LTC6953 Output Termination

Hi all

i want to use the device in my design. the output need to be configured as AC coupled CML. the datasheet shows the output termination shall be a parallel 100ohm resistor and then 2 series capacitors.

is the resistor a source termination? or is it a mistake and the resistor need to be near the load?

Parents
  • 0
    •  Analog Employees 
    on Dec 2, 2020 10:16 PM 2 months ago

    Figures 36 and Figure 38 in the LTC6953 datasheet describe this better.    The 100 ohm resistor and the caps need to be near the load.  However, we are recommending the 100 ohm before the caps.  This helps with the first clock pulse generation after a sync operation.

  • the load is a RFSoC which is, a BGA chip. so, placing a 100 ohm resistor is easy (even not needed because ODT feature), but the series capacitors (even in 0201 package) cannot be placed under the RFSoC i.e. near the load (no place for VIAs) so, i have to place the capacitors near the RFSoC (but not near the relevant balls i.e. load). in that case, do i need to place the resistor before the capacitors? it seems like not a good idea. can you advise what to do?

  • +1
    •  Analog Employees 
    on Dec 8, 2020 1:25 PM 2 months ago in reply to Shimon_Otsri

    Hello, if you only need a continuous clock and are not concerned about the exact timing of the very first clock pulse you can place the 100 ohm resistor on either side of the caps close to the load as possible.  Or if the RFSoC has an internal 100 ohm resistor that works also.

    When the 100 ohm resistor is on the LTC6953 side of the caps the correct differential voltage is created when the clocks are stopped.  So the first clock cycle can starts as expected.  IF the 100 ohm resistor is on the load side of the caps, then the LTC6953 differential voltage drifts to 3.3V when the clocks are stopped, so sometimes the first clock pulse will be corrupted since the 6953 output transistors  are compressed, but after the first clock cycle it works fine.

    Very few customers are concerned about the first clock cycle.  So most likely the 100 ohm location can be on the RFSoC side, but you'll need to let me know if this true.

Reply
  • +1
    •  Analog Employees 
    on Dec 8, 2020 1:25 PM 2 months ago in reply to Shimon_Otsri

    Hello, if you only need a continuous clock and are not concerned about the exact timing of the very first clock pulse you can place the 100 ohm resistor on either side of the caps close to the load as possible.  Or if the RFSoC has an internal 100 ohm resistor that works also.

    When the 100 ohm resistor is on the LTC6953 side of the caps the correct differential voltage is created when the clocks are stopped.  So the first clock cycle can starts as expected.  IF the 100 ohm resistor is on the load side of the caps, then the LTC6953 differential voltage drifts to 3.3V when the clocks are stopped, so sometimes the first clock pulse will be corrupted since the 6953 output transistors  are compressed, but after the first clock cycle it works fine.

    Very few customers are concerned about the first clock cycle.  So most likely the 100 ohm location can be on the RFSoC side, but you'll need to let me know if this true.

Children
No Data