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.

TMS320C6747: Series resistors on the EMIFB and EMIFA bus of the OMAP-L137 EVM

Part Number: TMS320C6747
Other Parts Discussed in Thread: OMAP-L137

I'm designing a board with limited real estate. I reviewed the schematic for the OMAP-L137 EVM. Series resistors are in line with all the data and address lines. Our EMIFB will be running at 133MHZ, the EMIF A at 33MHz. Is this necessary or just a nice thing to have when you have the space?  Thank you, Tony

  • Hi Tony,

    We recommend series termination resistors mostly for clock outputs, especially when a pad loopback clock is used to latch incoming read data synchronously. Pad loopback refers to when a signal goes through the output buffer of the IO onto the board trace, and then loops back in through the input buffer of the IO.

    Let me check whether the EMIFA or EMIFB clocks use these pad loopback IOs and get back to you.

    The IO buffers have a certain signal propagation delay (that varies with process, temperature, and voltage). Control signals sent to the memory are similarly delayed by equivalent IO buffers. Since the memory reacts to these delayed signals and clock it provides its data slightly later than the non-delayed internal clock. So by using the delayed pad loopback clock to latch incoming data, setup time is increased allowing the peripheral to close timings at higher frequencies.

    The problem with pad loopback signals is that they are susceptible to signal reflections from the PCB traces and terminations. Installing a series resistor close to the IO buffer driving these signals absorbs most of that reflection, preventing it from arriving at the input buffer. If the reflections were allowed to appear at the input buffer side of the pad-loopback IO, it can appear as glitches or overclock the peripheral (if its the clock signal).

    A signal may also ring as the reflections bounce from source to destination and back. This ringing can cause glitches on the address or data lines. With a series termination resistor, the ringing is quickly attenuated. A value of 22 ohms is typical of series term resistors, and it is typically placed near the source driver.

    At a minimum, place series term resistors near the clock outputs, and any other pad-loopback signals.

    If you have any kind of multi-destination routing, it is recommended to terminate the signals at the end of the transmission line.

    Control signals like CSn, WEn, OEn should have pull-up resistors to control the state of these signals when the drivers are high impedance - like when the device is held in reset or pin muxed to another peripheral (if possible). In particular, pulling CSn high is most critical.

    Consider using resistor packs to save board area instead if individual resistors. Follow general routing practices like keeping adequate trace separation between clock traces and other traces to prevent crosstalk/ coupling.

    Hope this helps,
    Mark
  • Hi Mark, Thank you for looking into this matter. I will proceed with the control signals and clock recommendations. I'll wait to see what you find out about the PAD  loop back I/Os before adding series resistors to all the data and address lines of the EMIFA and EMIFB.

    Thank you, Tony

  • Hi Tony,

    As far as I can tell, the EMIFA and EMIFB on C6747/ OMAP-L137 do not utilize pad-loopback clocks. Its still a good idea to put a series term on the clock trace close to the C6747 pin to cut down on ringing and reflections.

    I suspect the abundance of series termination resistors is to provide debugging access to all of the signals.

    In your design, does the EMIFx_CLK go to multiple devices or just to one endpoint?

    You can always simulate the effect of the series term and the PCB trace impedance with available IBIS models.

    See this E2E post about having pull-downs on the EMIF address lines.

    e2e.ti.com/.../39480
    - OMAP-L137 & C6747, there are not enough address pins, and the ROM bootloader does not configure or drive the GPIOs. This is why you need pulldowns during boot for this device.

    Regards,
    Mark
  • Hi Mark,

    Thanks for following up on the series terminations and the IOPAD loop back. The EMIFx clocks go to one device each.

    I will use series terms for control and clocks, That was a good recommendation abot the pull downs for the extra address lines on the EMIFA boot process.

    Regards, Tony