The repetitive algorithm is similar to the INCOS algorithm; the application calculates a local noise value for each data point using the same method that it uses for the INCOS algorithm.
The application saves all local noise/intensity value pairs in an array. The application does not accept nor decline a local noise event until all noise events are recorded (unlike INCOS, where only events whose noise values are smaller than a value that is 8 times the square root of the minimum of the current three scans are accepted as noise).
The starting value for the noise is the average of all local noise events.
In each iteration (a maximum of 10), the local noise values of the intensities that are stronger than the average intensity of all remaining noise events are discarded, and a new average noise is calculated. If the new average noise is smaller than the value of the old noise divided by 1.1, the application performs another iteration; if not (or if there were already 10 iterations), the process is finished.
In a final step, the application discards the third of the remaining values that are the most intense and the third of the values that are the least intense. The application then uses the remaining values to calculate the noise.