HMC394LP4, CML Interfacing to Inputs

What is the common mode voltage of the inputs on pins 13 and 14. Do I have to terminate the CML inputs to Vcc and AC couple? Then do I have to provide a bias to the inputs? Also, in the pin description for pin 13, it says "RF input 180° out of phase with pin 14 must be AC ground." What does that mean?? You have to AC ground an input? Is it impossible to put in a differential signal? It sounds like it was written by a person which does not understand English very well.

Parents
  • 0
    •  Analog Employees 
    on Oct 16, 2018 3:42 PM over 2 years ago

    Hello Mr Berris,

    I apologize for the delay in answering this and the state of the datasheet.  I've confirmed with design that the resistor between pins 13 & 14 (!IN and IN) is a 50 ohm resistor.  This implies this device was targeted at single ended 50 ohm environment and is not the best selection for a 100 ohm differential input.  I think I would stick with the Single Ended schematic on page 4-6 of the HMC394 datasheet, assuming your CML device can support SE operation.

    The differential CML devices I've used can support single ended operation as shown below.  Is this an option for the CML device your using?

  • Yes, I can use single ended CML and am doing so. (1) I don't believe that the /IN pin should be called an input. It's actually an output / fixed voltage. It's not even good for biasing. (2) On you're diagram, The terminated CML- output should not be AC coupled to a 50Ω resistor to ground. It should go to +Vcc. Otherwise, there will be a differential spike on the power supplies during switching. CML works fast because there are no differential currents during switching because the sum of both + and - termination currents is zero.

  • In order to improve the input to the HMC394, it would be better to have 50Ω resistors from the internal bias point to each of the inputs. Then it would be a true differential input. I don't know why they wouldn't do that. Maybe a layout problem on the die??

  • 0
    •  Analog Employees 
    on Oct 16, 2018 8:55 PM over 2 years ago in reply to richardberris

    Some of the other HMC dividers have an input network similar to what you described (as an example HMC705 or HMC492), but this part went a different route for some reason.  I also would tend to agree with you the /IN naming convention.

    In regards to the diagram, the reason I recommended cap to GND was the following.  

    Consider the SE CML+ signal to the HMC394.

    CML+ out --> AC coupling CAP --> PCB trace --> HMC394 IN pin into device --> internal HMC394 50 ohm --> HMC394 /IN pin out of device  --> AC coupling to GND. 

    It could make sense (especially if its a long trace) to reference this signal to GND.  I suspect it will be easier to reference this trace to a clean GND plane rather than a clean Vcc plane, especially if it is a long trace between the two devices.  Once the signal is AC coupled, both Vcc and GND act as an AC GND, so I thought it made sense to use the plane that is most likely the cleanest.   As result the CML- can follow suit with ac-coupling to GND.

    I just want to relay the thought process.  I'm not saying one way is better than another, just something to think about.  If anything else its an interesting discussion...

Reply
  • 0
    •  Analog Employees 
    on Oct 16, 2018 8:55 PM over 2 years ago in reply to richardberris

    Some of the other HMC dividers have an input network similar to what you described (as an example HMC705 or HMC492), but this part went a different route for some reason.  I also would tend to agree with you the /IN naming convention.

    In regards to the diagram, the reason I recommended cap to GND was the following.  

    Consider the SE CML+ signal to the HMC394.

    CML+ out --> AC coupling CAP --> PCB trace --> HMC394 IN pin into device --> internal HMC394 50 ohm --> HMC394 /IN pin out of device  --> AC coupling to GND. 

    It could make sense (especially if its a long trace) to reference this signal to GND.  I suspect it will be easier to reference this trace to a clean GND plane rather than a clean Vcc plane, especially if it is a long trace between the two devices.  Once the signal is AC coupled, both Vcc and GND act as an AC GND, so I thought it made sense to use the plane that is most likely the cleanest.   As result the CML- can follow suit with ac-coupling to GND.

    I just want to relay the thought process.  I'm not saying one way is better than another, just something to think about.  If anything else its an interesting discussion...

Children
  • I agree with you that the problem is interesting. No where on the data sheet does it say what that Pin 13 to Pin 14 resistor is. It's fair to assume that it's 50Ω. It should be. I'm now thinking that the most correct way to satisfy the CML outputs and possible noise spikes is for the CML- output to be AC coupled and 50Ω resistor terminated to the HMC394 /IN terminal. The /IN terminal should also be AC grounded. If you ignore the diff amp input transistors, the currents will sum to zero at pin 13. Everything else should be happy. No spikes anywhere. I think that's what I will do in my circuit.

    By the way, I measured the resistance between pins 13 and 14 with the power off and my ohmmeter read 52Ω. I hope I didn't damage the IC.

  • 0
    •  Analog Employees 
    on Oct 17, 2018 3:10 PM over 2 years ago in reply to richardberris

    Seems reasonable to me.  Again I apologize for all the confusion we created for you.