CDCE706: Used as a Tracking Clock Over an Octave+

Part Number: CDCE706
Other Parts Discussed in Thread: CDCE6214

Tool/software:

Hello E2E Experts,

Good day.

June 25, 2024
I am designing a general-purpose bit synchronizer that receives an analog bit
stream of data (typically contaminated with noise) and outputs digital data with
a clock that can be used to sample the digitized data.
The design requires a clock with the following requirements:
1. Frequency range: 100 MHz to 320 MHz
2. Frequency resolution: better than 50 ppm
3. Frequency tracking range: +/- 10%
a. The initial programmed frequency should track changes in the
analog input data rate without any glitches or phase
discontinuities.
b. The rate of the frequency change will be less than
1%/ millisecond.

We are considering the use of the TI CDCE706 with an AD9850 DDS from
Analog Devices. The DDS has very high resolution and can provide a high
quality clock that can be varied, for example, from 6 MHz to 16 MHz.
Frequency tracking will be performed by connecting the CDCE706's clock
input to the DDS's output (note that the AD9850 can output a digital clock).
This will cause the VCO to smoothly track small changes in the DDS
frequency.

The CDCE706 data sheet states that the Normal Speed VCO range is 80
MHz to 200 MHz and that the High Speed VCO range is 180 MHz to 300
MHz. However, note 3 on page 7 states that the VCOs can go lower in
frequency with a degradation of phase noise. Figure 29 on page 33 shows
that the Normal Speed VCO can actually range from less than 60 MHz to
more than 240 MHz and that the High Speed VCO can actually range from
less than 100 MHz to more than 320 MHz. Can these extended ranges be
guaranteed? The degradation in phase noise has no significance for our
application.

Please let me know if the above solution is viable or problematic.

Regards,

CSC

  • CSC,

    We cannot guarantee these extended frequency ranges. We have only validated normal part operation over the ranges specified in the datasheet.

    Thanks,
    Kadeem

  • Hello Kadeem,

    Good day.

    Our application requires that the VCO cover an octave (with spare) without sub-bands or any kind of switching.

    So just to be sure:

    Can the CDCE706 output continuously track an input reference that varies continuously from 5.625 to 9.375  in order to output

    5.625 x 32 = 180 to 9.375 x 32 = 300 MHz without any "hiccups"?

    Regards,

    TI-CSC

  • Hello,

    The clock gen team is out for US holiday. We will return Monday, July 8. Thank you for your patience. 

    Best,

    Cris

  • TI-CSC,

    The lock time is detailed in the datasheet on pages 35 and 36, and provided here as well:

    Thanks,

    Kadeem

  • I have read pages 35 and 36 carefully, but the graph does not answer my question. 

    It shows small jumps in frequency over a span of 60 microseconds, but does not describe phase changes.

    I need to better understand what happens during the lock-time.

    For example, suppose that the input reference frequency is changed continuously at a rate of 1%/millisecond.

    If we look at any 2 neighboring output clock  periods,  will their duration differ by more than say 0.01% ?

  • Dave,

    We will take a look on the bench and get back to you Thursday.

    Best,

    Cris

  • Dave,

    I used the CDCE6214 as the input to the CDCE706, and changed the output divider on the CDCE6214 (which is glitchless) to generate the 6MHz to 16MHz clock. I have the CDCE706 set up such that the full range from 6MHz to 16MHz input is supported (120MHz to 320MHz output). When varying the clock from the CDCE6214, I do not observe a drop of the clock from the CDCE706. 

    The below images show the periods of adjacent clock cycles while the PLL adjusts to the new input clock frequency (from 16MHz to 7.4MHz).

    Thanks,

    Kadeem