1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

Problems with Ethernet PHY

Discussion in 'Embedded' started by Rene, Apr 21, 2004.

  1. Rene

    Rene Guest

    Hello,

    I am debugging a nasty problem related to a 10/100 Ethernet PHY. While the
    design works well in 80-90% of the tested combinations, I have problems
    receiving data at 100 Mbps in other combinations.

    Extensive research so far showed the following phenomenons:

    - 10 MBps is always working properly
    - Tx is always working. I receive each packet on the PC I ping.
    - I get Rx errors during packet receiption. Longer packets are often faulty,
    while short packets (say < 100 bytes) seem to pass most of the time.
    - The PHY "sees" the Rx Errors. It's RxError Counter reflects this.
    - Tweaking the crystal frequency (25MHz) to approx. +16ppm above nominal did
    not help.
    - Replacing line termination resistors (the 49.9 Ohms) with other values
    never showed a difference.
    - Playing around with case ground did not show a difference.

    And now comes the part I don't understand.
    - Adding 100m of Cat6 cable makes the whole link work perfectly for several
    hours. Use a 5m cable and it fails again.

    I know it is hard, but does anyone have an idea what could go wrong here.

    Yours
    - Rene

    PS. The PHY is a National DP83847, connected to a MPC859.
     
    Rene, Apr 21, 2004
    #1
    1. Advertisements

  2. Rene

    Tauno Voipio Guest

    One cabling trouble that behaves like that is the cross-pair problem:
    the receive line does not have the signal and ground lines properly
    paired (twisted together). 5 meters is a suitable length to create
    standing-wave effects with the 125 Mbit/s data stream of 100BASE-TX
    (62.5 MHz, just a bit less than 5 meters in free-space wavelength).

    HTH

    Tauno Voipio
    tauno voipio @ iki fi
     
    Tauno Voipio, Apr 22, 2004
    #2
    1. Advertisements

  3. Rene

    David Guest

    Hi,
    had the same problem with an AMD ethernet controller.
    The reason was a bad layout of the pcb. Consider that this is a high
    frequency
    application! Did you check the layout recommendations for your chip?

    David
     
    David, Apr 22, 2004
    #3
  4. Rene

    Rene Guest

    The PCB has a population option for either a Kendin Switch or the
    National single port PHY. The connection between PHY (or Switch
    port) to the RJ45 (with integrated magnetics) is about 10mm.
    The stub length from the Kendin port to the National PHY is no more
    than 3mm.

    I checked the layout several times and reworked/patched one board
    extensively. It never made a difference...

    The layout works perfectly well for the Kendin, and for most
    combinations of the National PHY with other link partners. However
    some combinations just produce Rx errors all the time.

    I got from National a software patch to enable a different algorithm
    in the PHY. The exact change is unknown to me, but it should
    improve receiver reliability. In my case it fixed some combinations
    while others got worse :-( We are now building some boxes using
    a newer version of the PHY chip. According to National the new
    mask version should be more reliable when using short cables.
    We will see.

    - Rene
     
    Rene, Apr 22, 2004
    #4
  5. Rene

    Rick Merrill Guest

    Rene wrote:
    ....
    Perfectly reasonable: it may change your total length of LAN. To
    keep 100mb working you must stick to the standards, even if things
    "seem" to be working fine some of the time. (ok, call it magic if
    you want) But rules is rules. Many times
    I've seen ethernet wiring that would work for 10mb but not 'always'
    for 100mb because the correct pairs were not wired together. (ok, I
    did it myself the first time.) - RM
     
    Rick Merrill, Apr 23, 2004
    #5
  6. Rene

    Rene Guest

    Hi Rick, what do you mean with "perfectly reasonable". Of course I
    checked the cables. They are all CAT5 UTP (or STP) where 1,2 and
    3,6 are properly twisted and connected (the other wires also).
    I also use the same cables for other devices, which work perfectly
    well.

    Do you have any special rules in mind (i.e. minimum cable length).
    Can you give me more details.

    Thanks
    - Rene
     
    Rene, Apr 23, 2004
    #6
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.