If the effects of soft errors are a concern for your product's deployment, each component in the system design usually needs a FIT budget. To calculate the FIT for an AMD FPGA, use the SEU FIT rate estimator available in the SEU lounge and see the XAPP472 SEU Estimator on how to use the FIT rate estimator.
To calculate FIT for a deployment, at the minimum you need:
- Target device
- Estimated number of the device to be deployed
Besides providing the estimated FIT for the device, the estimator also predicts the number of soft errors expected for a given time frame.
By using an implemented design, a more accurate estimation of the FIT is achieved. This is possible by entering the number of block RAM used, whether the block RAM ECC feature is employed to detect and correct soft errors, and the percentage of essential bits in the design. Essential bits are defined as configuration RAM bits that are used to define the function on the FPGA. If an essential bit is changed unintentionally by a soft error, it is possible that the function in the FPGA does not behave as intended.
On the other hand, if a non-essential bit is changed, there is no impact to the function. The following steps determine the percentage of essential bits in a design:
- Set the following property in Vivado:
set_property bitstream.seu.essentialbits yes [current_design]
- Regenerate the bitstream for your design.
- The essential bits percentage is printed in the Vivado Tcl console and consequently the Vivado log while the bitstream and essential bits data are being generated. Here is an example:
Writing bitstream ./sem_ultrap_v3_1_example.bit...
Creating bitstream...
Writing bitstream ./sem_ultrap_v3_1_example.ebc...
Creating essential bits data...
This design has 707717 essential bits out of 143015456 total (0.49%).
After a FIT estimation has been completed, it must be checked to determine if the FIT requirement for the deployment is met. Other design approaches might need to be implemented to reduce the FIT.
If there is not a FIT target, then it is unclear what design changes (and trade-offs that come with them) need to be made to mitigate the soft error effects, including whether a deployment benefits from using the SEM IP. If that is the case, it is important to identify the benefits obtained using SEM IP before integrating it.