0
\$\begingroup\$

For the CDCE62005 clock generator IC, what does the internal or external loop filter accomplish? Thanks in advance for any input.

Page 44 of datasheet, and pages 1-2 and 6 or the eval board user guide.

CDCE62005 Datasheet: http://www.ti.com/lit/ds/symlink/cdce62005.pdf

CDCE62005 Evaluation Board User Guide: http://www.ti.com/lit/ug/scau024/scau024.pdf

\$\endgroup\$
  • \$\begingroup\$ Can point us to what page you're talking about on the spec sheet? I honestly don't feel like going through more than 100 pages of reading to see what you're talking about :) \$\endgroup\$ – KingDuken Feb 3 '19 at 17:53
  • \$\begingroup\$ Page 44 of datasheet, and pages 1-2 and 6 or the eval board user guide. Also can control F "loop filter". \$\endgroup\$ – user127250 Feb 3 '19 at 18:16
  • \$\begingroup\$ Edit that into your question if you want readers to see it. \$\endgroup\$ – Transistor Feb 3 '19 at 18:20
0
\$\begingroup\$

Page 44 of 85 shows a 3 caps thus 3 different 1st order LPF’s filter chosen at different breakpoints . So that it starts as a 1st order LPF with R2C2 and then Xc2 drops below R2 at some frequency just below mixer f . This is essential for phase margin and jitter in the defined BW of 2x LPF BW, as 2nd order systems must never become 3rd order if loop gain >1 and the mixer converting frequency to phase error is already an integrator=k/s .

The other cap C1 is 1/10th to 1/5th smaller typically to roll off when the loop gain <1. R3C3 is usually well past 1/f of mixer to reduce jitter further but not induce much phase margin due to phase shift at unity gain.

So it starts as a 1st order filter then flattens out near 0 order then the next LPF’s take effect when the gain is near or past unity gain. The loop gain is affected by the choices of f, the product of MHz/V on the VCO and N divide ratio for multiplying the mixer f to VCO output and phase detector gain in V/cycle

These choices tradeoff jitter, capture to lock time, clock SNR and noise BW

\$\endgroup\$
  • \$\begingroup\$ Thanks for the reply. So if I want a 1 GHz clock from the VCO, what would be a proper configuration for the loop filter? \$\endgroup\$ – user127250 Feb 3 '19 at 18:25
  • \$\begingroup\$ Let’s see you do some Bode Plots 1st for amplitude and phase or define every tradeoff I mentioned as a spec. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Feb 3 '19 at 18:28
  • \$\begingroup\$ I'm learning/refreshing some of the material related to these questions as I go, so it will probably take me awhile. But I appreciate your input so far. \$\endgroup\$ – user127250 Feb 3 '19 at 18:34
  • \$\begingroup\$ You can get formulae but best to define your jitter and lock time 1st \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Feb 3 '19 at 18:52
0
\$\begingroup\$

Loop filters can be 3rd or 4rth order, producing 270 or 360 phase-shifts, but the loop gain must be LARGE in the regions of high phase shifts, to prevent oscillation.

G_large at phase_shift_large

=====================================

1 + G_large at phase_shift_large * H

may be quite stable, because the G_large is in charge and lets the "H" set the behavior.

When the "G" is low, then the "1" plays a prominent part in the behavior of the closed loop stability.

Its no accident that Harry Nyquist used "1" in his stability criterion.

\$\endgroup\$

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.