The internal LDO for VCORE also takes a few 10µs for startup
• The SVS/SVM takes up to 100µs for release (in Low Power Mode)
• The REFO-oscillators which sources the reference for the DCO has about 25us startup-time.
• More important: You want to be sure that when your code execution starts the MCU power supply is stable to avoid unexpected program behaviour by e.g. BOR, corrupted RAM contents, etc. The MSP430F5438A is not that cirtical with respect to the voltage supply but especially smaller models (e.g. F2xx-series) do only have one fixed BOR-voltage which cannot be modified even when executing code with high MCLK frequency.

Long resets (i.e. 1ms and more) are possible but not useful because

• Depending on what your MCU controls you have to take into account that your port pins will have to be initialized after reset to avoid system shortcuts, unnecessary loads (especially during startup!), etc.

The 2.2nF-value is especially important if you program your device with the 3-wire Spy-Bi-Wire-interface: There the RST-pin is clocked by the debugger and - of course - a large capacity combined with a certain source-resistance of the debugger pin causes a significantly distorted clock edge.