As a result, the measured waveform represents both the constructive and the destructive combinations of the noise components and any residual ringing from the previous cycle Socially-aware caching strategy for content centric networking. This effect is manifested in the waveform as slight inflections in the ringing of the large spikes. From the measurement, these large noise spikes occur every, corresponding to the rising edges of the clock. Compared to the switching distribution of the CBL PRNG in , there appears to be little stratification in the measured waveform to reflect the individual transitions in the design. Instead, a single ringing pulse represents the lumped effect of the transitions at this clock edge. Socially-aware caching strategy for content centric networking In order to get more insight into the component of noise caused by the CBL PRNG, the clock frequency was reduced to allow adequate time for settling. The center plot of shows the noise waveform obtained for this measurement. The rise/fall time of the clock was kept constant for comparative purposes. As seen in this plot, there is a distinct difference in the noise waveform corresponding to the rising and falling edges of the clock. Recall from that all of theswitching activity occurs at the rising edge of the clock. This indicates that the smaller noise spike seen in this waveform for the most part represents the noise injected capacitively at the falling edge of the clock. Meanwhile, the noise at the rising edge reveals the complete settling response of the switching noise from the CBL PRNG. As before, there is simply a single ringing pulse for each clock transition. The ringing frequency and settling response of this noise spike should simply reflect the characteristics of the package parasitics. Socially-aware caching strategy for content centric networking However, further analysis shows that this ringing is also a function of the switching distribution at each clock edge. For example, if two gates are switching at the exact same instance, then the resulting noise is the superimposed value of the two pulses. Thus, there would be an increase in magnitude but no change in the resonant frequency. Now suppose that the switching times of the two gates are offset by a small fraction of the resonant period. As illustrated in the resulting waveform will show a change in both amplitude and frequency. The actual change will be dependent on the difference between the resonant frequency and the pulse length of the distribution. Effect of a time offset between the two waveform components. The solid line is the resulting waveform when there is an offset and the dashed line shows the resulting waveform when there is no offset.Frequency domain comparison of the ringing characteristics in the substrate noise for high inductance package top synchronous PRNG with a clock and bottom asynchronous PRNG. The bottom plot a closer look at the ringingfrom substrate noise caused by the NCL PRNG. Socially-aware caching strategy for content centric networking Aside from having a magnitude that is much less than the noise from the CBL PRNG, the ringing in this waveform appears to have an approximate period of This value agrees very well with the periods of high switching activity found for the NCL PRNG switching distribution in However, the shape of this waveform does not seem to exhibit the exponentially decaying ringing typical of injected substrate noise waveforms. Socially-aware caching strategy for content centric networking Instead, the waveform simply looks like an amplitude modulated sinusoid. This waveform shape can be explained by relying on a similar argument as that presented above.