Dear Raj,

Since the dead-zone of a phase detector is commonly a fixed time width, if you examine the loop performance (i.e. output phase noise for example) using the same basic loop bandwidth but with a set of increasing reference clock frequencies, the presence of the dead-zone on a unit-interval basis will be increasing. Hence, its impact on the output phase noise will be more evident.

For example, if you start with a reference clock of 10 MHz, a loop bandwidth of 1 MHz and jitter peaking of < 1 dB, and a loop divider of, say, 8. Examine its phase noise and then compare it to the following 2 cases:

1. Loop bandwidth of 1 MHz, jitter peak < 1 dB, reference clock frequency of 20 MHz, divider of 4
2. Loop bandwidth of 1 MHz, jitter peak < 1 dB, reference clock frequency of 40 MHz, divider of 2

If the dead-zone of the phase detector is significant relative to periods of 1/10 MHz, 1/20 MHz and 1/40 MHz, its presence should be evident in the output phase noise. Intuitively, the presence of the dead-zone means that for phase differences between the reference clock and feedback clock that are within the dead-zone, the VCO will be running open loop and hence the low frequency phase noise will mimic the open loop VCO phase noise.

An alternative is to apply sinusoidal phase modulation with some amplitude Aj in UI to the reference clock at a frequency less than the loop bandwidth and measure the output phase modulation at the same frequency (a filtered measurement is preferred to reduce noise). As the amplitude Aj approaches the dead-zone of the phase detector, the output phase amplitude will no longer track the input phase amplitude.

I hope this helps Raj.

Shawn