• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 84 subscribers
  • Views 2594 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

F28335 DMA Pipeline with XINTF

Hiroshi Nomura
Prodigy 150 points
Other Parts Discussed in Thread: TMS320F28335

Hi, Experts,

I'd like to know how the F28335's DMA pipeline (illustrated in Figure 3 on SPRUFB8D) changes when DMA source and/or destination is XINTF.

For instance, with a XINTF zone configured with Lead/Active/Tail timing being 1/3/1 (XTIMCLK=SYSCLK), and use this as DMA source, then "Read SRC data" cycle in the diagram (which takes 2 cycles on the fiture) replaces 5 cycles?   Or it becomes 6 cycles?    How about the case when such XINTF is set to DMA destination? 

 

 

  • 0
    TI__Expert 8500 points

    The DMA baseline is 4 cycles per transfer. Each waitstate on the XINTF (L/A/T) will add a cycle. So, for the 1/3/1 timing above the DMA will take 9 cycles to transfer.

    Regards,
    Dave Foley

     

  • 0 in reply to David Foley
    Prodigy 150 points

    Thank you.  This clarification is highly appreciated. 

  • 0 in reply to Hiroshi Nomura
    Expert 1725 points

    Hi communiTI!

    I would like to use this thread for another question, regarding the TMS320F28335 DMA pipeline specifications.

    There's said that the DMA needs 4 cycles for a single read, than I suppose that the pipeline creates a delay when reading from memory or from an external interface.

    With Pipeline I would think that there's a pipeline delay of 4 cycles and that (burst-) reads need 1 clockcycle, with the result delayed clock 4-cycles. Just like a pipeline-ADC. I would think that when DMA reads 16 words, that this would take 16Read + 4PipelineDelay cycles.

    On a pipeline ADC the sampleresults are delayed by a pipeline because of its architecture, but it's a continuous stream equal to the ADC sys-clock rate.

    Could someone please clear this up?

    Best regards,

    Tjarco Boerkoel

  • 0 in reply to David Foley
    Prodigy 150 points

    Dear Experts,

    I want to clarify something. So even if we use DMA, we still need to assign the minimum required wait cycles for the external interface XINTF, the L/A/T, right? So this makes it longer to read external memory using DMA, right?

    Thanks