I am going to be working with the Freescale MC9S12XE processor and trying to figure out the oscillator configuration. They offer the following options:
Loop Controlled Pierce Oscillator
Full Swing Pierce Oscillator
They recommend the Loop Controlled for most situations but suggest the Full Swing for high frequency crystal operation.
The documentation states the following about the general Pierce Oscillator and noise immunity:
The XOSC will contain circuitry to dynamically control current gain in the output amplitude. This ensures
a signal with low harmonic distortion, low power and good noise immunity.• High noise immunity due to input hysteresis
But it does not say if one configuration offers more noise immunity over the other.
My question is, would one of these configurations offer more noise immunity than the other? Why or why not?
Best Answer
The Pierce oscillator has no noise immunity right at the transitions, and the amplifier likely has a pretty good high frequency performance.
In both cases, the input of the amplifier is connected to ground through C1, and the slew rate through the zero crossing should be fairly rapid, so it would take a lot of energy at high frequency at just the right time to disrupt the waveform. There is, however, no hysteresis so the noise immunity measured in volts is low. The current at the transition is high due to the capacitor, so the noise immunity measured in mA is not so low.
So you could say that the arrangement with hysteresis should have higher noise immunity.
If you define "noise immunity" to be the voltage applied in series with the EXTAL input that would disrupt the waveform that's probably not so useful in terms quantifying immunity to externally generated EMI. However, if you had a non-ideal PCB layout so that there was actually such a voltage due to the ground bouncing around, this could be important.
Early on-chip crystal oscillator circuits used hysteresis (eg. Intel MCS48) but that method was widely abandoned in favor of the now-standard Pierce circuit (fewer problems). It's nice to see that on-chip oscillator evolution continues.