Using ADN4650 for Isolating TMDS

Thank you for reading below my inquiry.

I'd like to use ADN4650 for Isolating TMDS (of HDMI by mainboard in computer), but ADN4650 is for isolating LVDS.

I found that TMDS's electrical specification is CML.

So, I need to converting circuits both CML->LVDS and LVDS->CML.

I'd like to designing compact circuit by passive parts like Resistor or Capacitor.

Motherboard(TMDS output of HDMI) -> (CML) -> Passive Circuit1 -> (LVDS)

-> ADN4650 -> (LVDS) -> Passive Circuit2 -> (CML) -> TV(TMDS input of HDMI)

Is it available for realizing this concept?

In Google, I found many solutions.

But, I want to find compact and proper solution for ADN4650.

How can I realize the schem for using ADN4650?

  • 0
    •  Analog Employees 
    on Feb 8, 2018 10:19 PM


    ADN4650 can indeed be used to isolate HDMI, using appropriate coupling networks for TMDS/CML to LVDS and vice versa as you suggest. We will have a live demonstration of this at Embedded World show in Nuremberg, Germany, from 27 Feb to 1 Mar, at the Analog Devices booth.

    Some application-level circuit design and verification may be needed depending on the purpose (e.g. cable quality/length, HDMI sink capabilities), but the circuits below have been verified to allow isolated HDMI using ADN4650. AC coupling is required in both cases, and then for connecting to an LVDS receiver, the signal is attenuated and a common mode introduced (via resistor network on Din-). Conversely the LVDS outputs have a termination network to allow operation with some CML receivers.



    I'd recommend start with the components above, and then you can analyze the signal and see if performance is as desired for the application. If needed, the coupling networks can be adjusted further, or indeed a HDMI re-driver could be added.

    Best regards,

    Conal Watterson

  • Thank you for answering.

    Howerver, I'm sorry that I can't understand a little bit about your answer.

    Q1. In 1st figure, I found that common-mode introduced by resistor network only Din- Line.

       I heard that common mode voltage of LVDS is 1.2V.

       So, I re-ask about that. How can it be available by this scheme?

     Q2. In 1st figure, I found that power used for getting common-mode voltage is +5V_ISOPOWER.

         Does 5V_ISOPOWER means isolated 5V power or non-isolated 5V power?

       I want to isolate between Left side(suffix 1 Side) and Right side(suffix 2 side) of ADN4650 to all signal and power and ground.

     Q3. In 2nd figure, I also found that common-mode introduced by resistor network only TMDS- Line.

       As I mentioned like Q1, I re-ask about that. How can it be available by this scheme?

    Q4. Does the scheme must need AC coupling for isolating HDMI using ADN4650?

  • 0
    •  Analog Employees 
    on Feb 12, 2018 2:53 PM


    Sorry, I didn't take enough time to correctly identify the sections of circuit I posted. The first one is actually LVDS to TMDS.

    The second circuit has AC coupling on the TMDS input, then to introduce the LVDS biasing, the pull-up and pull-down on Din-. And to answer your Q1, now we can see it introduces a common mode due to Din+ being tied to Din- with 200 Ohm (split termination with common mode biasing could also be used, but the biasing on Din- should be just as effective).

    Regarding Q2, our circuit assumed full isolation of a HDMI source. So on the isolated (cable) side, you need a power supply for isolator Vdd2s, as well as 5V/100 mA for HDMI to the sink. If you have sufficient 5V power on the source circuit, you can isolate using ADuM5000 x2. One ADuM5000 will provide the 5V/100 mA for the HDMI sink, and then the other will provide power for ADN4650 x2 and ADuM1250 x2. The latter will use a 5V supply, and the ADN4650 will require 2.5V on Vdd2 (ADP5302 will efficiently convert the voltage, allowing isoPower to be 5V->5V instead of the less efficient 5V->3.3V and then avoiding the LDO drop 3.3V to 2.5V using Vin2 to Vdd2 on ADN4650).

    For Q3, but looking at the first circuit, I think actually R5/R7 may not be required. The CML input at the HDMI sink will reference both TMDS+ and TMDS- to 5V.

    AC coupling is required as the CML/TMDS signals in HDMI are referenced to 5V, while LVDS receiver and driver operate at lower voltages, and either require a common mode around 1.2V on the input, or provide one on the output.

  • Thank you for reply.

    I have more question about that.

    LVDS-to-TMDS/CML(First picture corrected)           (As you uploaded)

    TMDS/CML-to-LVDS (Second picture corrected)        (As you uploaded)

    Q5. Is TMDS referecenced by +5V?
        Could I use TMDS reference voltage 3.3V?
        (+5V_ISOPOWER => +3.3V_ISOPOWER,      +5V => 3.3V)
        and it looks like that TMDS/CML use 3.3V as below link (AVCC).

    Q6. I will use long length (about 5m) HDMI Cable.
       If I need tune cirucuit by long length, Which resistors could I change?

    Q7. You mentioned "LVDS-to-TMDS/CML(First picture corrected)".
        Do I need to 100ohm termination between TMDS+ pin and TMDS- pin of input side of ADN4650?

    Q8. I'd like to isolate TMDS Clock Line also.
        Could I use ADN4650 for isolating TMDS Clock (No TMDS Data) with proper passive component network converting CML(TMDS) to LVDS?
        If yes, I will use two ADN4650 for connecting below.
         HDMI CLK+- => (Passive Component Network) => ADN4650(#1) DIN1+-
         HDMI   0+- => (Passive Component Network) => ADN4650(#1) DIN2+-
         HDMI   1+- => (Passive Component Network) => ADN4650(#2) DIN1+-
         HDMI   2+- => (Passive Component Network) => ADN4650(#2) DIN2+-
        Is it available for connection?

    Please reply for me

  • 0
    •  Analog Employees 
    on Feb 27, 2018 3:59 PM


    I checked with colleagues who support our HDMI products; the TMDS lines do use CML referenced to 3.3V, as you suggest. We'll adapt our solution with ADN4650 to match that.


    I'm not sure this will work well with a long cable, due to the capacitive load and the AC coupling already used. I don't have a recommendation here right now, further testing would be required.


    In picture two above (TMDS to LVDS), the input side of ADN4650 is terminated by the complete network of the 100 Ohm pull-ups, the AC coupling, and the 200 Ohm differential (this would normally be 100 Ohm, except we are adjusting to account for the rest of the coupling network).

    In picture one (LVDS to TMDS), the assumption is that the output of ADN4650 will be terminated at the CML receiver which would have internal pull-ups to Vcc (should be 3.3V as you mention).


    TMDS clock can be isolated the same way as the data lines, but you can potentially adjust the component values (e.g. AC coupling capacitor) as the speed is 10x less than the data lines.

    Two ADN4650 will cover all four TMDS lines as you suggest.

    One final note - while this circuit might allow isolation between two HDMI nodes, it's not clear yet if it would allow an isolated port that's fully HDMI compliant. That's due to the AC coupling affecting the signal levels (required because of the LVDS I/O on the isolators). Regarding jitter, the precision timing of ADN4650 should still allow full compliance. Using a HDMI redriver at an external port, after the isolation circuit, might take care of compliance to all specifications including signal amplitude.