Connect with us

PCB routing

Discussion in 'Electronic Design' started by yy7d6, Sep 3, 2006.

Scroll to continue with content
  1. yy7d6

    yy7d6 Guest

    Hi i would like to know anyone about PCB routing experience with
    Altera/Xilinx FBGA.
    My aim is to know if which is easier to route, and w/c one has lower
    layer count for 676-Ball

    Thanks.
     
  2. mostafa

    mostafa Guest

    dear sir/madam

    I am a pcb designer in high frequencies in texas university of
    technology. please give me more information to help.

    high hopes.
    M.Darvishi
     
  3. Genome

    Genome Guest

    Don't know about them.....

    If they both have 676 balls with the same dimensional spacing and size and
    you apply the same PCB layout design rules to them then they will be
    'Exactly' the same.

    Your mission, should you accept it, is to 'think' about how you might play
    with your internals, and externals, to iterate the final solution. Do not,
    under any circumstances, give it to a PCB monkey. Do the job yourself.

    I assume that since you know about beginning sentences with capital letters
    the word 'iterate' will not be a problem.

    I think it might be called Engineering.

    DNA
     
  4. John  Larkin

    John Larkin Guest

    Figure that you can get the outer two rows of balls out on layer 1,
    and the next two rows on layer 2, and one more row per layer after
    that. Then you need a ground plane and probably two power planes. The
    Alterx/Xilinx thing depends on the details.

    We did a Xilinx FG456 on an 8-layer board, with 6 rows deep of signal
    balls all around. I could post pics.

    John
     
  5. PeteS

    PeteS Guest

    Layer count is going to be about the same for both. They both use a
    full grid approach, rather than leaving the centre zone clear.

    I've put both of them down on different boards and from a PCB layout
    perspective there's not really any difference between them. The biggest
    single thing is being able to route the signals you need to the
    external balls, but if you need all of them (otherwise you wouldn't be
    in a 676) then figure 8 layers at a minimum.

    I had (on both those designs) 16 layers for other reasons, so layer
    count was not an issue for me.

    In any case, for both devices, use non-SMD (solder mask defined) pads.

    Cheers

    PeteS
     
  6. yy7d6

    yy7d6 Guest

    Genome,
    Application Notes tells that Xilinx 676 BGA was routed with 6-layers,
    while the competing 672 BGA was routed in 10-layers, have you tried any
    of these? Just want to make sure.
     
  7. yy7d6

    yy7d6 Guest

    Thanks can you show some?
     
  8. yy7d6

    yy7d6 Guest

    Hi PeteS,
    I have a gerber of an Altera 672-ball 8-layers, but Xilinx
    Applications notes says that their 676-ball can be routed with 6-layers.
     
  9. John  Larkin

    John Larkin Guest

    Can you see posts to a.b.s.e.? I could post there.

    John
     
  10. John B

    John B Guest

    To misquote Archimedes.

    "Give me tracks thin enough and vias small enough and I'll do it in two
    layers."
     
  11. PeteS

    PeteS Guest

    Application notes are the greatest lies in the electronics universe,
    and can be the greatest truths.

    What was being routed? What were the signal definitions? What impedance
    controls were required? What power? There are so many variables even
    for a single device that it's not possible to say how many layers are
    required without knowing the specifics of a design.

    Cheers

    PeteS
     
  12. John  Larkin

    John Larkin Guest

    With tracks thin enough, you could get all the signals out on one
    layer, with no vias. For my FG456, about 2 mil design rules would do
    it, but I'd still want power and ground planes with vias.

    John
     
  13. qrk

    qrk Guest

    Xilinx has an application note on suggested routing. In fact, the
    Xilinx PCB routing application note is informative for people doing
    BGA PCBs. I doubt that you will be using all the I/O pins, so you can
    adjust your number of layers accordingly. The Xilinx parts have most
    of the power/ground pins near the center. I would assume that the
    Altera part is similar which means that the number of layers and
    routing strategy will be similar. If the Altera part has a significant
    number of power pins towards the outside, that can cause routing
    problems.
     
  14. Jamie Morken

    Jamie Morken Guest

    Hi,


    Did you use blind vias on the pads, or vias "between" the BGA pads?
    I know it costs extra to have blind/buried vias, but also it could allow
    you to use fewer layers possibly. Do you guys autoroute or manually
    route? Thanks!

    cheers,
    Jamie
     
  15. John  Larkin

    John Larkin Guest

    We use dogbones, undrilled pads for the bga balls with a very short
    trace to a thru via, for all pads except the two outer rows, which are
    the undrilled pads alone.

    We mostly manually route, since autorouters usually make a bigger mess
    than they're worth. But we do assign all the fpga pins to make routing
    easy, with minimum crossovers. I usually let my layout guy assign the
    fpga pins.

    John
     
  16. Robert

    Robert Guest

    South Africa of all places.

    In looking around I found a reference on Digg that claims they are making
    Copper Indium Gallium Diselenide (CIGS). The breakthrough doesn't produce
    that much more power for the same area just (supposedly) a much lower cost
    for the Solar Panel. A factor of 4-5 is quoted in the second reference.

    http://digg.com/tech_news/Newly_created_high-efficiency_solar_panels_from_South_Africa This link has some details at the bottom. http://www.engineeringnews.co.za/eng/news/today/?show=92093
     
  17. Eeyore

    Eeyore Guest

    AFAIK these panels are still at the 'lab' stage.

    If they perform as advertised it will indeed be a major breakthrough. Since they're using Germanium I'd be concerned over their ability to take much in the way of high temperatures though.

    Graham
     
  18. Don't bother. Doing the math (50 watt panel costing $70 to build, $100
    retail), show the panel makes about $1 of electricity a year, the cost
    of borrowing the money, at least $7.

    And that doent include the cost of installation, maintenance,
    batteries, inverters, etc.
     
  19. Robert

    Robert Guest

    Germanium?

    I see a bunch of materials quoted but no Germanium.

    Perhaps you meant the Gallium?

    Quoting:
    "As it uses no silicon, costs are dramatically lower. It makes use of normal
    window glass as a substrate, with - and this is where it gets complex -
    molybdenum applied as back contact, followed by the core component, being a
    compound semiconductor comprising five elements - copper, indium, gallium,
    selenium and sulphide, replacing the silicon - with cadmium sulphide as a
    buffer layer, followed by an intrinsic zinc oxide layer and, finally, a
    conductive zinc-oxide layer."

    They mentioned a Production Plant (verified) of 25MW. That's certainly small
    compared to a commercial generation plant but I don't know if it's still in
    the "Lab" stage. If you can believe what they say.

    Robert
     
  20. Robert

    Robert Guest

    Since they quote a cost of 4 to 5 times larger for imported Solar Panels,
    and say there is a thriving Solar Industry using them, I certainly think the
    new panel will make a lot of changes in the current Industry. No one ever
    said they were a replacement for the Power Grid in First World Countries.
    And that's if you can believe what they say.

    Robert
     
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.
Electronics Point Logo
Continue to site
Quote of the day

-