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.

Using TLK100's in a simple point to point range extender application

Other Parts Discussed in Thread: TLK100, TLK110

I guess Mr. Ofir Ben-Josef isn't with TI anymore since his email bounces.
Anyway, I'm looking to make an Ethernet extender for use between IP cameras and a switch.  I see your dual PHYs that support this mode, but those PHYs are only rated for 137 meters; therefore, I'd like to use two TLK100's because they tout 200 meters under typical applications. 
 
I think I may be able to just cross couple the MII Tx signals to the Rx signal besides the clocks as shown below or use the far-end loopback mode.
Option 1: more complicated, but I saw something like this in AN-1509 section 3.4
Phy 1       Phy 2
TXD[3:0]    RXD[3:0]
TX_EN       RX_DV
RX_DV       TX_EN
RXD[3:0]    TXD[3:0]
CRS         not used on either PHY
COL         not used on either PHY
TX_CLK      not used since both PHYs will be on same crystal/clock
RX_CLK      not used since both PHYs will be on same crystal/clock
I plan on using CLK25OUT to drive the XI pin of the second phy.
Option 2: preferred if possible
I was also thinking I may be able to use the Far-End loopback mode with each chip acting as a repeater in one direction which would save many external connections.  For the clock I would again use the CLK25OUT to drive the XI input of the second PHY.
 
If you can comment on the two options and correct any of the MII connections and clocking it would be much appreciated. 
I’m assuming I need a simple MCU to program the devices with either option, but if they could be hardwired please let me know.  I do want to be able to force setting 10M FD, but I think that is pin strappable, but not the far end loopback mode.
Other aspects of the design are that I plan on using the chip in a single 3.3V supply mode if that matters.  The customer application will always use two of these circuits so that both the local and remote ends of the link have this higher performance PHY so the application would be using 4 of the TLK100's overall.  I'm willing to pay a little more for longer lengths.
Since I'm controlling the interface between the intermediate link of the two PHY's all time do you think changing each TLK100's transmit amplitude to be larger than the normally specified levels will help improve performance and if so can this be done by changing the value of the Rbias resistor connecting to pin 3?
 
Any help you can provide will be much appreciated,
 
Jeff Lee
  • Hi Jeff,

    Option 1 :

    not recommended since RX_CLK is a remote node recovered clock, and TX_CLK is the PHY (TLK) Free-Running Clock.

    due to PPM differences between remote node clock and TLK clock ( CLKOUT25 is also based on free- running clock),  this scheme will not work.

    I can offer new scheme, using RMII Interface ( support by TLK110)  so you can connect two PHY sharing the same XI and connect TXD to RXD and RX_DV to TX_EN between them.

    In RMII PPM difference between remote node and local node XI clock is compensated internally using a Async_fifo.

    Option 2:

    It seem like very good and creative idea, using far end loop for each direction will cancel the need to use MII Interface ( you can left the RX and TX unconnected

    And avoid using the CLOCKS since it is an internal loop back).

    Beside of carful routing of the PHYs TD and RD pairs ( Node1 is transmitting to PHY1, PHY1 receives and looping it out to its TD pair that should be connected to

    Node2 RD pair and vice versa) a MDIO support is needed, since you have no other option to put TLK in loop mode.

    -          Setting the device into FORCE_10_FD is possible using Strap pins.

    -          Changing the pin3 voltage is not recommended. TLK should support 200 meters by default.

     

    Best Regards,

    Yaniv

     

  • Yaniv,

     

    Thanks for the quick feedback.  I was hoping option 2 was possible. 

     

    Regards,

     

    Jeff Lee