SEU Mitigation User Guide: Agilex™ 5 FPGAs and SoCs

ID 813649
Date 9/20/2024
Public
Document Table of Contents

2.7. Retrieving SEU Statistics using Mailbox Command

To retrieve SEU statistics (timing of SEU event) in Agilex™ 5 devices, send the mailbox command with the following syntax:
READ_SEU_STATS (0x40)

Examples of the READ_SEU_STATS (0x40) Mailbox Command:

To retrieve the SEU statistics via the Configuration Debugger tool, you can send 0x40 as the Command Code and 0x30000 in the Parameter List field.

0x30000 refers to the sector 0x3, which is the same sector as the injected SEU error.

Figure 7.  READ_SEU_STATS (0x40) Mailbox Command in the Configuration Debugger Tool

The SEU statistics are measured in cycles. To obtain the timing of SEU event, use the following equation to convert the number of cycles into milliseconds (ms):

Figure 8. Timing of SEU Event Equation
The internal clock frequency refers to the Configuration Clock Source in the General tab of the Device and Pin Options. Altera recommends that you use external configuration clock source (OSC_CLK_1) for better accuracy.
Configuration Clock Source Internal Clock Frequency
Internal oscillator 160–230 MHz
External configuration clock source (OSC_CLK_1)

200 MHz for low-powered devices.

OPN with core speed 6 and power S/X.

For example, A5Exxxxxxxxxx6Sxx and A5Exxxxxxxxxx6Xxx

250 MHz for other Agilex™ 5 devices.

For example:

The T_seu_cycle is 0x000B991C in hexadecimal and it is equals to 0x760,092 in decimal. To convert 760,092 cycles into milliseconds at a frequency of 200 MHz, the timing of SEU event obtained is as follows:

T_seu_cycle = 760,092 cycles / 200 MHz = 3.80046 ms

Table 6.  SEU Statistics
SEU Statistic Description
T_seu_cycle The time between the start of the most recent SEU detection cycle to the previous cycle.
T_seu_detect The most recent time spent in performing the SEU detection within a SEU cycle.
T_seu_correct

The most recent time spent in performing the SEU correction (internal scrubbing) within a SEU cycle if a correctable error is detected.

0x FFFFFFFF if no correction is done.

T_seu_inject_detect

The most recent duration from the SEU error being injected to the time the SEU is detected.

0x FFFFFFFF if no injection is done.

T_sdm_seu_poll_interval The time between the start of the most recent SDM check cnoc_return_interrupt line (if there is an SEU error) to the previous one. Use T_sdm_seu_poll_interval to measure the SDM SEU error polls interval.
T_sdm_seu_pin_toggle_overhead The time for SDM to toggle the SEU_ERROR pin after SDM detects error when polling the SEU error (check_cnoc_return_interrupt).
  • To obtain SEU Inject to Detect Cycle (best case), inject a safe SEU error by sending this mailbox command: INSERT_SAFE_SEU_ERROR [0x41] with the T=01 option (inject before next detection cycle), followed by sending READ_SEU_STATS [0x40].
  • To obtain SEU Inject to Detect Cycle (worst case), send the INSERT_SAFE_SEU_ERROR [0x41] mailbox command with the T=10 option (inject after next detection/correction cycle).