3

I think I understand the general operation of PLLs in the case where a single reference clock generates a single output clock. Once the PLL is locked, there is a deterministic relationship between the input and output clock frequency and phase. I understand the negative feedback loop, VCO, etc. However, I am uncertain about the theory of how PLLs behave when they are separated into physical components on a PCB.

After reading this question, I still am uncertain how fractional PLLs behave with lower output frequency than the reference.

Question

If I configure 2 separate (but identical) PLLs on a PCB such that:

  • I supply the same 10MHz reference clock to both PLLs
  • I control the nets on the PCB to be identical length, width, ground plane coverage, etc.
  • I configure the PLLs "identically"

After both PLLs are locked (Fout1=Fout2), will the signals Fout1 and Fout2:

  • Be identical in phase if Fout > Fin and the relationship is a fractional multiple?
  • Be identical in phase if Fout < Fin and the relationship is a fractional multiple?

In other words, is both the rising and falling edge of Fout1 aligned to the rising and falling edge of Fout2?

pll

cplusruss
  • 85
  • 3

1 Answers1

1

There is no guaranteed phase relationship, except that it should be stable once locked.

To synchronise the outputs of Fractional-N PLLs, you need to synchronise the sigma-delta modulators. If the reference frequency is low enough, and you power them up and program them at the same time, then they may reliably start in sync, but I don't know how you guarantee it.

Various manufacturers have ways of doing this, here is a random document I found on the ADI site: phase-resync-phase-programmability-and-phase-coherence-between-multiple-fractional-n-plls

Tesla23
  • 2,928
  • 1
  • 3
  • 9
  • Ah yes, phase coherence is the exact term I was looking for! With a purely integer PLL, this is much more straightforward. Thanks, Tesla. This is definitely points me on the right path. – cplusruss Sep 25 '21 at 03:34