• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 31 subscribers
  • Views 957 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

TMS570LS3137 debug connection issue

ilhan taygurt
Prodigy 235 points
Other Parts Discussed in Thread: TMS570LS3137

Hi All dear my colleagues,

In our new design we will use TMS570LS3137 with BGA package (ZWT) and we want to use ETM feature.

For this purpose is it possible to use 20-pin ARM 20-Pin Cortex Debug + ETM connector? 

If yes how do i connect the following signals?;

*pin#6 SWO / TDO / EXTa / TRACECTL

*pin#12 TRACECLK

*pin#14 TRACEDATA[0]

*pin#16 TRACEDATA[1]

*pin#18 TRACEDATA[2]

*pin#20 TRACEDATA[3]

i mean which connector pin to where on Hercules?

There are 32-bit ETM data on chip and if 20-pin Jtag with ETM is possible to use which ETM data pins should be used for this 4-bit TRACEDATA?

If available reference schematic would be perfect.

Best regards

  • 0
    TI__Guru 68950 points
    Theoretically maybe but practically no.
    There are 8 dedicated trace pins on the device and you really want more than 8 if at all possible.. Otherwise you'll get FIFO overflows in your trace listing. This means you'll see blocks/areas where due to the on chip ETM fifo getting filled - there are some 'black-out' areas where you lose trace for a while until you re-sync.

    Because the trace port frequency isn't as fast as the CPU's (close to 1/8) even though it's DDR you are in a situation where you want to use all the trace pins possible. There are 32 data pins available ... 8 dedicated the rest muxed'. The PROTRACE can only capture 19 data lines though.

    The more you try to trace the more data lines you need. Instruction trace compresses pretty well except for cases like "B SELF" believe it or not.. but if you start emitting cycle accurate trace info the bandwidth requirement goes up. Then if you want to use the most powerful feature (IMO) which is DATA trace [yes you can see the values the CPU writes to memory which is sooooo helpful] the bandwidth really goes up because this doesn't compress well at all....

    I would use the HDK design as an example, and use the 60-pin MIPI standard debug header.
  • 0 in reply to Anthony F. Seely
    Prodigy 235 points
    Dear Mr. Seely,

    Thank you for your compherensive explanation.
    Accordingly it seems I need to find another way.

    If you have could you pls send me tutorial&app notes which is absent in the internet but at the same time unrestricted? I think there is no any detailed info in the Hercules published docs.

    Regards.
  • 0 in reply to ilhan taygurt
    TI__Guru 68950 points

    Hi Ilhan,

    Sorry I can't really point you to any information like this.   We don't have anything published and we haven't done any sort of 'benchmark' to characterize trace bandwidth required versus application conditions.

    It's more of the other way  -- you have to get the trace box,  give as  many pins as possible, and do emperical testing with your code.  Because bandwidth depends on the amount of trace information generated by the code execution and how well this compresses.

    Also when you have bandwidth issues you can restrict / limit tracing to certain ranges as another way to get the information you need.

    There isn't any real clear 'thesis' explaining all these aspects unfortunately.