1. Welcome to our site! Electro Tech is an online community (with over 100,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

CY7C68013A layout on TWO-layer board, instead of FOUR

Discussion in 'Microcontrollers' started by Alex_rcpilot, Apr 6, 2007.

  1. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Hi all, I'm trying to build this USB2.0 board with a cypress CY7C68013A in TQFP128 package. According to the datasheet, I have to use at least a four-layer board to ensure the performance. And then it gives several pieces of recommendations on how to handle this four-layer board.

    The problem is, this USB2.0 stuff is the only thing that requires extremely high EMC standard, and that I don't think I will waste the whole board for this minor area. I want to use a two-layer board, and I don't expect too much quality loss, is it possible?

    Here are the recommendations from the datasheet:
    The following recommendations should be followed to ensure
    reliable high-performance operation.
    • At least a four-layer impedance controlled boards are required
    to maintain signal quality.
    • Specify impedance targets (ask your board vendor what
    they can achieve).
    • To control impedance, maintain trace widths and trace spacing.
    • Minimize stubs to minimize reflected signals.
    • Connections between the USB connector shell and signal
    ground must be done near the USB connector.
    • Bypass/flyback caps on VBus, near connector, are recommended.
    • DPLUS and DMINUS trace lengths should be kept to within
    2 mm of each other in length, with preferred length of
    20–30 mm.
    • Maintain a solid ground plane under the DPLUS and DMINUS
    traces. Do not allow the plane to be split under these
    traces.
    • It is preferred is to have no vias placed on the DPLUS or
    DMINUS trace routing.
    • Isolate the DPLUS and DMINUS traces from all other signal
    traces by no less than 10 mm.

    If I use a two-layer board, will it be as simple as following EVERY BUT the first instruction? Thanks for your time.
  2. ericgibbs

    ericgibbs Well-Known Member Most Helpful Member

    Joined:
    Jan 4, 2007
    Messages:
    19,849
    Likes:
    521
    Location:
    England.
    ONLINE
    hi alex,
    What are the dimensions of the pcb area that will require a four layer pcb layout.?

    You could consider the main pcb as being double layer and a small 4 layer 'piggy back' pcb for the USB2 module.

    Leave sufficient area clear on the main pcb for the 4 layer pcb.
    The 4 layer pcb would be mounted on pickup pins on the main pcb.

    I have used a similar method for mounting smaller pcbs to a main pcb.

    If this explaination is not clear enough, please ask.
  3. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Hi, thanks for your advice. You've been illustrative and I totally understand what you mean. I just had this 6 layer ARM9 PCB shipped here which I can plug into my test board. It's a lot cheaper than a big ARM9 board with everything on it.

    I did think about stacking a mini 4-layer PCB above a motherboard. But the USB interface is the only thing in my system that requires a 4-layer PCB, i.e, this 4-layered area only covers a corner of the TQFP128 chip, and extends less than 2 inches beneath D+, D- tracks, to the USB connector. There's a diagram in the attachment which shows its shape.

    Such a small area is worthy of the 4-layer material, but it's no right way building a circuit board. To carry the chip, this mini board has to cover at least the area of the entire TQFP128 area, and then there comes the problem with the rest of the pins. So either I implement a 4-layer PCB with a lot of chips that don't actually deserve it, or I keep everything including the USB2.0 tracks on a 2-layer PCB, and take extreme precautions to maintain signal integrity.

    Do you think it has to take a 4-layer PCB for USB2.0 layout? Thank you for your time.
  4. 3v0

    3v0 Coop Build Coordinator Forum Supporter

    Joined:
    Jul 14, 2006
    Messages:
    9,358
    Likes:
    213
    Location:
    OKLAHOMA USA
    Alex,

    Maybe this is old hat but.

    Have you seen this thread over on the eda board about building ICD2s using a PIC 18F4550 instead of a cypress chip ? It is way long but contains at least two working designs including PCB layout.

    EDIT: The above link points to the middle of the thread. You should start reading from the first post. They guys there will get upset if you ask a question already answered.
    Last edited: Apr 7, 2007
  5. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Thanks 3v0. I'm checking this link out. And I'll be careful with those ppl there.

    BTW, 18F4550 is USB2.0 compatible, but it doesn't work at high-speed. The CY7C68013A complies with USB2.0 spec and also works at 480Mbps high speed, hence requires something at a higher std?

    I'll return after reading the thread. Thank you again.
  6. 3v0

    3v0 Coop Build Coordinator Forum Supporter

    Joined:
    Jul 14, 2006
    Messages:
    9,358
    Likes:
    213
    Location:
    OKLAHOMA USA
    I doubt that the USB will be the bottleneck.

    FWIW I like the designs that use an OP AMP instead of transitors. I think the OP AMP is a bit more rugged and easier to replace if you do blow it out.
  7. ericgibbs

    ericgibbs Well-Known Member Most Helpful Member

    Joined:
    Jan 4, 2007
    Messages:
    19,849
    Likes:
    521
    Location:
    England.
    ONLINE
  8. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Well my guess is, that not following the exact specifications won't make significant difference, not even noticeable. Maybe with everything by the book, an error caused by EMI may occur once in a year's time, and with slight modification, the chance may be higher as once in 8 month's time. But either way it's acceptable for common use.

    I've always liked more integrated solutions, as long as they don't raise much cost. It makes the design tight and more reliable, and yes, easier to repair.
  9. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Hi Eric, I'm reading it right now. It's quite informative, and I think I borrow a lot of ideas from this document. Thank you, I think with this example, I will be able to figure out a way to lay out a great PCB.

    Plus, I just came up with this idea, that I may attach a small PCB to the main board over the USB tracks, with intact single copper layer connected to GND on the main board. This might act as a shield and keep the USB tracks in an interlayer. I don't know if it will work the way I think it would, but I'm sure a single layer PCB or just a piece of metal piggy back is much, much cheaper than a four layer one with larger area.
  10. ericgibbs

    ericgibbs Well-Known Member Most Helpful Member

    Joined:
    Jan 4, 2007
    Messages:
    19,849
    Likes:
    521
    Location:
    England.
    ONLINE
    hi Alex,
    Sounds OK to me, the use of a grounded backplane is common method of improving the noise immunity of a pcb.
    You can purchase, thin pcb board, tin plated overall, if you go the small pcb add on route.

    Good luck.
  11. Alex_rcpilot

    Alex_rcpilot New Member

    Joined:
    Jan 30, 2004
    Messages:
    237
    Likes:
    0
    Location:
    Beijing, China
    Hi Eric, now I have a clue of what I should do. I'll carefully investigate the document you offered, and add a backplane to improve overall performance. You guys have been really helpful. Thank you.

Share This Page