The most likely case is that there are some noise sources or random number sequences that are not starting with exactly the same values for each run. In fact, it is good practice to perform multiple simulations and take the average BER over them. There is more information on BER estimation and measuring at the following link:
If reproducing the exact BER after each run is necessary, you can set the noise to be exactly the same for each run by forcing the pseudo random number generators to start with the same seed. Here’s how:
- In each of all the components that generate pseudo random sequences.
- Go to the Random numbers tab
- Uncheck Generate random seed
The Generate random seed option is enabled by default so as to better simulate a truly random situation.
If the BER Test Set component is being used to measure the bit error rate, please remember that longer bit sequences will produce more accurate results. The Sequence length parameter can found in the Main tab of the BER Test Set.