The inherent device noise (RMS Voltage Noise) can be thought of as a constant and the jitter results from this can be lower for higher slew rates as implied by the following equation:

If noise of the input buffer dominates and increasing the slew rate (while holding frequency constant) improves significantly, then the phase noise is said to be slew rate limited. At some point, the phase noise from other parts of the device dominate and the phase noise is said to be not slew rate limited. At this point, increasing the slew rate brings diminishing returns, such as only 0.1dB for doubling the slew rate.

Table 3-1 shows fundamental trends for flicker noise and noise floor as a function of output frequency and slew rate. The total phase noise is the sum of the slew rate limited and non-slew rate limited phase noise sources. Derivations for this table are in Section 8.

In the case of a sine wave or clipped sine wave, the slew rate is proportional to the frequency. Applying this assumption to Table 3-1 gives Table 3-2. One observation is that when the frequency is higher, the device tends to not be slew rate limited and the flicker noise increases with frequency at a faster rate than the noise floor. For this reason, flicker noise becomes a much larger consideration at high frequencies (>10GHz) for buffers.

Data for Table 3-3 data was taken from a graph in the Texas Instruments LMK00301 which demonstrates the behavior of a slew rate limited noise floor. As predicted, quadrupling the frequency while holding the slew rate constant results in about a 6dB degradation of the noise floor. Doubling the slew rate while keeping the frequency constant results in a 6dB improvement in the noise floor.

Figure 3-2 gives a hypothetical example that illustrates general trends for input buffer noise assuming a noiseless sine wave input clock of constant amplitude. The flicker noise starts out as constant versus clock frequency because the fact that the noise is slew rate limited counterbalances the general tendency to increase in frequency. After about 100MHz, the flicker noise increases due to not being slew rate limited. The noise floor is slew rate limited at lower frequencies and therefore actually improves for a while as the frequency is increased. Around 40MHz, the slew rate is sufficient and then the noise floor starts to degrade with frequency.

Figure 3-3 shows the noise floor data taken from the Texas Instruments LMX1214 high frequency divider buffer and demonstrates the general bowl shaped trend for noise floor similar to Figure 3-2.

Table 3-4 can be derived by assuming that frequency is constant, power is changing, and applying Equation 5 taking Table 3-1. This illustrates that both the flicker noise and noise floor are impacted in the same way by input power for this case. Figure 3-4 illustrates this in terms of a phase noise degradation.