(for calculating BER confidence levels, as well as determining how many bits to measure...)
In the lab, we don't need to know the true BER of our system. We simply need to measure enough data to have some confidence that the BER is lower than some specified level. The question then becomes, if we repeatedly transmit N bits, and detect E errors, what percentage of the tests will the measured BER (=E/N) be less than some specified BER
(BERS)? We call this percentage the BER confidence level (CL x 100), and calculate it using the Poisson distribution as follows.
In other words, CL x 100 is the percent confidence that the system's true BER (i.e. if N = ) is less than the specified BER (BERS). That is, if the measurement is repeated an infinite number of times, the measured BER will be less (that is, better) than the specified BER for CL x 100 percent of the tests.
Since we cannot measure for an infinite length of time, the BER confidence level is always less than 100% (at least theoretically). Before starting a BER measurement, one must identify a target confidence level. Some industry standards specify this level (many do not) but 95% is a good target. All industry standards specify a maximum system BER (BERS).
Use the calculator below to determine the confidence level for a BER lab measurement by entering the specified BER, the data rate, the measurement time, and the number of detected errors. For reference, the number of transmitted bits (N) is shown as the data rate (BPS) multiplied by the measurement time (T).
Alternatively, one can determine how many bits must be measured in the lab (that is, how much time is required to measure data) to achieve a specific confidence level, assuming a certain number of errors (usually, 0 errors) -- simply enter BERS, BPS, and E, then change T until the desired confidence level is found.
Enter numbers below as integers, or use scientific notation (for example, enter 123 as 123, 1.23e2, or 1.23E2).