This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMK04832: Synchronizing outputs of Single LMK04832 and clock tree verification for three ADC12DJ3200 synchronizations and deterministic latency

Part Number: LMK04832
Other Parts Discussed in Thread: ADC12DJ3200, LMX2594,

Hi TI,

We are using three ADC12DJ3200 in our design.

For ADC's device clock we are using one LMX2594 and ,one HMC987(buffer) to buffer the output of LMX2594 and feed the device clock to three ADCs.

Sysref and FPGA clocks are provided by LMK04832 only.

The clock tree is shown as below:

The dotted lines signals are optional and are not connected default.

In the above image SYNC of LMK04832 is connected to FPGA.

We need to synchronize the all the outputs of LMK04832 and LMX2594/HMC987.

Question 1:

Can we do synchronize/fixed relatioship between LMK outputs when SYNC is controlled by FPGA and there is no relation between input reference clock to LMK04832 and SYNC from FPGA?

I am asking this because, I have read on this forum itself that SYNC should meet setup hold time with respect to input reference clock.

Question 2:

Also we want to align/synchronize the LMX2594 ADCs sampling clocks with SYSREFs clocks from LMK04832. So we have connected LMX2594 sync pin with SDCLKout of LMK04832.

So we are assuming that :

SYNC from FPGA will give a fixed relationship/synchronization between all outputs of LMK04832 and SYNC from LMK04832's SDCLKout to LMX2594 will give a fix relationship/synchronization between LMX2594 output and outputs of LMK04832.

So, if we follow above scheme there will be a fixed relationship between all clocks, which is a requirement for achieving synchronization between all three ADCs and also for deterministic latency. 

Is our understanding correct?

Question 3:

This is regarding the SYNC for LMX2594.

Is the SYNC from LMK04832 to LMX2594 need to be generated from VCO using SYSREF dividers or the SYNC which is coming from FPGA to LMK04832 SYNC pin need to be routed to LMX2594 internally through LMK04832?

Question 4:

This is regarding the dotted lines signals in the clocking tree diagram.

We are also planning to give an option to use LMK in distribution mode only by power down both PLLs in LMK.

Actually since it is written in LMK's datasheet that in single loop mode, reference clock for PLL2 can be taken from PLL1(CLKinx) clock inputs, thats why if we need to use PLL2 only of LMK04832 then we can take reference clock from CLKin0 as shown in above diagram.

Now the clock required at CLKin1 pins for using LMK in distribution mode need to be generated from LMX2594.

But the catch here is that we cannot lock the LMX2594 PLL with external reference clock in distribution mode since in distribution mode LMX2594's reference clock is drive from OSCout, which is buffer copy of OSCin, of LMK04832 and only VCXO is connected to OSCin of LMK04832.

Is there any way to connect reference clock input at CLKinx(PLL 1 inputs) to OSCout pins of LMK04832 so that we can use LMK in distribution mode by locking LMX2594 to an external reference clock.

An early response will be highly appreciated.

Thanks,

Lalit 

  • Hi TI,

    Can you please update on this?

    We want to close the clocking scheme and release the gerber as early as possible.

    Thanks,

    Lalit

  • Hello Lalit,

    1. Applying the synch input asynchronous to the reference will work to have teh outputs phase deterministic.We give some guidelines so no glitches, seen on the output. The timing is random you may have a runt pulse on one or more of the outputs.

    2. Yes, this is correct in that when the LMK04832 output drives the LMX2594 there will be a deterministic phase relationship between the LMK04832 outputs and the LMX2594 outputs once the LMX2594 is synchronized. The propogation delay between the paths will not be equivalent however, so expect some differences in the phase variation of the LMK04832 outputs and the LMX2594 across temperature.

    3. If the SYNC pulse for the LMX2594 is generated from the LMK04832 sysref output in single pulse mode this will allow for aliging the synch pulse input to the OSC input so the timing of the SYNC to OSCin is deterministic.

    4.In the LMK04832 distribution mode, only CLKin1 is active as input. so it is not possible to feedback another frequency domain

    Regards,

    Liam

  • Hi Liam,

    Thanks for the reply.

    I need some clarification on the answers:

    Liam Keese said:

    1. Applying the synch input asynchronous to the reference will work to have teh outputs phase deterministic.We give some guidelines so no glitches, seen on the output. The timing is random you may have a runt pulse on one or more of the outputs.

    What do you mean by "runt pulse on one or more of the outputs"?

    Will this runt pulse available on outputs for the duration when synchronization process is going on or runt pulse will also available on outputs even after synchronization is completed?

    Will this runt causes any issue in achieving synchronization/ fixed relationship between outputs and if so how we can avoid this?

    Liam Keese said:

    4.In the LMK04832 distribution mode, only CLKin1 is active as input. so it is not possible to feedback another frequency domain

    My query was, there any way to route signal connected to CLKin0 and CLKin1 pin to OSCout pin?

  • 1, Start an asynchronized SYNC trigger, would generate a non-completed cycle. That's the runt pulse.

    In SYNC, mute all output, no runt pulse.

    Exit SYNC, output normally, no runt pulse

    2, There is no internal path for CLKin0/CLKin1 to OCSout in LMK04832.