2394723_en-US

キャンセル
次の結果を表示 
表示  限定  | 次の代わりに検索 
もしかして: 

2394723_en-US

2394723_en-US

Regarding 8M Plus Real time Accuracy

Hi,  We would like to lean more about 8M Plus capabilities regarding its real-time applications. Are there any test reports or related materials available concerning TSN or real-time performance, or jitter?

i.MX 8M | i.MX 8M Mini | i.MX 8M NanoRe: Regarding 8M Plus Real time Accuracy
Area Available material What it contains
Real-time CPU/RTOS latency Harpoon User’s Guide Measured real-time latencies on i.MX 8M Plus / Zephyr , including IRQ latency and task latency in ns. Example results: no-load IRQ latency min/avg/max/stddev = 625 / 796 / 11,125 / 1,798 ns ; task latency = 2,583 / 2,671 / 13,041 / 6,045 ns . Under Linux CPU + memory load, IRQ latency = 625 / 798 / 4,250 / 4,674 ns , and task latency = 2,583 / 2,670 / 14,333 / 10,407 ns .
Real-time benchmark method Harpoon User’s Guide — rt latency application Defines the benchmark as the time delta between hardware IRQ events and software actions, measured with a hardware timer and sub-microsecond precision.
TSN capability i.MX 8M Plus product / reference material i.MX 8M Plus includes dual Gb Ethernet, with one Ethernet supporting TSN, and uses the integrated 800 MHz Arm Cortex-M7 for industrial real-time control.
TSN hardware standards i.MX 8M Plus Reference Manual TSN support includes IEEE 802.1Qbv Time-Aware Shaper , 802.1Qav Credit-Based Shaper , IEEE 1588v2 PTP , and the Ethernet block implements 802.1Qbv-2015 , 802.3br , and 802.1Qbu frame preemption-related TSN functions.
TSN test / validation environment
Real-Time Edge User Guide
Describes a TSN test environment for evaluating i.MX 8M Plus TSN capabilities, including traffic generation/analysis and monitoring of latency, jitter, and synchronization accuracy.
TSN jitter / latency example Real-Time Edge User Guide — TSN endpoint sample app Provides TSN endpoint statistics including traffic latency min/mean/max and notes latency around 503 µs with latency jitter around 300 ns in the shown example.
TSN application demo AN13588 Demonstrates a GenAVB/TSN real-time control application. It describes a 2 ms cycle , a 400 µs reserved/guaranteed control-traffic window , and a statistics thread for scheduling, processing timing, traffic correctness, and latency.
TSN 802.1Qbv demo AN13995 Demonstrates TSN 802.1Qbv using i.MX 8M Plus and explains how time-aware shaping uses fixed repeating cycles to provide deterministic latency; it also includes Linux tc / taprio configuration examples.
 
Re: Regarding 8M Plus Real time Accuracy

@yipingwang 

Thank you for providing the information. Could you also please provide us with information about the iMX8M Plus EVK?

Thanks.

Re: Regarding 8M Plus Real time Accuracy
We understand that no public "real-time performance report" has been officially released for i.MX95 EVK. However, NXP does internally perform real-time benchmarking on i.MX95 platforms. Internal benchmark documents indicate that cyclictest and EtherCAT performance evaluations have been executed on i.MX95 LPDDR5 EVKs running Real-Time Edge software with PREEMPT_RT Linux. Reported examples include a maximum cyclictest latency of approximately 38 µs during a 6-hour stress-ng test and EtherCAT filtered maximum jitter of approximately 12 µs under the documented test conditions. Because real-time performance depends strongly on BSP version, kernel configuration, CPU isolation, workload, and network traffic, these values should be considered reference measurements rather than guaranteed application-level limits.
 
as supported benchmarking platforms and provide detailed cyclictest, stress, and rt_latency procedures.
Re: Regarding 8M Plus Real time Accuracy

Our distributor informed us that there is currently no official performance report available for the NXP i.MX95 EVK regarding real-time performance or jitter.

However, they also provided documentation describing the testing methodology (e.g., cyclictest) for evaluating real-time latency.

This leads to some confusion on our side. Since a standardized testing methodology exists, we assume that such tests must have been performed internally—at least on the reference EVK platform.

Therefore, we would like to clarify:

  • Has NXP conducted any internal measurements of real-time performance (e.g., latency, jitter) on the i.MX95 EVK?
  • If so, are there any reference or baseline results that could be shared?

We understand that real-time performance may vary depending on system configuration and workload. However, even a baseline result under controlled conditions (e.g., default BSP, minimal load) would be very helpful for initial evaluation.

Thank you for your support.

hankwang_0-1784018178178.png


Re: Regarding 8M Plus Real time Accuracy
  1. REALTIMEEDGEUG (Real-Time Edge Software User Guide) (https://www.nxp.com/docs/en/user-guide/REALTIMEEDGEUG.pdf)

. Real-Time Edge Software (most relevant)

NXP's Real-Time Edge Software officially supports the i.MX8M Plus EVK and includes:

  • PREEMPT_RT Linux
  • TSN stack
  • IEEE 802.1AS (gPTP) synchronization
  • TSN traffic shaping and scheduling
  • EtherCAT, OPC-UA, CAN-related industrial protocols
  • Heterogeneous real-time operation using Cortex-A53 + Cortex-M7

The internal document REALTIMEEDGEUG states that Real-Time Edge Software provides:

  • Real-time networking (TSN)
  • Real-time Linux (PREEMPT_RT)
  • Pure RTOS/Bare-metal options
  • Jailhouse partitioning
  • Industrial protocol support

Support for i.MX 8M Plus LPDDR4 EVK


  1. NXP Application Note AN13995 – TSN 802.1Qbv Demonstration using i.MX 8M Plus
  2. The Real-Time Edge Six Pack materials describing PREEMPT_RT + TSN support
Re: Regarding 8M Plus Real time Accuracy

@yipingwangThank you very much for the information you provided.

タグ(1)
評価なし
バージョン履歴
最終更新日:
昨日
更新者: