This website uses cookies. By using this site, you consent to the use of cookies. For more information, please take a look at our Privacy Policy.
Home > FPGA Technical Tutorials > Designing with Xilinx FPGAs Using Vivado > Stacked Silicon Interconnect (SSI) > Pinout Considerations for SSI Designs

TABLE OF CONTENTS

Xilinx FPGA FPGA Forum

Pinout Considerations for SSI Designs

FONT SIZE : AAA

Pinout Considerations for SSI Designs  

Another important up-front consideration is the pinout selection. Whether using an  automatic or manual partitioning style, selection of which I/O pins are located in  which SLR has a substantial impact on the associated logic placement and routing  of the design. A well thought-out pinout selection will lead to good dataflfl ow through  the FPGA leading to good implementation results, providing better utilization, timing, and power. The best approach to determine a good pinout starts with examining  the expected dataflfl ow of all portions of the design and how that maps to the I/O  resources of the device. All associated control signals such as clocking, enables, and  resets should be considered in conjunction with this data flfl ow.   

The data path represented in Fig. 13.2 originates at SLR 0 (bottom SLR), must  go to the SLR 3 (top SLR) in order to buffer the data to an external memory, and  then travel back to SLR 1 to exit the device. This pinout selection has some obvious  drawbacks. First off, the data path is required to traverse six SLRs to complete the  data cycle. This could cause possible timing and resource issues. Also notice that  the clock and reset signals that must drive all of the logic are located in SLR3. Since

Simple.png

these are high fanout signals, a better selection would be in the center SLRs so that  the signals can be more evenly distributed improving the overall timing paths of  both signals. Placing the reset in the center would likely reduce the overall delay  from source to destination as the overall distance between these points is minimized. Moving the clock to a more centralized location will also reduce the insertion delay but, more importantly, it balances the overall clock skew as well. With the  clock placed in the top SLR, the data path entering the chip to the external memory  interface must travel against the clock. This increases the amount of negative clock  skew for that portion of the path resulting in reduced timing margin. The best placement for the clock has to do with what portion of the design is expected to have the  least timing margin and placing the clock so that it is either in the same SLR or the  clock travels with the data. Having the clock travel in the same direction as data will  improve setup timing margin by making use of useful (positive) skew.

Figure 13.3 shows the same data path with the memory interface moved to SLR1  and the clock and reset relocated to SLR0. The benefifi ts of this change should be  fairly evident. The overall SLR crossings reduced from 5 to 1. The high fanout reset  now only needs to reach two SLRs rather than all four and is located closer to the  logic it must drive. The clock is placed in SLR 0 so that the overall data path is  traveling with the clock rather than against it, promoting better skew characteristics  for this path. This pinout will likely consume less routing and fewer logic resources  as less pipelining and resource impacting optimizations like logic replication should  be necessary. Such simple pinout changes could have a dramatic impact on the over all implementation results of this design.

Often pinout decisions must weigh a balance between board layout considerations and internal dataflfl ow optimization. Due to the size of SSI device packages,  often there are numerous high-speed connections that must be routed out of a dense  ball grid array that pose its own diffifi culties in PCB routing and power delivery. This 

Alternative.png

often leads to an iterative approach between the digital design team and the printed  circuit board team trying to fifi nd the best compromise between internal dataflfl ow and  external PCB routability. Extra time and effort spent at this point can pay large dividends later in terms of easing design timing closure and reducing the overall design  implementation cycle while also requiring fewer device resources and less power  consumption as once pinout decisions are fifi xed, it is very diffifi cult to change later.


  • XC5VLX110-1FF1153C

    Manufacturer:Xilinx

  • FPGA Virtex-5 LX Family 110592 Cells 65nm Technology 1V 1153-Pin FCBGA
  • Product Categories: FPGAs (Field Programmable Gate Array)

    Lifecycle:Active Active

    RoHS: No RoHS

  • XC5VLX110-1FF676I

    Manufacturer:Xilinx

  • FPGA Virtex-5 LX Family 110592 Cells 65nm Technology 1V 676-Pin FCBGA
  • Product Categories: FPGAs (Field Programmable Gate Array)

    Lifecycle:Active Active

    RoHS: No RoHS

  • XC5VLX110-1FFG676C

    Manufacturer:Xilinx

  • FPGA Virtex-5 LX Family 110592 Cells 65nm Technology 1V 676-Pin FCBGA
  • Product Categories: Embedded - FPGAs (Field Programmable Gate Array)

    Lifecycle:Active Active

    RoHS:

  • XCV200E-7PQG240I

    Manufacturer:Xilinx

  • FPGA Virtex-E Family 63.504K Gates 5292 Cells 400MHz 0.18um Technology 1.8V 240-Pin PQFP
  • Product Categories:

    Lifecycle:Obsolete -

    RoHS:

  • XC3090-100PC84C

    Manufacturer:Xilinx

  • FPGA XC3000 Family 6K Gates 320 Cells 100MHz 5V 84-Pin PLCC
  • Product Categories:

    Lifecycle:Obsolete -

    RoHS: No RoHS

Need Help?

Support

If you have any questions about the product and related issues, Please contact us.