I'm curious about this situation and how I would go about understanding it better. I have an input clock of 100Mhz with a maximum total jitter of 1ns. I want to put that into a pll to double the frequency to 200Mhz. My hope is that after lock the 100Mhz and 200Mhz clocks will never get out of sync by a full bit period of the 200Mhz clock.

I feel like over time as the input jitter and output jitter adds up that the 200Mhz clock will slip to the next bit. I know this kind of things happens in serial links where over time the your jitter closes your eye and you eventually get a bit error.

Will that sort of thing happen with a pll or will these two clocks always stay in sync?

  • 3
    \$\begingroup\$ If the PLL control loop is properly designed you should have no issues, a clock doubler PLL is pretty straightforward and has no reason for the reference to drift with time relative to the output. \$\endgroup\$ – John D Nov 10 '14 at 23:07
  • 1
    \$\begingroup\$ I think you are confusing jitter with drift. Jitter will not cause a problem over time because it is a variation centered on the nominal value. Drift, on the other hand, is a variation that moves the nominal value as time progresses. In any case, even if the 100 MHz clock drifts from 100 MHz, the 200 MHz clock, since it is phase locked to it, will simply also drift in the same direction but will stay in sync. \$\endgroup\$ – Barry Nov 11 '14 at 0:59

If the input frequency is approximately a sinewave ie not rich in harmonics, you could make a frequency doubler by using a fast full wave rectification circuit and a fast comparator. This might be easier and more reliable than a PLL if the jitter is significant.

Without knowing the peak to peak jitter it's really hard to say but if the 1ns figure you gave is p-p (rather than RMS) then pretty much any PLL circuit designed correctly will work.

| improve this answer | |
  • \$\begingroup\$ Frequency doubler also works with square waves, provided they have good (50%) mark-space ratio. This variety uses a 90 degree delay (2.5 ns here) and an XOR gate. Credit : Peter Alfke. pldworld.com/_xilinx/html/tip/sixeasypieces.htm \$\endgroup\$ – Brian Drummond Nov 11 '14 at 16:21
  • \$\begingroup\$ @BrianDrummond I forgot about the XOR doubler. Good man. \$\endgroup\$ – Andy aka Nov 11 '14 at 20:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.