• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 101 subscribers
  • Views 1017 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.

Performance of ARM3359 PRU in Ethercat configuration

Luca Melandri
Prodigy 100 points
Other Parts Discussed in Thread: AM3359, TLK110

Hi,

I'd like to have more information about performances in Ethercat implementation with ARM3359.

I noted that the PRUs are 200MHz RISC processors and I'd like to have comparison, if it's possible, with hardware Asic ET1100 of Beckhoff.

One of the information I'm interested about, is the time tMM (MII PORT to MII port delay) that is 335ns in worst case of data to be processed in ET1100 (page III-93 of the Et1100 Hardware Description databook)

Which is this time in ARM3359? 

Thanks 

Luca

  • 0
    TI__Intellectual 1835 points

    We have measured wire-to-wire latency. That means, the numbers below include the PHY latency (the delay between between the Ethernet cable and MII interface) because the test equipment to measure latency is connected on the wire and not on the MII interface.

    For EtherCAT processing path, we see 620ns average latency.

    For EtherCAt auto-forwarding path, we see 610ns average latency.

    In our measurements, this is inline with ET1100 once PHY delays are added to the published delay data. 

    The results above are from measurements done in last week of May 2012. 

    Thanks.

  • 0 in reply to Maneesh
    TI__Mastermind 30110 points

    Hi Maneesh,

    Is there a way with the Industrial SDK and the IDK/ICE boards to do some latency testing (like is there a basic example available for this purpose)?
    What equipment did you use for the wire to wire latency? It might be useful to publish a little bit more on the test bench used.

    What type of latency measurement are possible to measure just using the SDK and IDK/ICE boards (without using custom code, without the measurement equipment you mentionned)?
    Ideally customer would like to do the testing and compare it with the latency published in the ET11xx hardware datasheet:
    http://www.beckhoff.com/english.asp?download/ethercat_development_products.htm

    Thanks and best regards,

    Anthony

  • 0 in reply to Maneesh
    Guru 20205 points

    Hi Maneesh,

    I would like to get latency for only PRU processing because our customer uses only PRU firmware which is included in Industrial SDK with 3rd party's software.

    Best regards,

    Daisuke

     

  • 0 in reply to Daisuke Maeda
    Guru 20205 points

    Daisuke Maeda said:

    I would like to get latency for only PRU processing because our customer uses only PRU firmware which is included in Industrial SDK with 3rd party's software.

    For latency in PRU processing, is there the information?

     

  • 0 in reply to Maneesh
    Intellectual 595 points

    Hi Maneesh

    I read your reply about the wire-to-wire latency..

    but I still have a few question about your test environment.. because I want to measure etherCAT(AM3359 ICE) throughput ..

    above post you wrote about ' between Ethernet cable and Mii Interface' but I Can't fully understand " From where to where"

    please let me know your test environment and test method(Processing path and auto-forwarding path)

    thanks in advance

    Moonseon

  • 0 in reply to Moonsun Park
    TI__Mastermind 47541 points

    Hi,

    I recommend to use Industrial SDK  release1.0.0.8 (http://software-dl.ti.com/dsps/dsps_public_sw/am_bu/AM335x_SYSBIOS_Industrial_SDK/latest/index_FDS.html) for any benchmarking related to process path /auto-forward path forwarding latency as there are optimizations in firmware to improve the same.

    MII to MII latency configured by PRU firmware process path latency: 340ns auto-forward path latency : 280ns

    Above measurements are done using hilscher netanalyzer (passive network sniffer) by measuring the time difference between incoming packet time stamp (master request) and outgoing packet time stamp (slave response). This includes PHY latency (~ 280ns for TLK110) as well on top of MII latency .